Prepared-aggregator facades built via __new__ lack _agent. Accessing
self._agent raised inside the planner try and bool-coercion of a missing
snapshot forced False, suppressing config fallback for cache_ttl=off.
Pass a tri-state value and add a no-agent/config-off regression.
Blank SimpleNamespace stubs used by MoA decoration and
plan_cache_sections_for_destination never set _cache_disabled, so
anthropic_prompt_cache_policy re-injected cache_control markers after
operators turned caching off. Stamp the disable onto those stubs from
an explicit flag or the live config, and pass the agent flag from the
MoA aggregator path.
Fixes#76085
Hermes installs runtimes for itself — `uv` at `$HERMES_HOME/bin/uv`, Node
at `$HERMES_HOME/node` — and neither directory is on an arbitrary
process's PATH. Every `shutil.which("node"/"npm"/"npx"/"uv")` in Hermes's
own code therefore has two failure modes: the managed runtime is invisible,
so the caller reports "not installed" or degrades to a slower tier on a
machine that has exactly what it needed; and when a system copy also
exists, the one Hermes does not own wins.
Routed the Hermes-owned call sites through managed-aware resolvers:
- `agent/lsp/install.py`, `hermes_cli/dep_ensure.py`, `hermes_cli/main.py`
(`_make_tui_argv`), `hermes_cli/tools_config.py` (`_run_post_setup`) now
use `find_node_executable()`.
- `hermes_cli/tools_config.py::_pip_install` and `hermes_cli/setup.py`'s
vercel install use `ensure_uv()` (installing uv is in scope during setup,
and the Windows installer's `uv venv` does not seed pip, so the fallback
tier is "No module named pip"). `tools/lazy_deps.py` uses `resolve_uv()`
— a lookup, not a bootstrap, because it runs mid-turn for an optional
dependency and downloading a runtime as a side effect exceeds what the
caller asked for.
- `hermes_cli/gateway.py`: extracted `_append_node_dir_for_service()`,
shared by the systemd unit and launchd plist generators, which appends
the managed dirs before the PATH-resolved one. A service definition is
written once and survives reboots, so resolving a system Node that
happens to lead the installing shell's PATH bakes the wrong interpreter
in permanently. Managed dirs are profile-scoped, so each profile's unit
still names its own Node; the existing symlink-parent rule (don't
`.resolve()`) is preserved verbatim.
- `tools/environments/local.py`: the terminal tool's subshell PATH gains
the managed dirs, appended alongside the sane entries rather than
prepended — a tool the user deliberately put on their own PATH still
wins, and the managed one only fills a gap. This is also what makes the
bare `which("uv")` in `tools/env_probe.py` correct: that probe reports
the environment the *model* sees, and the model can only run what is on
that subshell's PATH.
`scripts/install.ps1`: the persisted User PATH update becomes
`Set-ManagedNodeFirstOnUserPath`, a move-to-front rather than an
add-if-missing. Installs made by an older install.ps1 already have the
managed dir in User PATH — at the tail, behind a system Node — and an
add-if-missing check sees it present and leaves that ordering in place
forever, so the users the bug hurt would never be repaired. Unrelated
entries keep their relative order (empty segments included; a trailing
`;` is legal and the installer's other PATH code preserves them),
duplicates collapse, and it writes only when the string actually changes.
Tests:
- `tests/test_managed_runtime_resolution.py` — AST guard that fails any
new bare `which()` for a managed runtime, with a short justified
allow-list and a companion test that fails when an allow-list entry goes
stale. Reading source is banned by AGENTS.md and this is the documented
exception: the property is "no call site anywhere spells it this way",
which no runtime seam can observe.
- `scripts/ci/test_install_ps1_path_migration.ps1` — behavioral, not a
source regex: it lifts the real `Set-ManagedNodeFirstOnUserPath` out of
install.ps1's AST and rewrites only the two registry calls into an
in-memory store, so the shipped split/dedupe/prepend/change-detection
logic executes for real. Not in the default lane (Linux runners have no
PowerShell host); runs under `pwsh`. 13/13 assertions pass.
The sequential executor's KeyboardInterrupt handlers emitted a cancelled
post-tool-call event for the current tool, called agent.interrupt(), then
re-raised — WITHOUT appending a tool result message for the interrupted call
or any remaining calls in the batch. The assistant tool-call turn was left
with no matching tool results, a message-role alternation violation that
malforms the next provider request (relying on downstream repair passes to
patch it, which don't run on every path).
The cooperative-interrupt block (_interrupt_requested) and the concurrent
executor already emit a result for every call_id; this brings the two hard-
interrupt handlers into line via a shared _append_cancelled_tool_results
helper that appends a cancelled result for the current + remaining calls
before re-raising.
Verified live before/after (0 tool results -> 3 for a 3-call batch
interrupted on the first tool) and with a sabotage-checked regression test.
52 interrupt/executor tests pass.
Salvaged from #51604 (@JoaoMarcos44, issue #51603): resolve_anthropic_token()
and run_oauth_setup_token() in agent/anthropic_adapter.py read
ANTHROPIC_TOKEN / CLAUDE_CODE_OAUTH_TOKEN / ANTHROPIC_API_KEY via bare
os.getenv(), bypassing agent.secret_scope — a cross-profile over-read in
multiplex mode. Every other provider routes through
runtime_provider._getenv -> get_secret; the adapter now does the same via
a local _getenv wrapper (identical to os.getenv when multiplexing is off,
scope-authoritative + fail-closed when on).
Dropped from the original PR: the cron scheduler hunks (superseded by
fdab380a1a which installs the per-job profile scope) and the unrelated
hermes_logging Windows hunk (scope creep).
Includes the PR's RED->GREEN scope-isolation test file (6 tests).
The parallel tool-batch planner treated search_files as unconditionally
parallel-safe (_PARALLEL_SAFE_TOOLS) with no path reservation, so a
batch of patch(path=X) + search_files(path=dir(X)) landed in one
concurrent segment and the search could observe pre-mutation file
content — a same-block write->read stale-read race.
Fix the class, not the site: path-scoped reservations now carry a
reader/writer role.
- search_files joins _PATH_SCOPED_TOOLS as a READER, reserving its
search root (default '.', matching the tool's default) instead of
bypassing path checks entirely.
- Overlap only conflicts when a WRITER is on either side: a write into
a searched/read subtree splits segments (ordered behind the write),
while reader<->reader overlap — previously split needlessly — now
stays parallel (concurrent reads commute).
- write_file/patch keep their existing writer barrier semantics.
Prior art surveyed for this design: Codex CLI's RwLock read/write
barrier (readers share, writers exclusive), Claude Code's
isConcurrencySafe partitioning, and gemini-cli's contiguous
parallelizable batching — all converge on reader-shared/writer-
exclusive with contiguous-order preservation, which this planner
already had for read_file/write_file/patch; this closes the
search_files gap and adds the missing reader/reader concession.
Verified by sabotage run (tests fail against the old planner) and an
E2E script exercising the real planner + real file I/O.
- Reuse telemetry['middle_window_tokens'] for the skip's middle estimate
(is-None fallback to a fresh estimate) so log and telemetry agree
- Declare prellm_skip_count in the base telemetry schema (fixed shape)
- Defer _derive_auto_focus_topic into the non-skip branch (user-turn scan
was wasted work on every skip)
- Document the skip in compress()'s Algorithm list and force: arg doc
- Drop dead call_llm patches from 7 tests (unreachable with
_generate_summary mocked)
When the middle section is < 10% of threshold tokens AND at least one prior
real-usage ineffectiveness strike has been recorded, skip the expensive LLM
summarization call and fall through to the deterministic message-dropping
path. Without this guard, a tool-heavy session where the protected tail
already holds most of the tokens can burn 500+ seconds on a summary call
that replaces a few lightweight messages with negligible token savings.
Key design decisions per GottZ review on PR #60451:
1. Separate _prellm_skip_count counter — never increments
_ineffective_compression_count (the strike counter that latches at >=2
to disable compression entirely). One real strike + one skip must NOT
permanently lock out compression until /new.
2. feasibility_skip sentinel flag — exempts skips from the abort branch
(abort_on_summary_failure / _last_summary_auth_failure /
_last_summary_network_failure). A stale failure flag from a prior
cycle must not turn a deliberate skip into a full abort.
3. reason=None for feasibility-skip fallbacks — a stale _last_summary_error
from an earlier real failure must not be embedded into the skip's
deterministic fallback marker.
4. info-level logging for feasibility-skip fallbacks (not warning) — this
is an intentional optimization, not a failure.
Skipped when force=True (manual /compress) so auth/error handling paths
are always exercised on explicit user request.
Adds 6 regression tests (TestPreLlmFeasibilityCheck) covering:
- Strike counter isolation
- Stale auth/network failure flag immunity
- force=True bypass
- No-skip when no prior strikes
- Counter reset on session reset
Co-authored-by: Cursor <cursoragent@cursor.com>
Co-authored-by: TRON <tron-agent@agentmail.to>
_CFG_DOTTED_RE's nested quantifier (?:[A-Za-z0-9_\-]+\.)+ backtracks
exponentially on long non-matching dotted runs (doubles every ~4
segments). Flatten it and use possessive quantifiers (py3.11+) in
_CFG_DOTTED_RE and _YAML_ASSIGN_RE wherever the successor is disjoint.
Zero behavior change: equivalence fuzz-verified over 120k structured
and random inputs comparing full sub() output including groups. Adds a
ReDoS regression test.
_call_fallback_candidate_sync replans messages/tools for each resolved
destination, but its async mirror still shipped the caller's decorated
sections verbatim — the primary destination's markers (including a
direct-native tool marker) leaked to fallback candidates with different
cache contracts, and the relay saw the display label instead of the
resolved provider/api_mode. Mirror the sync path: resolve the
destination, replan both sections, thread provider/api_mode through
_relay_async_completion, and replan again for the auth-refresh retry
client. Mutation-checked: the new parity test fails on the verbatim
pass-through shape.
Follow-up to #76032 (#20880).
The count walked every message part and tool schema on every request but
is consumed only by tests. Compute it on demand via a property instead.
Follow-up to #76032 (#20880).
Three copies of the same logic landed with #76032:
- MoA's _call_prepared_aggregator and auxiliary_client's
_replan_synchronous_cache_sections both implemented stub → policy →
strip → plan for a resolved destination. Extract
plan_cache_sections_for_destination() into agent_runtime_helpers (which
already owns the policy functions) and route both through it. Also
removes a redundant full-transcript deepcopy+strip per request (the
caller pre-stripped what build_prompt_cache_plan strips again).
- The fallback_chain[N] label regex + chain-entry lookup lived in
_fallback_entry_timeout AND _fallback_destination. Extract
_fallback_chain_entry() and reuse.
MoA's cache-plan failure log is promoted debug → warning: the call-block
site skips MoA, so this block is the aggregator's only decoration path —
a silent failure ships an undecorated request (the 0%-cache MoA bug class).
Behavior-preserving; 195 targeted tests green.
Follow-up to #76032 (#20880).
_apply_static_prefix_marker duplicated _apply_system_cache_markers' split
logic minus its empty-suffix guard: when the stored system prompt equals
the static prefix exactly, the tool-cache plan emitted a two-part split
with a trailing empty text block — HTTP 400 on native Anthropic. Fold the
tool-cache layout into the existing helper via mark_suffix /
fallback_to_whole flags; the empty-suffix case now marks the prompt as
one whole block. Behavior-parity verified against the merged planner for
every non-empty-suffix shape.
Follow-up to #76032 (#20880).
Review pass 2 (reuse reviewer HIGH): the step-3b probe-down fallback for
custom/local endpoints returns the same silent 256K default but only
logged at INFO - invisible by default, and it is the MORE common path
for small local models (the exact users the warning exists for).
Extract _warn_context_length_fallback() (deduped per model+base_url)
and call it from both fallback sites, per the fix-the-whole-bug-class
rule. Regression test drives the custom-endpoint path and fails without
the widening (mutation-checked).
Review follow-up:
- Warn once per (model, base_url) at the step-9 fallback via a module-level
dedup set (established _WARNED_* idiom). The fallback result is
deliberately never cached, so the un-deduped warning fired on every
resolution - e.g. once per gateway message via the session-hygiene path.
- Replace the three inline-mock pool-cleanup tests (which reproduced the
try/except block against a MagicMock and passed even with the production
code reverted) with a parametrized test that drives the real
BatchRunner.run() with a patched Pool; drop the CPython stdlib
signature change-detector test.
- Add a once-per-model warning regression test; clean up dead imports.
All tests verified to fail against pre-PR batch_runner.py/model_metadata.py
and pass with the fix (mutation check).
Salvage of #6629 by aaronlab (kshitijk4poor reworked against current main).
Three concerns from the original PR, reworked to address review feedback:
1. Context-length fallback diagnostic (agent/model_metadata.py):
get_model_context_length() silently returned 256K when all 9 detection
methods failed. Users with small-context models (8K, 32K) would get 256K
silently, causing hard-to-debug API context-length errors. Added a
warning log at the step 9 fallback with model name, base_url, and the
correct config override hint (model.context_length, not context_length).
The token-estimation ceiling-division fix from the original PR already
landed on main (5c2ecdec) with CJK handling — not duplicated here.
2. Fsync for batch trajectory writes (batch_runner.py):
Trajectory entries were written without flush/fsync, but the checkpoint
immediately marked them as completed. A crash between write and disk
sync would leave the checkpoint claiming completion with no trajectory
data on disk. Added flush() + os.fsync() before checkpoint update.
3. Pool cleanup on interruption (batch_runner.py):
Ctrl+C during pool.imap_unordered() relied on context manager cleanup
which can hang. Added explicit pool.terminate() + pool.join() for both
KeyboardInterrupt and Exception paths. The original PR used
pool.join(timeout=10) which is invalid — CPython's Pool.join() takes
no timeout parameter. Fixed to use pool.join() without arguments.
Tests:
- test_warning_emitted_on_fallback: verifies warning fires at step 9
- test_no_warning_when_cached: verifies no false warning when cache hits
- test_trajectory_entry_is_synced_to_disk: verifies os.fsync is called
- test_pool_terminate_called_on_exception: verifies cleanup on RuntimeError
- test_pool_terminate_called_on_keyboard_interrupt: verifies cleanup on Ctrl+C
- test_pool_join_called_without_timeout: verifies no timeout arg to join()
- test_real_pool_join_accepts_no_timeout: integration check on CPython API
Co-authored-by: Aaron Lab <aaronlab@users.noreply.github.com>
Setting prompt_caching.cache_ttl to a falsy value (false, null, off,
disabled, no, none) now fully disables prompt caching instead of
being silently ignored.
The disable propagates through anthropic_prompt_cache_policy() (early
return when _cache_disabled flag is set) and restore_primary_runtime()
(override after snapshot restore), so it survives /model switches and
fallback re-derivation — the gap that caused #56105 to be reverted in
#56126.
Salvage of #33555 by @BB-light, with model-switch/fallback survival
gap fixed on top.
Co-authored-by: BB-light <BB-light@users.noreply.github.com>
The micro-compaction defrag pass (_defrag_rolling_summary) rewrites the
newest MICRO marker's content and pops _DB_PERSISTED_MARKER from the
LIVE dict in place — the same in-place pop class finalize_turn's fill
site was fixed for in #75170. Without invalidation the bounded
flush-scan cursor identity-skips the rewritten marker row and the
defragged rolling summary never reaches state.db (resume rehydrates a
stale summary).
The compressor holds no agent reference, so the pop site raises
_flush_scan_cursor_invalidated and the finalize_turn micro-compaction
block consumes it, setting agent._db_flush_scan_prefix = None.
The module-scope pop sites (context_compressor.py:175/224) operate on
fresh copies — identity-breaking by construction — and need no flag.
Follow-up to #75170 (fix-the-class sweep of _DB_PERSISTED_MARKER
in-place pops).
The bounded flush-scan in _flush_messages_to_session_db_unlocked skips
the identity-matched prefix of its previous snapshot, on the documented
assumption that no code path pops _DB_PERSISTED_MARKER from a live dict
in place. finalize_turn's pure-tool-call-tail fill is exactly that path:
it pops the marker so the filled content gets re-persisted — but the
cursor then skips the row anyway, so the delivered final response never
reaches state.db and /resume replays content="" (the #43849/#44100
class resurfacing via the perf cursor). Invalidate the cursor at the
pop site so the filled row is re-examined.
Companion to the preflight fix: _estimate_msg_budget_tokens charged
reasoning_details at chars/4 via _REPLAY_BUDGET_KEYS, so the signed/base64
envelope (measured 72% of the reasoning mass on Anthropic-wire sessions)
consumed the tail budget and _find_tail_cut_by_tokens summarized away real
transcript to make room for tokens that are never sent (69 messages on the
measured session; up to ~4.8x budget inflation on thinking-heavy histories).
Per the #51800 counter-argument, actual thinking TEXT stays visible to the
budget: _reasoning_details_text_chars counts thinking/text/summary fields
and skips signature/data/encrypted blobs, and the text is skipped entirely
when reasoning/reasoning_content already carries the identical prose (so it
is charged once, not twice). codex_reasoning_items remains fully charged —
Codex Responses genuinely replays it every request (#55572).
Sabotage-verified: restoring reasoning_details to _REPLAY_BUDGET_KEYS fails
the new envelope and double-charge tests.
The reasoning_details field (OpenRouter/Anthropic thinking blocks +
opaque cryptographic signature blobs) inflates the rough token estimate
by ~4x. Providers do not bill these envelope bytes as prompt tokens.
In a measured Kimi K3 session, reasoning_details held 2,124K chars
vs 281K chars of actual thinking text. The estimator reported ~533K
tokens when real prompt_tokens was ~140K — triggering compression at
~27% of the configured threshold.
Fix: skip reasoning_details in both _estimate_message_chars and
_estimate_message_tokens_without_images, alongside the existing
_anthropic_content_blocks exclusion.
Fixes#73298
Fix#75588
## Root cause
When a short conversation ends in a tool-call/result group and the
protected head alignment reaches the end of the message list,
_find_tail_cut_by_tokens() could return len(messages) + 1. This
happened because the final return used max(cut_idx, head_end + 1)
which could push past the array length when head_end >= len(messages).
The out-of-range value then propagated into _find_context_summaries()
which iterated range(start, end) and indexed messages[idx] without
clamping, raising IndexError and failing the active gateway turn.
## Fix
Two-layer defense:
1. Source fix: _find_tail_cut_by_tokens() now clamps its return to
min(n, ...) so it never exceeds len(messages).
2. Defensive clamp: _find_context_summaries() now bounds start/end
to [0, len(messages)] so even if a future caller passes bad values,
it cannot crash.
## Verification
- 7 new regression tests for the exact boundary conditions
- All 214 existing test_context_compressor.py tests pass
Review follow-up on #75102. The shadow substituted the sidecar whenever
the ``api_content`` key was merely PRESENT, but the wire only substitutes
a non-empty string sidecar on a user/assistant row (see
``turn_context.substitute_api_content``). For any other shape the sidecar
is popped and discarded while the clean ``content`` is sent -- so the
shadow dropped real content from the estimate and UNDERcounted, the
dangerous direction: compaction fires too late and the turn dies on a
hard context-length error instead of merely compressing early.
Gate the substitution on the same predicate, and cover the divergent
shapes (None, empty string, int, list, non-user/assistant role) with a
test that fails against the unconditional version.
Also rename the image test: it never carried a sidecar, so it was not
testing what its name claimed. It is a non-regression pin on the flat
per-image accounting that moved into ``_wire_message_shadow()``, and is
now named for that.
`api_content` is a SUBSTITUTE for `content`, not an addition to it.
`turn_context.substitute_api_content()` pops the sidecar and overwrites
`content` at every API-bound message-build site (the `api_messages` build
in `conversation_loop`, the max-iterations summary in
`chat_completion_helpers`, the chat-completions transport), so exactly one
of the two is ever sent to the provider.
The preflight estimator counted both, because both `_estimate_message_chars`
and `_estimate_message_tokens_without_images` walked every key of the
persisted dict with a single-entry denylist (`_anthropic_content_blocks`).
Any message whose sidecar differs from its clean stored content was counted
twice — exactly 2.00x on a 40KB sidecar.
The sidecar exists to keep the provider prompt-cache prefix byte-stable, so
it is written on precisely the long, cache-pinned messages where the
doubling hurts most. Because `estimate_messages_tokens_rough()` also feeds
the compaction threshold via `context_compressor` and `conversation_loop`,
the inflated estimate makes compression fire on phantom bytes.
Fix: substitute rather than sum, mirroring the wire. The two estimator
helpers had drifted into near-identical copies of the same shadow-building
loop, so this factors the shared logic into `_wire_message_shadow()` and
fixes the class once instead of patching one site and leaving the other.
Image accounting is unchanged: base64 payloads are still replaced with a
placeholder and charged at the flat `_count_image_tokens` rate, and the
`_multimodal` text_summary path is preserved.
Tests: three cases in `TestEstimateMessagesTokensRough` — sidecar equal to
content is counted once, a sidecar that DIFFERS is still counted (a lower
bound, so it fails if the field were dropped rather than substituted, which
would undercount the real request), and a sidecar cannot smuggle raw base64
past the flat image rate.
Verified on Linux (Python 3.11): 53 passed in
tests/agent/test_model_metadata.py, 57 passed with
tests/agent/test_context_breakdown.py, 656 passed / 3 skipped across the
compression/context/token/estimate/prune surface of tests/agent.
Mutation-tested: reverting the substitution fails the new equality test.
`scripts/check-windows-footguns.py` is not applicable — no file I/O,
process management, terminal handling, subprocesses, or signals.
agent_init.py's init-time fallback and chat_completion_helpers.py's
try_activate_fallback() still read key_env via raw os.getenv(), missing the
per-profile secret scope installed by the multiplexed gateway (same bug
fixed for fallback_config.py/auxiliary_client.py in this PR). Both now
delegate to hermes_cli.fallback_config.resolve_entry_api_key(), and the
Ollama Cloud OLLAMA_API_KEY read now goes through
agent.secret_scope.get_secret() too.
agent_init.py's fallback loop had no try/except around key resolution
(unlike the other three call sites), so a fail-closed UnscopedSecretError
under multiplexing would have crashed init instead of skipping to the next
fallback entry — added the same skip-and-continue handling.
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
resolve_entry_api_key() and the duplicated _fallback_entry_api_key()
read key_env via a raw os.getenv(), bypassing per-profile secret
scoping in the multiplexed gateway. Under multiplexing this can hand
a fallback request another profile's credential. Both now resolve
through agent.secret_scope.get_secret(), which reads the active
profile scope when multiplexing is on and falls back to os.environ
unchanged when it's off, so single-profile behavior is preserved.
Closes#74311
shutil.move() moves into an existing destination directory instead of
replacing it. When the snapshot extract failed after creating some of its
output, the recovery path moved each staged entry onto a path the extract
had already created, burying the user's own skill one level deeper
(skills/alpha/alpha/) and leaving the snapshot's partial content in its
place. rollback() then returned "snapshot extract failed (state restored)".
Clear the failed extract's output before moving the staged copies back:
entries the original tree never had are dropped, and each staged entry's
destination is removed before the move. When an entry still cannot be
restored, keep the staging dir and name the entries in the message rather
than reporting a restore that did not happen.
AIAgent.__init__ calls set_current_session_id(self.session_id), which
mutated both the task-local ContextVar and the process-global os.environ.
Because _build_child_agent wraps construction in delegated_child_context(),
the ContextVar write is harmless (task-local), but the os.environ write
clobbered the parent's HERMES_SESSION_ID for the rest of the process —
leaking the child id into parent tools and subprocesses spawned after
the child was built.
Root cause of HermesPRDelegationSessionContext: parent
20260729_212118_5d797e dispatched child 20260730_160515_736ea1; later
parent terminal inherited HERMES_SESSION_ID=the child.
Fix: set_current_session_id() skips the process-global os.environ write
when called from within a delegated_child_context(). The child's own
tools and subprocesses still resolve their id through the ContextVar
(task-local), while the parent's process-wide env keeps the parent's
session identity. Root agents (CLI, gateway, cron) retain both paths.
Adds 7 regression tests covering single child, concurrent children (8
parallel), parent-tool observation after construction, and root-agent
session rotation backward compatibility. All pass; ruff clean.
_sync_device_code_entry_to_auth_store used key-presence on the profile
store to decide whether to write-through rotated tokens to the global
root. _store_provider_state unconditionally creates that key, so every
refresh after the first self-disabled the write-through — root kept a
revoked refresh token and every other profile died with
refresh_token_reused / invalid_grant.
Fix: use _load_provider_state_with_source to learn where the grant was
resolved from. When the source is the global root, write back only to
root and skip _store_provider_state so the profile never accrues a
shadowing providers.<id> key that blocks future root fallback.
Add regression test verifying write-through fires on refresh 2+, not
just the first call.
- Bound ALL reads of the on-disk JWT store through one _read_jwt_store()
helper (load, eviction, save-merge) — the 1 MiB cap previously only
covered the load path; eviction and save could still parse an
oversized/corrupt store and rewrite it back out (sweeper finding).
- Fix the class, not the site: the recovery gates checked the literal
provider == "copilot" while /model and profile configs can leave the
alias spelling in place (the reporter's own log shows provider=copilot
AND provider=github-copilot in one session — the aliased turns would
have silently skipped recovery). Single owner:
AIAgent._is_copilot_provider() (slug aliases + Copilot base-URL
fallback), used by both run_agent recovery methods and both
conversation_loop gates.
- Update the salvaged 401 test to current main's client-retirement
contract (release deferred to GC — no synchronous .close()).
- Add copilot_stale_cred_retry_attempted to the TurnRetryState field
contract test; add bounded-store and alias-gate regression tests.
Copilot degrades in two related ways that both abort a turn as non-retryable
and only clear on a gateway restart (a cold process re-runs the token exchange):
1. HTTP 400 model_not_available_for_integrator / model_not_supported — a
raw/degraded token routes to the restricted copilot-language-server
integrator whose allowlist omits enterprise-only models (e.g.
claude-opus-4.8). Because it is a 400 (not 401), the existing 401 refresh
path never fired. Prevented (retry-with-backoff exchange + on-disk JWT
persistence + header guard at the client chokepoint) and self-healed at
runtime (single-shot forced re-exchange + client rebuild + retry before
fallback).
2. HTTP 401 'IDE token expired: unauthorized: token expired' — the short-TTL
*exchanged* IDE token expires mid-turn. The clean-401 path DID fire and call
_try_refresh_copilot_client_credentials(), but that method only re-resolved
the stable raw ghu_ token and rebuilt the client — it never evicted the
cached exchanged JWT or forced a fresh exchange, so the retry put the SAME
expired token back on the wire, 401'd again, and the single-shot guard
aborted the turn. Fix: force a fresh IDE-token exchange (evict cached JWT via
evict_cached_exchanged_token + re-mint via get_copilot_api_token) before the
client rebuild, mirroring the merged auxiliary-path recovery (#59837) and the
400 recovery in this same PR. Graceful fallback to the resolved token if the
exchange endpoint is unreachable; picks up the enterprise base_url on
re-exchange.
Brings main-loop clean-401 recovery to parity with the merged auxiliary path
(#59837), using the newer on-disk-aware evict helper. Companion context: #58743
(this PR, expanded), #51313, #63204 (which assumed the 401 path already
recovered — it reached the method but the method was too weak).
Tests: exchange retry/persist round-trip, restart-blip disk reuse, stale-cred
400 classifier, 400 recovery, and 3 new 401 cases (fresh exchanged token on the
wire; network-blip fallback to resolved token). 58 copilot tests green on
current main.
Disk-full / ENOSPC / SQLITE_FULL on first-message session persist used to be
swallowed as a debug log while prompt.submit still returned streaming, so the
send vanished with no error. Re-raise those failures, return a real RPC error,
and stamp session_persistence_failed turns with error so clients get a terminal
error frame.
build_moa_facade() reused agent.model as the preset name; a session
that had drifted to a fallback model crashed on restore with
MoAPresetNotFoundError. Validate the resolved preset against the
configured presets and fall back to the default preset.
Salvaged from PR #74903 by @liusencomic-cyber.
Narrow the moa_aggregator Relay bypass to CodexAuxiliaryClient and add
coverage that call_llm(stream=True) returns the provider's direct
create() result for Responses-shim clients.
Salvaged from PR #74903 by @liusencomic-cyber.
Convert a completed MoA aggregator response into one valid Chat
Completions delta chunk at the MoA facade boundary, normalize completed
message.tool_calls into indexed stream deltas, and classify these local
MoA adapter-shape errors as non-fallback format errors so a local
compatibility bug cannot silently drift the user's MoA route to a
single model (#55933 follow-up).
Salvaged from PR #74903 by @liusencomic-cyber.
Under managed Relay execution the provider factory runs lazily inside
provider_stream() on the Relay session's event loop. The MoA facade's
auxiliary call_llm(stream=True) is invoked from that callback, so the
eager final_response check added for the non-managed path never fires:
the inner ManagedLlmStream is returned to the outer stream, which then
synchronously iterates it on the same loop thread and dies with
RuntimeError: Cannot run the event loop while another loop is running
(the completed response effectively trapped one level deeper).
stream_current() now detects a running event loop and returns the raw
factory result instead of nesting a ManagedLlmStream: the outer managed
stream already provides Relay tracking for the enclosing attempt, and
its own completed_response_predicate traps the completed response as
final_response — the same contract the main streaming worker consumes
(chat_completion_helpers reads stream.final_response after the chunk
loop). Nested managed streams remain supported for genuinely streaming
providers via the outer stream's own iteration.
Adds managed-execution regressions using the retained relay_turn
fixture: direct completed-response trapping, and the nested
facade-shaped stream_current call.
Port from nanocoai/nanoclaw#2748: Docker's built-in 64 MB /dev/shm silently
breaks shared-memory-hungry workloads inside the sandbox — Chromium/Playwright
renderers crash tabs, and PyTorch DataLoader workers die with 'bus error' /
'insufficient shared memory'. tmpfs is lazily allocated, so the higher ceiling
costs nothing until actually used, and usage still counts against the
container's --memory cgroup limit.
- tools/environments/docker.py: --shm-size 1g in resource args (not
cgroup-gated; tmpfs mount option). Skipped when docker_extra_args already
sets --shm-size, or when configured empty/'0' (Docker default).
- terminal.docker_shm_size config key (DEFAULT_CONFIG + all three
config->TERMINAL_DOCKER_SHM_SIZE env bridges: CLI, gateway, config.py map)
- tests: default emit, custom value, opt-out, extra_args precedence,
helper edge cases (sabotage-verified: default/custom tests fail without
the emit)
Port from openclaw/openclaw#112954. The redactor knew GitHub, Slack,
Google, Stripe, AWS access-key-ID and ~25 other vendor prefixes but had
zero GitLab coverage — glpat-/gloas-/gldt-/glrt-/glrtr-/glcbt-/glptt-/
glft-/glimt-/glagent-/glsoat-/glffct-/glwt- tokens and legacy GR1348941
runner registration tokens passed through display and log surfaces
verbatim. Follow-up explicitly invited when #4541 was closed.
Each pattern keeps a full literal prefix so the _PREFIX_SUBSTRINGS
pre-screen (derived at module load) stays false-negative-free; routable
runner tokens allow dotted segments. Sibling site: skills_guard's
credential-exposure scan gains a gitlab_token_leaked pattern.
Phase 2 review findings on the salvage branch:
C1 (critical): batch and micro summary markers share
COMPRESSED_SUMMARY_METADATA_KEY, and compress() never reset micro state.
After micro absorbed exchanges 1..k, a batch compaction summarizing
1..m (m>k) could fire; the next micro pass's supersede then dropped the
batch marker (whose content the stale rolling summary does NOT contain)
and archive_and_compact immediately made the loss durable. Defrag had
the same hazard: it rewrote "the newest marker" even if that was a
batch marker. Empirically confirmed with a probe (batch marker content
destroyed in one pass).
Fix, three parts:
- Micro-created markers now carry MICRO_COMPACT_MARKER_KEY; supersede
and defrag only ever touch micro-tagged markers. Rehydration in
_resolve_compact_cursor tags the marker it absorbs (containment
proof), which safely covers adopting a batch marker as the new
rolling base after a reset.
- compress() success path resets micro rolling summary/cursor state so
a stale summary can never claim cumulativeness over a batch marker.
- Regression tests for both directions plus the reset.
W4: _splice_micro_compact_result no longer strips _db_persisted stamps
from surviving messages. Micro archives in place under the SAME session
id (unlike batch's child-session rotation, #57491), so surviving stamps
are accurate; stripping them meant an archive_and_compact failure left
every previously-persisted message unstamped and the next append-only
flush re-inserted them all as duplicate active rows.
W5: finalize_turn micro gate now checks agent._persist_disabled —
persistence-isolated fork agents (background review) must not burn an
aux call per review turn, and must never archive_and_compact the
canonical session rows if their compressor ever gains a DB binding.
W1: _serialize_one_exchange now delegates to _serialize_for_summary
(was a ~70-line near-verbatim copy; one serializer, one place to fix).
S4: _find_one_exchange boundary guard rejects only assistant/tool
boundaries (the actual alternation hazard) instead of requiring user —
a stray mid-list system/injected message can no longer wedge the
cursor forever.
5 new regression tests; 38 micro/prune tests, 400 compression-suite
tests, 61 finalize/persist tests pass; ruff clean.
tests/run_agent/test_proactive_prune_loop_wiring.py builds agents with a
MagicMock compressor; getattr(mock, '_micro_compact_enabled', False)
returns a truthy auto-attribute, so the hook called _micro_compact on the
mock and spliced its (empty-iterating) return over the transcript —
wiping all messages before persist (CI slice 7/8 failure).
Gate now requires _micro_compact_enabled is True, a callable
_micro_compact, and a non-empty list result before touching messages.
Same hardening protects production plugin context engines that don't
subclass ContextCompressor.
Two integration bugs found during review of #74522, both confirmed with
empirical probes against the production message-repair path:
1. Alternation: the summary marker was role="user" and an exchange was a
single assistant+tools group, so splicing between two user turns produced
user -> marker(user) -> user. The pre-request repair_message_sequence pass
(conversation_loop.py, runs before EVERY API call) then merged the marker
into the neighbouring real user message: metadata gone, cursor
unrecoverable on resume, and the summary text duplicated into the
transcript on every later pass (the transcript GREW every turn).
Fix: an exchange is now a full agent turn (assistant + tools + follow-up
assistant iterations, bounded by user messages), the marker is
assistant-role, and superseding an old marker deliberately merges the two
adjacent real user turns (plain-text \n\n-join, identical to repair
pass 2) so the returned transcript is alternation-valid by construction.
Probe result: repairs 0 (was 2), marker survives, no summary leakage.
2. Defrag destroyed user messages: _defrag_rolling_summary serialized the
whole remaining middle (user turns included) and spliced it away —
8 of 10 user prompts destroyed in one pass, contradicting the feature's
"your messages are never compacted" invariant. Fix: defrag now
re-summarizes only the rolling summary TEXT and rewrites the marker
content in place; transcript shape, cursor, and user turns untouched.
Probe result: 10 of 10 user prompts survive.
Also: marker provenance is now COMPRESSED_SUMMARY_HAS_USER_TURN_KEY=False —
micro markers absorb only assistant/tool content (#64650 invariant), and
real user turns remain in the transcript for provenance detection.
Adds 5 regression tests (repair-pass integration, alternation on
multi-iteration tool turns, defrag user survival, defrag input scope,
marker provenance); updates the two existing tests and the design doc to
the corrected semantics. 28 tests pass.
The on/off switch was the only knob. A pass fired after every completed turn,
absorbed exactly one exchange, and there was no way to ask for less. Since a
pass is also what breaks the prompt-cache prefix, "how often does it run" and
"how often do I pay a cache break" are the same question, and it had no answer.
Add `compression.micro_compact_every_n_turns` (default 1, clamped to >= 1). At 1
the behaviour is what it was; at 5 you get a fifth of the breaks and a fifth of
the reclaim rate. The counter advances per invocation rather than per committed
pass, so a turn that finds nothing to absorb still moves the cadence along and
cannot wedge it, and a bogus 0 or negative degrades to "every turn" instead of
silently disabling compaction.
Also expose `micro_compact_defrag_threshold_tokens`, which has been a hardcoded
attribute on the compressor with no path from config since it was added.
This does not give micro-compaction the prune's reclaim-size gate -- a pass
still commits whatever the single absorbed exchange saved. It makes the break
frequency tunable, which reaches the same end by absorbing less rather than by
waiting for a bigger win. The docs now say that plainly, including that a
reclaim threshold is the obvious follow-up and does not exist yet.
Tests cover the skip-until-due window, the cursor and prefix staying untouched
on skipped turns, the clamp, and that the feature is off unless enabled.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Review raised whether default-on can be reconciled with the prompt-cache
contract in AGENTS.md, which permits mutating past context only for context
compression and treats per-conversation caching as sacred. It cannot, and the
codebase already says so in its own words.
A micro-compaction pass rewrites already-sent history, so it invalidates the
cached prefix every turn rather than at an episodic boundary. That is the exact
cost the proactive prune gates against: `proactive_prune_min_reclaim_tokens`
exists, per its own config comment, to keep rewrites to "one big episodic break
instead of a tiny break every tool iteration." Micro-compaction has no
equivalent gate -- one exchange per turn means one break per turn, by design.
Default to off. An operator who wants the amortized stall can opt in with
`compression.micro_compact: true` and accept the tradeoff knowingly; nobody
inherits a per-turn cache break from installing an update.
Also register the key in config_defaults so it is discoverable and picked up by
the update path's new-options check -- it was previously read by agent_init but
declared nowhere -- and document the cache cost in docs/micro-compaction.md
instead of only the benefit. The measurements behind the feature (occupancy
plateau, zero batch compactions) never priced cache invalidation, and the doc
now says which numbers a reader would need to measure to justify enabling it.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The rolling summary lives only in memory. A resumed session starts with an
empty one while the marker carrying every previously absorbed exchange is
still in the transcript. The first pass after a resume therefore built a
marker from a single exchange and superseded the marker holding the entire
history -- silently discarding everything micro-compaction had accumulated.
This was introduced by the supersede fix. Before it, markers piled up
wastefully, but nothing was ever lost.
Two changes, so a single failure cannot lose data:
Rehydrate. When the cursor is recovered by scanning the transcript -- the
resume path -- also recover the rolling summary from that marker, so the
next pass merges into the existing history instead of replacing it.
Extraction uses rfind for the heading because SUMMARY_PREFIX references the
heading text itself, so the first occurrence is inside the preamble.
Gate superseding. Earlier markers are dropped only when this pass's summary
is demonstrably cumulative, i.e. the rolling summary was non-empty going in.
If rehydration ever fails, the pass keeps both markers: wasteful, but the
history survives.
Tests cover the resume path, the failed-rehydration fallback, and the
round trip of a summary through a marker.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The cursor was set to the pre-splice `exchange_end`. A splice collapses the
absorbed span -- an assistant plus its tool results, often four or more
messages -- into a single marker, and may also drop a superseded marker
further back, so every index after it shifts.
The stale cursor therefore overshot, landing inside a *later* exchange's
tool group. The next pass's `_find_one_exchange` walked forward from there
to the following assistant, so the exchange it had landed inside was never
absorbed at all. On tool-bearing conversations micro-compaction was
silently doing roughly half the work it should.
Traced on a 3-tool-per-exchange transcript: the cursor sat at index 6 when
the marker was at 2, and the message count stalled at 32 instead of
continuing to 28.
Derive the cursor from the marker's actual position in the spliced result
instead, which is self-correcting regardless of how much the splice moved.
Apply it on the defrag path too, which had the same staleness.
Found by a randomized long-horizon harness (480 conversation shapes x 25
passes, varying tool-group sizes and summarizer failure modes) asserting
structural and progress invariants after every pass. Existing tests missed
it because their fixtures have no tool results, so the absorbed span is one
message and nothing shifts. The regression test uses tool groups.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Tokens saved is the wrong headline for this feature. Micro-compaction is
not an efficiency optimisation — the same summarization work happens either
way. What it buys is (a) that work amortized across turns instead of one
stall, and (b) a window kept low enough that a session runs much further
before needing a hard compaction at all.
Neither shows up in "net tokens saved". A session can save nothing on paper
and still be a clear win on both counts.
So the telemetry now carries occupancy: tokens_after as a share of the
compaction threshold, plus the threshold and resolved window it was
computed from. That is the number that says whether a session has headroom
left. The report leads with it, and cross-references the batch
`compression_attempt` lines already in the log so it can show how often the
long pause actually fired — ideally never.
Occupancy is read from the cached threshold only. The public
`threshold_tokens` property resolves lazily and can issue a synchronous
/models probe (#32221); telemetry must never be the thing that blocks a
turn, so an unresolved window reports null. In practice a pass has already
resolved it via the tail calculation, so the field is populated. A test
pins the no-forcing behaviour directly against the emitter.
The report is pure ASCII: `scripts/check_subprocess_stdin.py` currently
dies on a cp1252 console before printing its results, and a diagnostic tool
that crashes on the platform it is diagnosing is worse than no tool.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The existing log line reports message counts, which is the least
informative number available here: absorbing one tool-heavy exchange can
drop hundreds of tokens while moving the count by one. There was no way to
answer "is this actually helping?" from a real session.
Emit one content-free JSON line per pass, in the same shape as the batch
compaction telemetry: before/after tokens, the delta, the size of the
absorbed exchange, the rolling summary size, duration, and running
per-session totals so a whole run can be read off the last line. No
transcript content rides along.
Add scripts/micro_compaction_report.py to aggregate those lines into
passes, outcome mix, net tokens saved, mean exchange size and durations,
with an optional per-session breakdown.
Measuring it immediately surfaced something worth documenting: the first
pass in a session normally *costs* tokens. The summary marker carries a
fixed ~400 tokens of scaffolding, paid on pass one against a single
absorbed exchange. From pass two the marker is replaced rather than added,
so the overhead is already paid and each exchange is close to pure saving.
Break-even is typically the second or third pass. Tests cover the
telemetry contract, the cumulative totals, and that first-pass/later-pass
shape so nobody reads a single turn and concludes it made things worse.
The estimator costs ~5 ms at 600 messages and ~20 ms at 1200, taken twice
per pass, post-turn — and only once an exchange is actually in hand, so
turns that no-op early pay nothing.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The `_micro_compact` docstring cited "#82483" for the resume double-load
problem. No such issue exists — the repository's highest number is 74323,
so the reference was invented rather than looked up.
The reasoning it was attached to is correct and stays: the session flush is
append-only, so an in-memory splice alone leaves the original rows active
and a resume loads both the summary and the messages it replaced. Only the
citation was wrong.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
`_find_one_exchange`'s docstring described an exchange as "(optional) user
message + assistant message + its tool results", but the walk skips past
user messages and starts at the assistant, so user turns are never absorbed
into the rolling summary.
The code is right and the docstring was wrong. Assistant output is largely
an account of what was done and survives summarising with little loss. The
user's messages are the intent everything else is derived from and cannot be
reconstructed from the work that followed — paraphrasing "use the existing
helper, don't add a new one" into a summary is how an agent ends up doing
the opposite six turns later. They are also cheap: a prompt is normally a
tiny fraction of what one tool result costs.
Correct the docstring, document the property (and its cost — a floor on how
small the middle can get, since user turns accumulate), and add a test so it
stays deliberate.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Batch compaction pauses a session for one large summarization once the
window fills. Micro-compaction spreads that cost out: after each completed
turn, `finalize_turn` folds the single oldest un-absorbed exchange
(assistant message plus its tool results) into a rolling summary, so the
work happens in small increments during post-turn idle time instead of one
long stall.
Mechanics:
- a cursor tracks the first message not yet absorbed, recovered from the
transcript's last summary marker when in-memory state is unavailable;
- protected head and tail windows are never touched, so the system prompt
and recent turns stay verbatim;
- the absorbed span is replaced by a marker carrying the usual
`_compressed_summary` metadata, so resume, handoff and `/compress`
treat it exactly like a batch summary;
- `archive_and_compact` keeps the session DB in step, otherwise the
append-only flush would leave the original rows active and a resume
would double-load both summary and originals;
- when the rolling summary itself passes a token threshold it is
defragged: re-summarized in one shot and the cursor jumps to the tail;
- an exchange the summarizer can't handle is retried a bounded number of
times, then skipped, so one poison exchange can't stall every turn.
Keep only the newest summary marker. The rolling summary is cumulative, so
each marker already contains everything the previous ones held; leaving them
stacked near-duplicate copies of the same text, each with its own heading and
end-marker scaffolding, and the transcript grew on every turn instead of
shrinking. Measured over six turns on a 12-exchange conversation with tool
output: 4104 -> 4797 tokens before, 4104 -> 2572 after.
Off switch: `compression.micro_compact: false` (default on).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
The compaction summary's role was selected against the LITERAL
neighbouring messages (compressed[-1] / tail_messages[0]). Mistral-family
chat templates (Devstral, Mistral Small 3.x, Magistral) enforce
user/assistant alternation but exempt the tool flow (tool results and
assistant messages carrying tool_calls) from the check, so a protected
head ending [user, assistant(tool_calls), tool] pinned the summary to
role="user" while the last role the template counts is "user": the
backend rejects the whole request with a Jinja alternation error
(HTTP 500). The summary persists in the stored conversation, every
retry replays the identical poisoned history, and the session is
permanently unrecoverable. Fires on EVERY compaction against a
Mistral-strict backend, captured byte-exact via a tee-proxy in front of
a llama.cpp/llama-swap Devstral deployment.
Fix: compute both neighbour roles through _template_visible_role(),
which skips template-exempt messages. The #52160 (Anthropic user-first)
and #58753 (zero-user-turn) forced-user guards are preserved; their
forced shapes (summary-user followed only by exempt messages) are
alternation-safe. When the visible head ends "assistant" and the
visible tail opens "user", no standalone role can alternate and the
existing merge-into-tail fallback now correctly fires (the literal
logic emitted a standalone user summary there: a second poisoning
shape).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01LGN45sMMbwM8cW9T9ga4ou
Reactions live in the existing messages.display_metadata JSON column (no new
table), with iOS Tapback semantics enforced DB-side: one reaction per author
per message, re-tap retracts, different emoji replaces. The desktop catches up
to the reaction contract five platform adapters already ship.
- SessionDB: set/get_message_reaction, latest_message_row_id (role + offset +
require_text so invisible tool-call-only rows are never targeted),
take_unseen_reactions (announce-exactly-once), get_message_role
- message.react RPC: accepts row_id or newest_role for live messages that
haven't learned their durable id yet
- react_to_message tool: desktop-gated (check_fn), defaults to the user's
latest visible message, messages_back for retroactive reactions
- Model context rides run_message only (beside the speech-interrupted note):
the persisted prompt stays clean, so no [The user reacted …] scaffolding in
transcripts, and no cached prefix ever changes
- Resume projection forwards row_id + reactions; _row_id is stripped from
outgoing API copies next to display_metadata
Use database.journal_mode as the sole non-secret operator setting, preserve the vulnerable-SQLite safety gate and existing WAL databases, validate explicit DELETE results, document the active config path, and cover real SQLite openers with behavioral tests.
Add HERMES_JOURNAL_MODE env / database.journal_mode config for
virtiofs/NFS/SMB where WAL is not crash-safe. Route 5 bypass openers
through apply_wal_with_fallback so a single setting covers every .db
(#68545).
Salvaged from #56085 (@Stoltemberg), rebased onto current main: sites
main had already converted (credential_pool, auxiliary_client MoA
paths, model_metadata, moa_loop, agent_runtime_helpers) resolve to
main's versions; the remaining ~29 read-only sites across 16 agent/
files swap to the no-deepcopy readonly loader (~135us saved per call).
Full per-site mutation audit performed (every enclosing function read,
escapes traced): 23 SAFE, 5 ESCAPES with read-only consumers, 1 UNSAFE
path (init_agent -> get_compatible_custom_providers -> normalizer
in-place alias writes) fixed by the preceding no-mutate commits, which
make the normalizer copy-safe for ALL callers.
The 1.0s sleep between sequential tool calls has been present since the
initial commit with no documented rationale. It sleeps between local
tool executions — the next LLM request goes out only after the whole
batch — so it rate-limits nothing, and the parallel read-only path
already runs with no delay. Every multi-tool turn pays (N-1) seconds
of dead time. Remove the sleep, the internal tool_delay plumbing, and
dead test assignments. AIAgent.__init__ keeps tool_delay as a
deprecated no-op keyword for one release so existing programmatic
callers construct cleanly; passing it emits a DeprecationWarning.
- Remove tests/-shadowing sys.path.insert(dirname/'..') from 11 test files:
it prepended the tests/ dir itself to sys.path, so 'import agent' /
'import hermes_cli' resolved to the test packages and collection died
with ModuleNotFoundError depending on import order (2 files failed in
every full-suite run; 9 more were latent).
- Patch call_llm in 5 context-compressor tests that called compress()
unmocked: each burned ~50s attempting live LLM traffic through the
relay before falling back (572s file — the slowest in the suite, and
flaky under the 300s per-file timeout). File now runs in ~5s.
- agent/redact.py: fix two catastrophically-backtracking regexes hit by
the compressor's redaction pass on large payloads —
_STRICT_URL_USERINFO_RE anchors on the mandatory '//' (optional-scheme
prefix backtracked O(n^2): ~55s on a 320KB payload, now sub-ms;
output-equivalence fuzz-verified on 20k random strings), and the
_CFG_DOTTED_RE/_CFG_ANCHORED_RE subs gain an exact linear keyword
pre-gate so secret-free text skips the quadratic pattern entirely.
- tests/gateway/test_feishu.py: version-guard the extra_ua_tags SDK
signature check; the repo pins lark-oapi==1.6.8 but stale local
installs (1.5.3) fail the assertion — skip below the pin.
- tests/tools/test_managed_browserbase_and_modal.py: stub
agent.redact + agent.credential_persistence in the fake agent package
(empty __path__ blocks all real agent.* imports added since the fake
was written).
- tests/gateway/test_startup_restart_race.py: raise wait_for timeouts
2s -> 30s; 2s wall-clock on a loaded 40-worker box flaked in the
baseline run (passes instantly when the box is quiet).
Resolves the PR's conflict with main (2252 commits). Two conflicts, both
"each side added an independent block in the same place" — kept both:
- gateway/run.py — the housekeeping loop. This branch adds the Skill Sync
pulls inside the CURATOR_EVERY branch (12-space indent); main adds a
stale-session auto-archive as a sibling `if` at loop level (8-space).
Different scopes, so the naive union would have mis-nested the archive
block into the curator branch; kept each at its own indent level.
- tools/skill_manager_tool.py — the _edit_skill result dict. This branch
appends the org auto-propose note; main appends
_add_description_prompt_preview(). Independent, order-insensitive.
No behaviour dropped from either side.
Verified: 3552 passed / 0 failed across 63 suites (scope regenerated to
include main's new maybe_auto_archive / _add_description_prompt_preview
consumers) via scripts/run_tests.sh. `hermes sync` and `hermes sync status`
still work against a live token, resolving the production plane default.
The Pyright Optional-parameter warnings in skill_manager_tool.py are
pre-existing on main (`content: str = None` etc.), not introduced here.
A transport error during the post-terminal SSE drain (used only to let the
Relay transport finalize the attempt) was sharing exception handling with
the pre-terminal assembly path, so it discarded an already-completed,
already-billed response and opened a brand-new physical request. Give the
drain step its own non-fatal error handling: log a finalization warning
and still return the terminal response.
Closes#74310
Three provably-safe optimizations for O(n)-per-iteration history walks:
1. sanitize_tool_call_arguments: optional identity-keyed cursor (strong
refs to the exact validated message objects) skips re-json.loads-ing
already-validated history each loop iteration. Any list rewrite
(compression, repair, undo, steer) breaks the identity prefix match
and forces re-scan from the divergence point. Wired via a per-agent
cursor dict in conversation_loop.
2. estimate_messages_tokens_rough: per-message memo keyed on a deep
identity fingerprint (strings pinned by strong reference so id()
aliasing is impossible; scalars by value; dicts/lists structurally
with key order). Equal fingerprints imply identical str(shadow)
bytes, hence identical estimates. Unfingerprintable shapes fall
through to direct compute. Bounded FIFO cache (4096 entries).
3. _flush_messages_to_session_db_unlocked: bounded scan that skips the
identity-matched prefix of the previous successful flush's snapshot.
Snapshot only taken on full success; cleared on exception. Compression
rewrites use fresh copies, breaking identity and forcing full re-scan.
Parity proven in tests/agent/test_cursor_optimizations_parity.py:
500-message synthetic histories with tool calls, malformed args, unicode,
element-wise old==new across 3 iterations incl. simulated compression.
Measured (median of 5): sanitize 0.097ms->0.011ms, tokens 1.145ms->0.853ms,
persist-scan 179.5us->10.0us at 500 messages.
Four hot-path consumers paid a full config deepcopy per read:
- telemetry gate relay_shared_metrics.enabled() — runs 2-3x per agent
turn (2x per API call from lifecycle hooks + 1x per tool call) and
called read_raw_config(), which deepcopies the whole raw config every
call. New read_raw_config_readonly() serves the cached dict directly:
248 us -> 4.6 us per call (54x) on Teknium's real 77-key config.
- interruptible_streaming_api_call local-endpoint stale-timeout branch
called load_config() once per API call for every local-model user.
- gateway get_inbound_media_max_bytes() + _get_ephemeral_system_ttl_default()
called load_config() on per-message paths. All three switched to
load_config_readonly() (345 us -> 12 us; PR #28866 lineage).
Together these account for ~90% of the ~1,900 deepcopy primitives per
turn measured in the 26-call stubbed-LLM profile.
read_raw_config_readonly() keeps the (mtime_ns, size) freshness key so
config edits are picked up next call, and preserves the identity
invariant (cache-miss returns the same object later hits serve) —
regression-tested with 'is', per the PR #28866 identity-bug lesson.
The mutable read_raw_config() is unchanged for save-path callers.
581 targeted tests green (config, relay metrics x2, ephemeral reply,
platform base, new readonly suite).
TUI (tui_gateway/server.py _load_approval_mode): now delegates to
tools.approval._get_approval_mode instead of re-reading config raw via
_load_cfg + _deep_merge(DEFAULT_CONFIG, ...) and normalizing locally.
Behavior fix, not pure refactor: the canonical load_config path applies
managed-scope config overlays and ${VAR} env expansion, plus a legacy
max_turns lift, which the TUI's raw YAML read bypassed — under a managed
config that sets approvals.mode, the TUI previously reported/toggled a
different mode than the approval gate actually enforced. Both surfaces
now agree by construction. Name/signature and the mode-vocabulary clamp
are preserved.
Codex (agent/transports/codex_app_server_session.py): read confirmed the
_decide_exec_approval/_decide_apply_patch_approval paths carry NO
Hermes-side mode/timeout reads — the Hermes resolution already flows in
from agent/codex_runtime.py via tools.approval.is_approval_bypass_active()
(auto_approve_* routing) and via the shared approval-gate callback. So no
code extraction was needed; added docstrings pinning that invariant and a
cross-reference on the protocol-semantic choice mapping
(_approval_choice_to_codex_decision), which intentionally stays local.
Adds tests/tools/test_approval_mode_parity.py: cross-surface invariant
test asserting the core resolver, the TUI path, and the codex bypass
derivation agree for synthetic configs (unset defaults, global mode set,
YAML-bool off, malformed values, whitespace/case), plus a delegation-seam
test proving the TUI has no independent config read left.
Note: gateway/run.py has sibling raw reads but is intentionally untouched
(multiple in-flight PRs); flagged as follow-up.
Four deferrals following the established truthy-skip / PEP 562
lazy-load patterns (PRs #22681/#22859 lineage). Rebased over #74194,
which independently landed the browser_tool half of this work — that
file is dropped here; the remaining four modules are untouched by it:
- tools/vision_tools.py: defer agent.auxiliary_client
(credential_pool -> hermes_cli.auth -> httpx -> rich, ~50 ms) to
first vision handler call. async_call_llm /
extract_content_or_reasoning stay patchable module attributes;
injected test mocks win over the loader.
- agent/model_metadata.py: defer 'requests' (+urllib3, ~27 ms of the
'import cli' waterfall) to the fetch functions. PEP 562 __getattr__
keeps patch('agent.model_metadata.requests.get') working.
- tools/browser_supervisor.py: websockets (~22 ms) imports on first
CDP connect; ClientConnection type under TYPE_CHECKING.
- cron/jobs.py: croniter (~15 ms) resolves on first cron-expression
use; HAS_CRONITER stays monkeypatchable (None = unprobed sentinel).
A/B vs current main incl. #74194 (median of 7, cold subprocess):
import cli 147 -> 132 ms (-10%)
import model_tools 244 -> 224 ms (-8%)
import run_agent 264 -> 244 ms (-8%)
Lazy-verify: importing the four modules no longer pulls requests /
croniter / websockets into sys.modules. 369 targeted tests green
post-rebase.
Local-model users paid a fresh probe waterfall on EVERY CLI cold start
inside AIAgent.__init__: detect_local_server_type (up to 4 HTTP GETs,
2s timeout each on a hung server) + /api/show (3s timeout). The
existing caches were in-process only, so back-to-back invocations
(chat -q, cron ticks, subagents) re-paid the network every time.
- New 300s-TTL disk L2 at HERMES_HOME/cache/local_endpoint_probes.json
for detect_local_server_type verdicts and query_ollama_num_ctx
results. Only SUCCESSFUL probes persist (a down server never pins a
negative verdict); stale entries pruned on write; corrupted cache
degrades to a miss; atomic writes. 300s is strictly fresher than the
1h in-process TTL that already accepts server-swap staleness.
- models.dev fetch timeout 15 -> (5, 10) connect/read tuple: a
blackholed connect stalled the first-turn critical path 15s; now
fails in 5s (matches the OpenRouter fetch convention, #46620).
- _auto_detect_local_model timeout 5 -> (2, 3): runs inside
_get_model_config() at startup against a LOCAL endpoint; a hung local
server cost 5s before the banner.
E2E (real HTTP server, two fresh subprocesses, isolated HERMES_HOME):
proc1 = 2 HTTP hits, proc2 = 0 HTTP hits, identical results
(ollama/131072), probe wall 74.5 -> 35.5 ms. 222 targeted tests green
incl. 9 new disk-L2 contract tests.
The streaming hot loop computed len(repr(chunk)) on EVERY chunk to feed
the retry-diagnostic byte counter — a full recursive pydantic repr at
5.5-8.8 us per chunk (measured), ~20-30 ms of pure CPU per 3,000-chunk
response, paid on every streaming response on every platform.
New _estimate_chunk_bytes() sizes the chunk from its delta payload
strings (content / reasoning / tool-call arguments) plus a 40-byte
framing floor: 2.1-2.4 us per chunk (~3x cheaper), independent of
pydantic field count, never raises on unknown shapes (Anthropic events,
stub providers fall back to the floor). Both call sites switched
(chat-completions loop + anthropic event loop).
The counter feeds only the stream-retry diagnostic log line
(agent/stream_diag.py) — an estimate proportional to traffic preserves
its purpose (distinguishing 'died at 0 bytes' from 'died mid-stream').
6 new contract tests; 64 targeted stream tests green.
- extract _commit_registry/_note_refresh_failure shared by the background
worker and foreground stage-4 (identical 4-step success + failure paths
were duplicated); worker now commits under _models_dev_fetch_lock so a
failing background refresh can never re-arm the backoff immediately
after a successful force_refresh committed (unsynchronized-write race)
- add should_clear_context_pin_async to hermes_cli/route_identity.py
(matching the get_model_context_length_async precedent) and use it at
the 4 async gateway sites instead of inline asyncio.to_thread wraps;
the sync _format_session_info site keeps the sync call (already
off-loop via its callers' to_thread)
- test the background-refresh success path (the PR's primary new
behavior): disk saved, mem cache swapped, backoff cleared, in_flight
reset — mutation-checked
- replace the race-prone spin-wait on _models_dev_refresh_in_flight with
a named-thread join in the backoff test
The branched call shape in get_provider_info/get_provider is deliberate:
~69 test sites across tests/hermes_cli and tests/gateway monkeypatch
fetch_models_dev (and get_provider_info) with zero/single-arg lambdas.
Passing allow_network= unconditionally broke 5 tests in CI slices 2/3/7.
Documented the constraint inline.
- _mark_stale_cache_grace only moves cache_time forward so a completed
background refresh is never rewound to a 5-minute grace window
- clear _models_dev_refresh_in_flight if Thread.start() raises so a
one-off thread-exhaustion failure doesn't disable refresh forever
- move empty-registry validation into _fetch_models_dev_from_network
(was duplicated in the background worker and the foreground fetch)
- pass allow_network through as a plain kwarg in get_provider_info and
hermes_cli.providers.get_provider instead of the branched call shape
- refresh the stale module docstring (no bundled snapshot exists; the
resolution order now describes stale-serve + background refresh)
Follow-ups on top of @DonutsDelivery's salvaged reaper commit (#64141):
- create_from_config() now parses lsp.idle_timeout (invalid values fall
back to DEFAULT_IDLE_TIMEOUT) — previously the constructor knob was
unreachable from config.yaml (config exposure adapted from #36892 by
@0xbWy and #68091 by @9miya20)
- canonical default declared in hermes_cli DEFAULT_CONFIG so config
discovery surfaces the knob (per sweeper review note on #64980)
- reaper loop survives transient sweep errors instead of dying and
silently re-opening the leak (gap flagged in #68091 review)
- eventlog.log_reaped(): one INFO line per sweep + clears the
log_active announce cache so respawns re-announce at INFO
- docs: replace the stale 'no idle-timeout reaper' paragraph with the
new lifecycle description + config reference
- tests: reuse-refresh protection (the regression teknium's sweeper
requested on #64141), reaper-survives-error, config propagation,
invalid-value fallback, DEFAULT_CONFIG/manager-constant sync
gemini-2.5-flash shuts down Oct 16 2026 (Google deprecation schedule)
and gemini-3-flash-preview is superseded. Update every hardcoded
default to the current GA flash model:
- gemini_native_adapter: probe_gemini_tier + _create_chat_completion
default params; free-tier guidance de-pinned from a specific model's
RPD number so it doesn't stale again
- auxiliary_client: gemini/kilocode fallback aux models,
_OPENROUTER_MODEL, _NOUS_MODEL -> google/gemini-3.6-flash
(verified live on both OpenRouter and Nous portal /models)
- provider plugins: kilocode + vertex default_aux_model
- hindsight memory plugin: gemini provider default
- setup wizard gemini list: 3-flash-preview -> 3.6-flash (matches the
curated picker catalog)
- tests: aux-client assertions that pinned the old default literal now
reference the _NOUS_MODEL constant, so the next default bump can't
break them (change-detector cleanup)
Fixes#32360.
* Revert "fix(credits): remove the 'Grant spent · $X top-up left' notice"
This reverts commit 5dc6a14c14.
* fix(credits): reintroduce grant_spent behind an in-session crossing gate
The removed notice nagged because its condition is a steady state for
accounts living on top-up, the latch is per-session, and the cold-start seed
runs the policy at every session open — every session re-announced
'Grant spent · $X top-up left'. Reintroduce it gated so only a session that
WATCHES the grant run out announces:
- seen_grant_unspent crossing gate, mirroring seen_below_90: opens when the
session observes the grant meaningfully unspent (>= GRANT_UNSPENT_MIN_MICROS,
1 cent — portal-seeded states derive micros from float dollars and can carry
sub-cent residue where headers report exactly spent). Seeds never prime it.
- The gate guards only the show branch and is consumed by the announcement —
one announcement per crossing. Header flicker (used_fraction None and back)
clears the sticky line but cannot re-announce; a renewal re-opens the gate.
- new_credits_latch() centralizes the latch shape (agent build, lazy re-init,
and the policy test helper all build through it).
- Tests: steady-state-open stays silent (seed + policy seams), live crossing
announces once, flicker/renewal/residual cases locked; the rendering test
primes the gate and asserts its leg count so a gate regression cannot
silently shrink coverage.
Salvage of PR #52188. The original PR raised RuntimeError when LM Studio
load was rejected or unverifiable, which would abort agent startup on
transient network failures. Replace with logger.warning + fallback to
configured context length, preserving the old graceful-degradation behavior.
The codex app-server namespaces MCP tool call ids as
codex_mcp__<server>__<tool>_<codex_call_id>. With an exec-<uuid> component the
built-in hermes-tools server alone overflows the Responses API's 64-char
call_id limit, so the request 400s with a non-retryable "string too long".
The offending item sits near the head of the transcript and replays every
turn, permanently bricking the session — the only recovery is /reset.
Sibling defect to #10788, which clamped input[*].id via
_MAX_RESPONSES_ITEM_ID_LENGTH. Apply the same treatment to call_id at both
Responses emit sites in _chat_messages_to_responses_input: a deterministic
surrogate (call_ + sha256[:32]) for ids over the limit, short ids unchanged.
Because the surrogate is a pure function of the original id, a function_call
and its matching function_call_output — which carry the same original id — map
to the same surrogate and stay paired without correlating the two items.
Follow-up to the #60063 salvage: the curated gemini list now carries
gemini-3.6-flash (aux default, #70416) and the vertex list carries
gemini-3.5-flash-lite (#68767) — both need snapshot pricing so direct
Gemini/Vertex sessions don't report cost=unknown.
Rates verified against https://ai.google.dev/gemini-api/docs/pricing
(2026-07-28): 3.6-flash $1.50/$7.50, cache read $0.15;
3.5-flash-lite $0.30/$2.50, cache read $0.03.
The grant_spent notice fired for every subscription user with top-up
funds the moment their cap was reached and camped in the CLI/TUI status
bar and desktop toasts with no action to take — the account keeps
working off top-up. Remove it everywhere:
- agent/credits_tracker.py: drop grant_cond + the emit/clear block;
the dev fixture state now (correctly) produces no notice
- TUI: keep the turn-start clear of credits.grant_spent as back-compat
for older backends that still emit the key
- Desktop: drop the demo step and stale comment references
- Docs/config comments: remove grant-spent from credits_notices text
- Tests updated: policy/cold-start now assert the key never fires
Usage bands, depleted, and restored notices are unchanged; /usage still
reports the full balance breakdown.
Review-pass findings on the lazy-init deferral:
- No-op guard: the codex app-server usage callback assigns
compressor.context_length on EVERY response (same window each time).
The setter unconditionally invalidated the derived budgets, wiping
runtime corrections applied directly to threshold_tokens /
tail_token_budget (aux-context threshold sync) — those persisted on
main's eager init. Same-value assignment is now a no-op.
- Re-floor on genuinely new window: the setter invalidates budgets but
previously kept the stale threshold_percent, so a codex window switch
recomputed threshold_tokens from the new window with the old model's
floored percent. Re-apply the raise-only small-context floor from
_base_threshold_percent so percent and tokens derive from the same
window (guarded with getattr for object.__new__ test instances).
- Init log extracted to _emit_init_summary_once() and also fired from
the setter path, so a consumer assigning context_length before any
read no longer strands the startup line forever.
- threshold_tokens getter resolves the window into a local before
reading threshold_percent — correctness no longer depends on
left-to-right argument evaluation order.
- reasoning_timeouts: fix inaccurate 'tuples are immutable' comment
(the container is a list; safety comes from build-once-at-import),
document why the slug stays in the tuple.
Adds TestContextLengthSetterCoherence (3 tests): same-value assignment
preserves overrides; new-window assignment re-floors both directions.
With the selective prompt-cache copy (#57046), un-marked messages on the
decorated api_messages list share their nested content parts with the
persistent conversation history — the per-message copy in
conversation_loop is shallow and decoration now deep-copies only the
marked messages. try_shrink_image_parts_in_messages previously wrote the
re-encoded image INTO the aliased part/source dicts, so an
image-too-large retry on an Anthropic route would silently replace the
original image bytes in agent.messages (and persist the degraded copy).
Replace the in-place writes with copy-on-write: rebuild the content list
with fresh part/source/image_url dicts and reassign msg['content'] — a
top-level write on the per-call copy that never reaches history.
Adds two regression tests simulating the aliasing; both fail against the
old in-place implementation (mutation-verified).
The lazy-init change in #32221 deferred get_model_context_length() out of
ContextCompressor.__init__, but the "Context compressor initialized" log
(emitted only when quiet_mode=False) reads self.context_length,
self.threshold_tokens and self.tail_token_budget. Those property reads
resolve the deferred value, so the synchronous model-metadata probe still
ran inside __init__ on the interactive-CLI path — the exact blocking the
PR removed, just narrowed to the non-quiet path.
Emit the informative line once, on first context-length resolution, so
construction stays non-blocking on every path. Add a regression test
asserting no probe fires in __init__ with quiet_mode=False and that the
init log is emitted exactly once on first access.
_match_any() was re-sorting _REASONING_STALE_TIMEOUT_FLOORS (21 elements)
on every call. This function runs per API turn via error_classifier,
chat_completion_helpers, and thinking_timeout_guidance.
Also fixes thread-safety: the old _PATTERN_CACHE was a mutable dict
accessed from multiple threads without locking. Pre-compiling all
patterns at module load time eliminates both the per-call sort and
the TOCTOU race condition. The resulting list is effectively
immutable after import, safe for free-threaded Python 3.13+.
(cherry picked from commit e8b006b853)
Replaces full deepcopy with selective shallow copy in apply_anthropic_cache_control.
Only the 4 messages that receive cache_control markers are deep-copied; the rest
stay as references.
Measured on a 100-message conversation (typical long session):
- Before: ~15ms per call
- After: ~2ms per call
- 7.5x faster, scales with conversation length
Memory impact is also significant — no need to duplicate dozens of unchanged
messages on every turn.
The contract is unchanged: callers still get an independent message list
they can mutate. Only messages we modify (by injecting cache markers) are
copied. The rest are shared references to immutable history entries, which
is safe since the agent never mutates past turns.
(cherry picked from commit 17892df6cc)
Every LLM call built a fresh openai.OpenAI wire client (new httpx pool,
TCP+TLS handshake, measured 19.2ms p50 / 35.5ms p95 per call at ~5 calls
per tool-loop turn) and closed it when the request finished. Cache ONE
reusable wire client on the agent, keyed by the effective client kwargs:
- _create_request_openai_client hands back the cached client when the
effective kwargs are identical; any change (credential rotation,
provider failover, vision default_headers) evicts and rebuilds.
- Only a request that produced a response reports a reuse close reason
(request_complete / stream_request_complete); error unwinds report
*_error_cleanup and really close, so a retry after a request error
always builds a fresh pool.
- Cross-thread aborts poison the slot: a pool whose sockets were
shutdown(SHUT_RDWR) from a stranger thread is never reused (#29507) —
the owner-thread close discards it and the next create rebuilds. The
holder read and the abort are atomic (under the holder lock) at all
three abort sites, so a late abort can never poison the NEXT request's
checked-out client.
- Worker-side interrupt breaks close the half-read SSE stream on the
owning thread before building the partial response; a failed close
poisons the slot (otherwise each interrupt leaked one checked-out
connection until the pool hit PoolTimeout). run_codex_stream gets the
same poison-on-close-failure handling.
- Single checked-out slot (in_use): a concurrent call gets an untracked
client with the old per-request lifecycle.
- release_clients() / close() really close the cached client when idle;
if a worker has it checked out they abort the sockets and detach the
slot, deferring the FD release to the worker's own close.
- MoA facade and Mock passthroughs never enter the cache; max_retries=0
is preserved on all request clients.
Every API call in the tool loop persisted its token/cost delta by
calling SessionDB.update_token_counts() synchronously on the turn
thread — a BEGIN IMMEDIATE sessions UPDATE plus a session_model_usage
upsert, measured in production at p50 3.3ms / p95 70.4ms per call and
up to 299ms against a cold multi-GB state.db. The tool loop stalls for
that long between calls, N times per multi-tool turn.
SessionDB gains queue_token_counts(): same signature and semantics as
update_token_counts(), but the critical path is a deque append plus a
condvar notify. A lazily started daemon thread applies deltas in
enqueue order through the existing update_token_counts ->
_execute_write path, so the established self._lock / BEGIN IMMEDIATE /
jitter-retry discipline is unchanged. When a backlog forms, adjacent
same-route incremental deltas coalesce into one UPDATE: token and
api-call fields sum, cost fields sum None-preservingly (an all-None
run stays None so COALESCE keeps the stored value), and absolute=True
deltas never merge and act as ordering barriers. Route equality is
required for a merge because those fields feed COALESCE backfill, the
last-non-None-wins status fields, and the per-model usage attribution
key — a merged apply is row-equivalent to sequential applies.
Correctness and durability:
- flush_token_counts() gives read-your-writes to token/cost readers
(get_session, list_sessions_rich, _get_session_rich_row,
list_gateway_sessions, InsightsEngine.generate) — a plain attribute
check when nothing is queued. The writer sets its busy flag before
popping the queue so the lock-free fast path can never miss an
in-flight batch.
- update_session_model / update_session_billing_route /
update_session_meta write the sessions row synchronously, bypassing
the queue, so they flush it first: a still-queued first-of-session
delta carries the pre-switch route, and applying it after the switch
UPDATE would trip the first_accounted_route branch (api_call_count
== 0 plus a route mismatch) and resurrect the old model/provider.
- AIAgent._persist_session flushes at turn finalize and every
error-exit persist point; close() stops and drains the writer before
the WAL checkpoint; an atexit hook (registered on first enqueue,
unregistered on close so closed instances are not pinned until
interpreter exit) drains at shutdown. Worst-case crash loss is the
in-flight call's delta — the same window as the old inline write.
- A flush trusts a live stop-flagged writer (its loop drains before
exiting) and only drains on the caller's thread when the writer is
dead or never started, claiming the same busy flag so concurrent
flushes wait instead of racing an in-flight batch.
- After close() has stopped the writer, queue_token_counts applies the
delta inline instead of parking it on a queue nothing will drain; a
closed-connection failure then raises at the call site, which
already guards for it, exactly like the old synchronous path.
- Writer apply failures are logged and never raise into a turn; the
writer thread survives and keeps applying.
Call sites switched to the queue: the per-call site in
agent/conversation_loop.py and both codex app-server sites in
agent/codex_runtime.py. In-memory per-turn counters
(agent.session_estimated_cost_usd etc.) stay synchronous, so live turn
displays never see the queue.
Tests: tests/agent/test_async_token_accounting.py (19 tests: enqueue
ordering, absolute-as-barrier, backlog coalescing with exact sums,
coalesced-vs-sequential row equivalence, merge unit rules, None-cost
preservation, read-your-writes, flush vs stop-flagged/concurrent
drains, inline apply after writer stop, close/atexit durability,
_persist_session drain, writer failure isolation);
tests/run_agent/test_token_persistence_non_cli.py updated to the
queue_token_counts contract.
With HTTP(S)_PROXY (and similar mounted transports), live connections sit
on client._mounts rather than the default _transport. force_close_tcp_sockets
only walked the default pool, so stranger-thread interrupt abort logged
tcp_force_closed=0 and left the request alive for minutes (#72975). Also
walk nested proxy _connection wrappers and WARNING when an abort finds no
sockets.
The auto-continue recovery note was typed only after run_conversation
returned, so its row sat untyped for the whole turn — and permanently
when the continuation was itself killed, which is the case it exists
for. persist_user_display_kind stamps the type on the live message
before the crash persist writes it, in the same insert as the content.
The flush also carries display_metadata through, which it was dropping.
A slash-skill invocation is persisted expanded — activation note plus the
entire skill body. _history_to_messages is the single display projection
every surface reads, so that payload rendered as a chat bubble anywhere a
session was resumed.
Project it here onto the invocation the user typed, and tag skill/bundle
dispatches with the same string so the live send matches. Rewind and
regenerate replay from what the transcript shows, so re-expand the
invocation server-side before running the turn: the replayed prompt is
identical to the original and no client ever holds the body.
A mid-stream steer persists an interrupted-turn checkpoint so the model knows
its reply was cut off. That scaffolding — "[This response was interrupted by a
user correction.]" and the "Visible response before the interruption:" header —
was written straight into message content, so every reload painted the raw
machinery as an assistant bubble (and merged it into the preceding tool-call
bubble). Steered transcripts became unreadable.
Reuse the existing display/replay split instead of inventing new surface:
- Carry the scaffolded form in the server-only api_content sidecar (the exact
bytes replayed to the provider), keep content the user's/agent's real words.
- When nothing reached the screen there is no clean form, so mark the row
display_kind=hidden — replayed to the model, dropped by every transcript
surface, exactly like compaction-reference rows.
- Honor display_kind=hidden in the gateway's _history_to_messages projection
(it only sniffed the [System: convention), so the checkpoint can't leak
through the live/resume path to the TUI/CLI either.
The model still receives the full interrupted context on the wire; the
transcript shows the partial reply and the user's correction.
A mid-stream steer/redirect cancels only the live model request and queues
the correction for a rebuild. But the retry-wait, error-handling, and
backoff-wait paths all treated the cancellation bit as a hard stop:
clear_interrupt() destroyed the pending correction and the turn died with
"Operation interrupted…" — the user's message silently lost. All three
sites now preserve the redirect and rebuild the iteration from it, exactly
like the InterruptedError handler.
tui_gateway also gets the two suppressions its sibling surfaces already
had: the "Operation interrupted: waiting for model response (…)" sentinel
is cancellation metadata and no longer ships as assistant prose in
message.complete (gateway/run.py and ACP already suppress it), and a
leftover pending_steer returned by the turn is requeued as the next prompt
instead of dropped (cli.py and gateway/run.py already do this).
session.steer now records the correction on the inflight turn like
session.redirect does, so a resume/reconnect mid-turn rebuilds the steered
user bubble instead of losing it.
Expanding an @file:/@folder: reference deleted the token from the message
it was typed in, leaving a hole in the sentence and no anchor for clients
to chip. The attached context block still names each ref, so nothing is
lost by leaving the token where the user put it.
The concept 'never send a turn that strict wire validation rejects as
empty' was forked across four sites, each with its own predicate and its
own blind spots:
1. build_assistant_message write-time ' ' pad — broke codex commentary
turns (content:'' is a designed state), and a DB-side pad can't
survive _rows_to_conversation's whitespace strip anyway. REMOVED.
2. conversation_loop send-time ' ' pad — main-loop only (summary path
uncovered), ordering-fragile (had to run after whitespace
normalization), assistant-only. REMOVED.
3. stream-stub '[response interrupted]' substitution — defeated the
loop's empty-stub guard (the stub no longer looked empty, entered
history, and the placeholder leaked into the stitched final
response via truncated_response_parts). REMOVED.
4. repair_empty_non_final_messages in sanitize_api_messages — the
unconditional pre-send chokepoint shared by the main loop AND the
summary path, covers user and assistant turns, non-final only,
copy-on-write. This is now the SINGLE OWNER.
The owner's payload predicate (_msg_has_payload) is extended to treat
codex_message_items / codex_reasoning_items as payload, so
designed-empty codex commentary turns are never rewritten on any
api_mode — the failure shape that broke site 1 in CI is encoded in the
owner, not special-cased at a call site.
Tests updated to pin the new contracts: builder stores textless turns
as-is; the empty stream stub stays recognizably empty for the loop
guard; poisoned resumed histories are repaired to the placeholder at
the send boundary; codex item carriers are never rewritten.
Sabotage-verified: unwiring the owner fails 3 regression tests.
Third layer of the empty-stub fix: full self-recovery. A poisoned transcript
(empty assistant stub or empty user turn already persisted before the write-
time guard, or fed in from a host history) previously 400'd every subsequent
request until it scrolled out — needing a manual DB edit + gateway restart.
sanitize_api_messages() (the unconditional pre-send chokepoint) now repairs
empty non-final messages on the per-call copy by substituting a minimal
'[response interrupted]' placeholder, so the session recovers itself IN MEMORY
on the very next send. The final message is left untouched (empty final
assistant is legal); stored history is never mutated; reasoning-only and
tool_call turns are preserved (negative controls).
Tests: production-shape repro (tool -> empty assistant -> user), empty-user
case, non-destructive guarantee, and negative controls. RED verified by
disabling the wire-in.
Add distinct, greppable log lines at both fix sites so the condition is
observable in the field instead of only inferable from the absence of the old
'Cannot compress further' spiral:
- chat_completion_helpers: warn when an empty partial-stream stub is replaced
with the placeholder (0 chars recovered, no tool call).
- error_classifier: warn when a malformed-body 400 is classified as
format_error rather than context overflow, with num_messages/approx_tokens.
Extend the litellm-shape classifier test to assert the warning is emitted.
A stream that dies after delivering deltas but with 0 recovered chars (and
no captured tool call) produced a content-less assistant stub. That empty
non-final message is invalid for the Anthropic schema, so every later request
400s with 'all messages must have non-empty content' (INVALID_REQUEST_BODY).
On a large session the 400 was misclassified as context overflow, dropping the
loop into a compression spiral that ends in 'Cannot compress further' — a
misleading context-size error on a session nowhere near its limit.
Root cause: substitute a minimal '[response interrupted]' placeholder so the
stub is never empty.
Misclassification: add _INVALID_MESSAGE_BODY_PATTERNS (checked before the
context-overflow heuristic) to classify these as non-retryable format_error,
and teach the body extractors the litellm/Bedrock errorMessage/errorCode/
errorArgs shape so descriptive proxy errors are not mistaken for generic ones.
Adds RED-verified regression tests for the stub path and three classifier tests
(including a guard that real overflows still compress).
Commentary-phase Codex turns persist with content:'' by design (their
text is delivered via the interim assistant callback), and the Responses
wire has no 'assistant must not be empty' validation — padding them
broke test_run_conversation_codex_continues_after_commentary_phase_message
in CI. Both the builder pad and the send-time pad now skip
api_mode=codex_responses. Keying on the ACTIVE api_mode preserves the
repair for codex-written sessions replayed through a strict
chat-completions provider.
A mid-tool-call stream drop with no delivered text produces a
partial-stream stub carrying content:'' and tool_calls=None. The
conversation loop's truncation path appended it to history as
{"role":"assistant","content":""} before the continuation nudge, and
strict providers (Moonshot/Kimi via OpenRouter) reject empty assistant
content with HTTP 400 ("the message at position N with role 'assistant'
must not be empty") on the next replay. Because the message is
persisted, every subsequent turn re-failed — the session was
unrecoverable.
Three layers, smallest blast radius first:
1. conversation_loop (length path): an EMPTY partial-stream stub is no
longer appended as an interim assistant message; only the
continuation user-message is. Stubs that delivered partial text are
still persisted so continuation stitching is unchanged.
2. chat_completion_helpers.build_assistant_message: never serialize a
textless assistant turn with content:'' — pad to a single space, the
same trick as the reasoning_content pad (#15250, #17400). Tool-call
turns are exempt (content:'' alongside tool_calls is accepted
everywhere).
3. conversation_loop send boundary: pad a textless assistant turn's
empty content to a single space AFTER all content-mutating passes
(surrogate sanitize, whitespace normalization, thinking-only drops),
before token estimation. This is the durable repair for sessions
ALREADY poisoned by older builds: the persisted stub rows are rebuilt
to '' on every reload (_rows_to_conversation strips whitespace, so a
DB-side pad can't survive) and only a send-time pad repairs them.
Verified: 485 tests pass across the four affected files; live replay of
a real poisoned session's resumed history against Moonshot via
OpenRouter returns HTTP 200 (was HTTP 400).
Direct tool HTTP calls already identified as Hermes-Agent, but the main
OpenAI-SDK chat path still sent OpenAI/Python. Set Hermes-Agent/<ver> for
api.x.ai clients (xai + xai-oauth) so normal text traffic is attributed correctly.
A /steer redirect during a thinking phase serialized the streamed
reasoning into the persisted assistant checkpoint ('Reasoning shown
before the interruption: ...'). An assistant turn exposing its own
chain-of-thought reads to Anthropic's output classifier as
reasoning-injection/prefill jailbreak, so every subsequent call on the
session deterministically returned 'Provider returned an empty
response' — and because the checkpoint is persisted and replayed, no
retry, nudge, or empty-recovery branch could ever escape it. Four
sessions were permanently bricked this way in the week of Jul 21-27
(42+ blocked calls; every reasoning-free checkpoint that week was
untouched — same mechanism as the prefill.json incident).
Class fix: streamed reasoning is now display-only state. The
_current_streamed_reasoning_text accumulator is removed entirely
(producer in _fire_reasoning_delta, resets, and init), so no future
path can serialize CoT into replayable content. The checkpoint keeps
only the visible response text; the model regenerates its reasoning on
the retried turn. Invariant documented in _apply_active_turn_redirect.
Regression tests: CoT never appears in either checkpoint shape,
reasoning-only interrupts produce a bare checkpoint, reasoning deltas
stay display-only.
The read-only org mirror broke the learning loop precisely where it matters
most. The system prompt tells every agent to patch a skill the moment it
finds a gap, and shared skills are the ones the most people use — but every
write to _org/ was refused, and the curator was excluded from them outright.
So org skills froze while personal skills kept improving, and the offered
alternative ("fork it into a personal skill, then propose the fork") is not
something an agent does mid-task. The refusal WAS the feature; improvements
were simply lost, and manual forks would have fragmented the shared set.
Edit in place:
- skill_manage patch/edit/write_file now work on org skills. Only delete is
still refused (the mirror is a view of org HEAD — a local delete returns on
the next pull; removing a shared skill is an admin action).
- Org skills are curation-eligible again, so the curator can improve the
highest-leverage skills in the system instead of skipping them.
- The load-time provenance header now says edits are allowed and kept,
instead of instructing the agent not to edit.
Local edits are never overwritten:
- pull_org_skills previously rmtree'd each skill dir and re-materialized it,
silently destroying local work on the next session start. It now records a
content fingerprint per skill (.org-baseline.json) when it writes one, and
SKIPS any skill whose local content diverges from that baseline.
- When upstream ALSO changed such a skill, it is reported in the pull
result's "conflicted" list and left untouched for the user to resolve
deliberately (propose the local version, or delete it and re-pull to take
theirs). A missing baseline is treated as unmodified so pre-existing
mirrors do not raise phantom conflicts.
- Fingerprints are content-based (path + bytes, sorted), so a touch/mtime
change is not mistaken for an edit.
Sharing back:
- Default: the edit stays local and the tool result tells the user to run
"hermes skills propose <skill>".
- Opt-in sync.org_auto_propose / HERMES_SYNC_ORG_AUTO_PROPOSE submits each
edit immediately. Defaults OFF — pushing every agent edit to a whole
organisation is not a safe default. A failed submission never fails the
edit; the change is saved and can be proposed later.
- "hermes sync status" lists org skills with unshared local edits;
"hermes sync pull" reports conflicts it declined to overwrite.
Tests: 25 in the namespace suite (was 15). The two that asserted the old
read-only behaviour now assert the opposite. New coverage for edit-applied,
share-back guidance, delete-still-refused, curation-allowed, edit detection,
missing-baseline tolerance, mtime-insensitivity, and the auto-propose
default. 428 passed / 0 failed via scripts/run_tests.sh.
Verified through the REAL pull path against a mock plane: pull v1 -> edit in
place -> upstream ships v2 -> pull leaves the local edit intact, reports the
conflict, and surfaces it in status.
The call-block decoration reads agent._use_prompt_caching / _cache_ttl /
_use_native_cache_layout directly; the redecoration helper wrapped each in
getattr with divergent defaults (e.g. or-'5m' vs verbatim _cache_ttl).
The flags are unconditionally initialized on AIAgent, so the defaults
served only test fixtures and would mask a real init bug as silent
cache-off. Align with the house style.
_peel_moa_guidance hand-implemented the inverse of moa_loop's
_attach_reference_guidance from a different module — a drifting separator
or shape would make the peel silently no-op and put the last cache
breakpoint on the turn-varying guidance block (the #72626 bug class).
Move the inverse into moa_loop.peel_reference_guidance directly adjacent
to the attach, keep a thin wrapper in conversation_loop, and pin the
contract with a round-trip test over all three attach shapes.
Also fix the empty-list residue: peeling a guidance-only content part now
drops the whole message (mirroring the appended-user-message shape)
instead of leaving an empty-content user turn behind.
guidance=None is a real prepared shape (all references failed / silent
degraded policy builds prepared_request without attaching guidance), and
the MoA facade sends prepared['messages'] — not api_kwargs['messages'].
Gating the rebase on 'and guidance' left the stale decoration in the
prepared object for the no-guidance MoA sub-path, so #72626 persisted
there. rebase_prepared_request already handles falsy guidance (copies
messages, skips the attach).
The static-prefix reconstruction pattern (build_system_prompt_parts ->
['stable'] -> startswith gate -> fail-open) existed in three copies:
session restore (conversation_loop), compression keep-prompt path
(conversation_compression), and the new failover redecoration helper.
Hoist it into agent/system_prompt.reconstruct_static_prefix and call it
from all three sites.
Also memoize failed rebuilds per stored prompt (_static_rebuild_failed_for):
the redecoration chokepoint runs at the top of every retry attempt, and a
persistent stable-tier mismatch (restored session whose SOUL.md/skills
changed since save) would otherwise re-run the full prompt build — SOUL.md,
context files, memory I/O — on every attempt of every API call for the
life of the session. A legitimately changed stored prompt retries once.
strip_anthropic_cache_control flattened ANY pure-text multi-part content
list with a separator-less join. Decoration only ever produces a single
text part or the 2-part [static, volatile] system split; organic
multi-part text (merged user turns, imported transcripts) got word-jammed
and parts carrying extra keys (citations) were silently dropped — on the
common no-failover path, since redecoration runs on every attempt.
Restrict the flatten to the exact decoration-produced shapes and make
marker removal copy-on-write on part dicts (the per-call message copy is
shallow, so parts alias the persistent history).
try_activate_fallback refreshes the cache policy flags for the new
provider, but the retry loop reused the primary's decorated api_messages.
Cache-off→cache-on shipped zero breakpoints; cache-on→cache-off left
stale markers. Strip and re-render at each retry attempt (same chokepoint
as reasoning-echo reapply), peel/rebase MoA guidance so the last marker
stays off the turn-varying block, and rebuild the static system prefix
when caching becomes active mid-turn (#72626).
Three code-reuse fixes applied during salvage:
1. Reuse _relative_time from hermes_cli/main.py instead of duplicating
the relative-time formatting logic in hermes_cli/status.py.
2. Extract _stamp_hygiene_compression_provenance helper in gateway/run.py
to deduplicate the two nearly-identical try/except blocks that stamp
compression timeout/abort provenance in the hygiene path.
3. Add ContextCompressor.record_timeout_failure() method and use it from
the in-agent compress_context timeout callback instead of re-implementing
the (60, 300, 900) cooldown ladder inline. The existing summary-LLM
exception handler already has this ladder — now both paths share one
method.
Three mechanisms to detect and notify when gateway sessions stall silently:
1. Mid-turn activity heartbeats stamped to SessionDB so hermes sessions list
and hermes status show progress during long turns without new message rows.
2. Stall watchdog: when a busy session has pending inbound and the shared
activity clock is idle past agent.session_stall_timeout (default 300),
log a WARNING and notify the user once to try /new. Notify-only; does
not kill the turn.
3. Compaction timeout: fenceless compress_context callers get a progress-aware
host budget (compression.context_timeout_seconds default 120 idle,
compression.context_total_ceiling_seconds default 600 ceiling). On timeout,
cancel via commit fence, skip compaction without dropping messages, and
continue the turn.
Closes#72016 (slices 1-3; slice 4 cumulative SSE stream-retry deadline
remains a follow-up).
Cherry-picked from PR #72424 by @fangliquanflq.
- codex app-server sibling path: surface a WARNING (was silent debug) when
the projected-message flush fails — same bug class as the main fix, but
codex output has already streamed so fail-closed and agent_persisted=False
are both wrong here (#860/#42039 duplicate-write hazard); loud durability
gap logging instead.
- map session_persistence_failed in _format_turn_completion_explanation so
the user sees an actionable reason instead of 'The request failed:
unknown error' + explainer test.
- contributors/emails mapping for elco@thedaoist.gg (attribution CI).
claude-fable-5 is a Mythos-class reasoning model (1M context, 128K output,
adaptive thinking per anthropic_adapter.py) but was missing from the
_REASONING_STALE_TIMEOUT_FLOORS table. Without a floor entry it got the
default 180s stale timeout (300s with context scaling), which is too short
for fable-5's thinking phase on large contexts.
Each stale kill bumped the cross-turn circuit breaker streak; after 5
consecutive kills _check_stale_giveup() fired immediately (elapsed: 0.00s),
aborting all calls with "Provider has been unresponsive for 5 consecutive
stale attempts." Users with 191K-token contexts hit this reliably.
Add ("claude-fable", 600) — deep-reasoning tier alongside o1/deepseek-r1/
nemotron-3-ultra. The claude-fable slug matches claude-fable-5 and future
variants via the existing right-anchor regex.
Update _pre_msg_count after adopting the durable transcript so the
post-compression log reflects the correct pre-adoption message count.
Also use the existing _live_child_id() helper in the updated test
instead of hand-rolled child-id extraction.
Follow-up to #72631.
is_truthy_value(..., default=False) and getattr(..., False) disagreed with
compression.in_place: true from #38763, so partial/failed config loads fell
back into rotation mode and re-armed the pre-lease drift path. Also report
compression.in_place in hermes dump overrides so stale false values are visible.
Busy sessions (memory review / shared session writers) kept outrunning the
in-memory snapshot, so rotation-mode compress aborted every attempt with
"changed before lease acquisition" and surfaced as a fake "No changes from
compression". Adopt the durable transcript and continue compressing instead
of returning the stale snapshot unchanged.
Every fallback/dedup/skip decision asks one question — 'is this candidate
the same backend as the one that failed, along the axis that failure
invalidated?' — but it was re-implemented inline at six sites across four
subsystems, each comparing whatever string was locally convenient. Each
incident fixed one site while the others kept the bug: #22548, #70893,
#59561, #72468, #62984/#54250/#57584.
agent/backend_identity.py now owns the concept: BackendIdentity (provider /
model / base_url axes), FailureScope (MODEL / CREDENTIAL / ENDPOINT — each
failure class invalidates a different axis), and should_skip_candidate().
Unknown axes never manufacture a skip (over-skipping strands failover; a
wrong try costs one RTT).
Migrated sites:
- chat_completion_helpers.try_activate_fallback: replaces the provider+model
early-exit (the #62984 bug: ignored base_url, stranding multi-endpoint
pools) AND _fallback_entry_is_same_backend_by_base_url (deleted)
- auxiliary_client._try_configured_fallback_chain +
_try_main_agent_model_fallback: replace label/model comparisons; auth and
payment map to CREDENTIAL scope, keeping the #59561 carve-out
- hermes_cli/fallback_cmd add: primary-match + duplicate checks now identity-
aware (#54250/#57584): same provider+model on a different explicit
base_url is a pool entry, not a duplicate
_mark_provider_unhealthy stays label-keyed deliberately: its only triggers
are confirmed 402s, which ARE credential-scoped.
Owner-level tests pin each incident's semantics by number; sabotage-verified
(removing the base_url axis fails the #62984 test).
Two defects found by manual testing on the branch.
1. ORG SYNC NEVER RAN. The org pull/mirror/gating machinery was fully
implemented and unit-tested but had ZERO runtime callers —
maybe_pull_org_skills() was referenced only inside a comment, and
`hermes sync` had no org path at all. Every code path fell through to
personal sync (refs/user/<sub>/), so org skills never loaded and the
feature looked like 'everything syncs to my personal org' even with a
valid org token. The unit tests could not catch this: they invoked the
functions directly, which is exactly the gap they left open.
- cli.py session startup now calls maybe_pull_org_skills() alongside the
personal maybe_pull_skills(), fail-quiet.
- Auto-pull is gated on real org membership: resolve_org_identity()
requires an org role on the token, only issued for multi-member orgs,
so a solo account never reaches the network.
- `hermes sync pull` refreshes the org mirror too (one pull, both
surfaces) and reports what it refreshed.
- `hermes sync status` exposes org_available/org_id/org_role/org_skills
plus a plain-language summary, so a user can tell whether the org
workflow applies instead of it being invisible.
2. INTERNAL JARGON LEAKED TO USERS. Help text and errors exposed internal
milestone/spec coordinates: 'Propose a skill ... (M2)', 'Personal skill
sync (HSP/1)', 'DEV-PHASE gate closed: your token lacks
tool_gateway_admin', 'contract §4.3', and an inert message describing our
internal personal-vs-multi-member design split. All rewritten in user
language. Feature-local comments/docstrings lost their internal
coordinates (§N, M1/M2, design.md, PR numbers) while keeping the
explanatory prose. Pre-existing issue references elsewhere in the tree
were deliberately left untouched.
Tests: 4 new guards, including two that assert the CALL SITES exist so the
org pull cannot silently become dead code again (verified failing when the
wiring is removed) and one that fails if user-facing help leaks jargon.
344 passed across the sync/skills/prompt suites.
Verified against live staging with a real org token: sync status reports
org_available=true, org_role=OWNER; sync pull performs the org refresh; the
.active_org marker is written with the org id from the token.
Follow-ups on the salvaged #63359:
- _finalize_child_results carries tool_call_history on subagent_stop
(the #62011/#72403 field landed after the PR branched; the shared
pipeline must emit it for both delegate_task and plugin-launched
children). Lifecycle test updated for the new payload field.
- The lifecycle executor uses DaemonThreadPoolExecutor — a wedged or
abandoned child must never block interpreter exit at atexit-join time
(same rationale as _run_single_child's timeout executor and the
async-delegation pool).
- delegate_task's batch path keeps live-transcript wiring while routing
child construction through the shared
_build_child_preserving_parent_tools helper.
Widen okalentiev's failed_model narrowing (#59561) to
_try_main_agent_model_fallback. The safety-net layer still skipped on a
provider-label match alone, so single-provider users whose aux compression
model and main model share one custom endpoint had ZERO fallbacks: the aux
model timing out exhausted the chain in one hop and compression aborted.
Real incident (0.19.0 debug dump): aux zai-org/glm-5.2 hung 324s and timed
out while main mindai/macaron-v1-venti on the SAME endpoint was serving
448K-token turns — the label-only skip discarded the one viable summarizer,
the session wedged over threshold, and the anti-thrash breaker tripped.
Same convention as the chain fix: model-specific failures (timeout,
connection, rate limit) pass failed_model so only the exact failed model is
skipped; provider-wide failures (auth 401 / payment 402) pass None and keep
the whole-provider skip. Both sync and async call_llm sites pass it.
Sabotage-verified: the new regression test fails on the provider-only skip.
Follow-up to the same-provider fallback fix: narrowing the configured-chain
skip to the exact failed model is only correct for model-specific failures.
Auth (401) and payment (402) errors are provider-wide — every model on the
provider shares the same broken credentials/account — so trying a sibling
model can't recover and merely burns another doomed request before the
aux task fails. Worse, returning that sibling client bypasses the
main-agent-model safety net that a provider-wide skip would have reached.
Only forward failed_model to _try_configured_fallback_chain for
model-specific failures (timeout, connection, rate limit, model-incompatible,
invalid response). Auth/payment keep failed_model=None (whole-provider skip),
preserving the pre-existing safety-net behaviour for credential/billing
failures.
Adds an integration test that a timeout forwards the failed model, and
updates the payment-error test to assert failed_model=None.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
_try_configured_fallback_chain skipped every fallback_chain entry whose
provider matched the one that just failed. A chain that intentionally lists
several models under the same provider (e.g. two more NVIDIA NIM models
after the primary NIM model times out) was therefore skipped wholesale,
falling straight through to the main-agent-model safety net instead of
trying the other configured models on that provider.
Add failed_model so the skip narrows to the exact (provider, model) pair
that failed. Callers that only know the provider (client-build failures,
where the whole provider is unreachable regardless of model) keep the old
provider-wide skip; the two runtime request-error call sites (call_llm,
async_call_llm) now pass the model that just failed.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Per Teknium: the caps should bound a single agent loop, not accumulate
over the whole session. Rename SessionCapConfig -> LoopCapConfig and the
config section session_caps -> loop_caps; move the counters into
reset_for_turn (invoked per turn via turn_context) so each turn starts
with a fresh budget; retune defaults 200 -> 50 (a single turn issuing 50
web searches / spawning 50 subagents is already pathological). Block
codes session_*_cap -> loop_*_cap and messages updated to drop the
/new-resets-the-budget guidance (irrelevant now that it resets per turn).
Tests flipped: the old persists-across-turn-resets assertion becomes
resets-each-turn.
Add per-session lifetime caps on web_search calls and subagent spawns
(defaults 200/200, matching Claude Code v2.1.212). Unlike the existing
per-turn tool-loop guardrails, these count over the whole session and
reset only when a fresh agent is built (/new, /clear). Hitting a cap
blocks the offending call and halts the turn cleanly.
- agent/tool_guardrails.py: SessionCapConfig + session counters on the
controller (in __init__, not reset_for_turn, so they persist across
turns). before_call() enforces caps first, independent of
hard_stop_enabled. delegate_task batches count each task.
- hermes_cli/config.py: tool_loop_guardrails.session_caps defaults.
- docs + tests (unit + E2E validated against a real AIAgent).
Port from nearai/ironclaw#5149 (the describe-first live-hardening fix in
their progressive tool disclosure work): when a model invokes a deferred
tool through the tool_call bridge without the schema-required arguments,
return the tool's parameter schema instead of dispatching blind.
Pre-fix, a blind call produced an opaque downstream failure
("[TOOL_ERROR] Tool execution failed: KeyError: 'document_id'") that
teaches the model nothing about what the tool expects — IronClaw observed
cheap models looping ~30 identical invalid calls until the iteration
budget died. Post-fix, the model repairs the call in one round-trip.
- tools/tool_search.py: new validate_deferred_call_args() — key-absence
check of schema 'required' fields only; no type checking (coerce_tool_args
already repairs types downstream); fails open on any validator error so
it can never block a legitimate dispatch.
- model_tools.py: probe after the scope gate in the bridge dispatch.
- agent/tool_executor.py: probe in both unwrap sites (concurrent +
sequential) before the underlying tool replaces the bridge; sequential
path flattens the payload to match its {"error": str} wrapping.
- tests: TestDeferredCallSchemaProbe — blind call returns schema (not
KeyError), valid/optional calls dispatch, unvalidatable tools fail open,
out-of-scope rejection unchanged.
When the empty-response ladder is fully exhausted (thinking-prefill
continuation, empty-content retries, provider fallback) and the model
produced structured reasoning but never any visible text, deliver a
clearly labeled excerpt of that reasoning instead of a bare '(empty)' —
the reasoning frequently contains the actual answer.
Delivery-only by design: raw chain-of-thought is never promoted to a
normal answer earlier in the ladder (prefill continuation still gets
first crack, retries and fallback still run), transcript persistence
semantics are untouched (the '(empty)' sentinel scaffolding keeps its
replay-safety behavior), and a truly empty exhaustion still returns the
existing terminal.
Idea credit: PR #48795 (@ligl0325) proposed falling back to
reasoning_content on empty content; this lands the safe kernel of that
idea at the one point in the ladder where it is strictly an improvement.
Port from nearai/ironclaw#6129: their sensitive-marker scrubber matched
markers as bare substrings, so tool results containing 'Secretary of the
Treasury' were scrubbed as 'secret' on replay, evicting legitimate content
and forcing the model into a re-fetch loop. Hermes' lowercase/dotted/YAML
config-key redaction patterns (_CFG_DOTTED_RE, _CFG_ANCHORED_RE,
_YAML_ASSIGN_RE) had the same false-positive class: their key classes allow
arbitrary alphanumeric affixes around the keyword, so ordinary document
text like 'Secretary: J.Smith', 'tokenizer: cl100k_base' (HF model cards),
and BibTeX 'author=Smith' got value-masked on the surfaces that run these
passes (browser snapshots, log lines, kanban summaries, CLI-echoed output).
Fix: post-match word-boundary validation of the keyword occurrence inside
the matched key. Boundaries: key edges, non-letters (_ - . digits),
camelCase transitions (clientSecret, secretKey, APIToken), plural 's'
(secrets:, tokens:). Concatenated real-world compounds keep matching via
explicit alternatives (authtoken, authkey, secretkey, accesstoken). ALL-CAPS
keys keep legacy embedded matching (MYTOKEN=...) — all-caps is almost never
prose, same rationale as _ENV_ASSIGN_RE. Same discipline the file already
applies to exact-match body/query keys (ported from ironclaw#2529) and the
deliberate 'auth' exclusion that keeps 'author:' from matching.
Port from openclaw/openclaw#110518 / #110956: some models reuse one call id
for different tool calls in a single batch (native Kimi Responses replays,
Ollama-compatible endpoints, degraded models at long context). Hermes kept
both calls but the pre-API sanitizer then dropped the later call/result pair
per id (#58327), so the second call's output silently vanished from every
replayed payload — the model never saw it and confabulated.
_uniquify_tool_call_ids renames later collisions to a deterministic <id>_d<n>
suffix at ingestion, before validation/dispatch/history build, so both pairs
survive. Composite Responses ids collide on the call half and keep their
response-item half. Deterministic suffixes preserve prompt-cache prefix
stability (no random UUIDs).
Port from Kilo-Org/kilocode#12162.
Google began rejecting unrestricted legacy 'Standard' Google Cloud API keys
on the Gemini API on June 19, 2026 (all Standard keys stop working in
September 2026). The rejection is a 401 whose message misleadingly tells the
user to supply an OAuth 2 access token. gemini_http_error() now appends
actionable guidance (check key type in AI Studio, mint a new Gemini API key,
temporary restriction bridge) on that narrow shape — matched via
google.rpc.ErrorInfo reason ACCESS_TOKEN_TYPE_UNSUPPORTED or the
'Expected OAuth 2 access token' signature. Plain invalid keys
(API_KEY_INVALID) keep their existing message.
Also fixes a latent sibling gap: _summarize_api_error() preferred re-extracting
the raw response body for errors carrying .response, which stripped adapter-
composed guidance (this one AND the existing free-tier 429 guidance) from the
user-facing summary. GeminiAPIError now surfaces its composed message.
Port from anomalyco/opencode#37848 (+ dev-branch twin #37840): expand
context-overflow patterns and guard against rate-limit messages that
mention tokens.
- 'Throttling error: Too many tokens, please wait before trying again.'
(AWS Bedrock / proxy shape) classified as context_overflow and routed a
healthy session into compression on every throttle. Added 'throttling'
to _RATE_LIMIT_PATTERNS, which the message-only path checks BEFORE the
overflow list.
- 'Input length N exceeds the maximum allowed input length of M tokens.'
(Together/Fireworks shape) fell through to unknown — no compression
recovery. Added 'maximum allowed input length' to overflow patterns.
- 'request_too_large' / 'Request exceeds the maximum size' (Anthropic 413
type re-wrapped without a status code by aggregators/proxies) fell
through to unknown. Added to _PAYLOAD_TOO_LARGE_PATTERNS.
All three shapes proven live on main before the fix; 265 classifier +
bedrock tests and 238 sibling rate-guard/compression tests pass.
Root cause: delegate_task children run through three nested daemon-thread
layers (async-delegation executor -> per-child timeout executor -> the
interrupt worker interruptible_api_call spawns). After multi-day gateway
uptime the deepest layer wedges BEFORE the socket opens — the same
fingerprint as the gateway-cron hang (#62151): zero stale-detector output
(the worker never reaches dispatch), all providers, foreground/restart
works. The cron fix (should_use_direct_api_call) explicitly excluded
delegation 'for lack of evidence' — #60203 is that evidence.
- should_use_direct_api_call: extend the inline gate to delegated
children, detected via the delegation ContextVar set by
_run_single_child (platform='subagent' stamp as fallback). Scope
unchanged otherwise: chat_completions wire only; Codex/Anthropic/
Bedrock/MoA keep their established workers. Interrupts still work —
the inline path registers _active_request_abort, which interrupt()
invokes cross-thread (same mechanism the #72227 stall monitor uses).
- _dump_subagent_timeout_diagnostic: dump ALL thread stacks (bounded,
40), not just the conversation worker — a pre-HTTP wedge is
indistinguishable from a slow provider without seeing where the
nested helper threads sit.
`apply_anthropic_cache_control` runs once per call block, before the retry
loop, and splits the system prompt into `[static prefix, volatile tail]` text
blocks carrying the cache_control breakpoints.
A failover fires *inside* that retry loop, and `_sync_failover_system_message`
assigns a bare string over `api_messages[0]["content"]` to refresh the
`Model:`/`Provider:` identity lines. That drops the block list and both
breakpoints. `convert_messages_to_anthropic` only emits system cache blocks for
list content (`isinstance(content, list)`), so the retried request ships
`system` as one plain string: zero breakpoints, nothing written to cache, and
the whole system prompt re-billed at full write price. The next call misses
too, so a failover costs two full-price prompts instead of one write + one
read.
Measured with the real functions, same history, native Anthropic layout:
normal turn system=BLOCK LIST(2) system breakpoints=2
after failover system=plain string system breakpoints=0
`_sync_failover_system_message`'s own docstring notes this fires on "every
gateway turn, since fallback re-activates per message while the primary is
down" -- so a gateway running on a degraded primary pays it on every message.
Fix: rewrite the decorated blocks in place instead of flattening them.
`rewrite_prompt_model_identity` only touches the LAST `Model:`/`Provider:`
lines, and those live in the volatile tail, so the static prefix stays
byte-identical and its cache entry keeps matching -- the failover retry keeps
both breakpoints AND the warm prefix. Shapes we cannot safely patch fall back
to the existing plain-string assignment.
This is the same class the retry loop already guards against eight lines below
the gap, for reasoning fields:
# api_messages is built once, before this retry loop, while the primary
# provider is active. [...] so the fallback request isn't sent with stale,
# primary-shaped reasoning fields.
agent._reapply_reasoning_echo_for_provider(api_messages)
There was no equivalent for cache decoration.
`apply_anthropic_cache_control` marks an assistant turn with non-empty text by
writing `cache_control` INTO `content` -- `_apply_cache_marker`'s list branch
puts it on the last content block, not at the top level.
`_convert_assistant_message`'s ordered-replay branch rebuilds the message from
`anthropic_content_blocks` and returns early. Its only cache sources are
`_relocated_replay_cache_control` (markers rescued from blocks the replay
sanitizer dropped) and the top-level `m["cache_control"]`; it never reads
`m["content"]`. So for an assistant turn that interleaves signed thinking with
a tool_use AND has preamble text, the breakpoint is dropped.
It is burned, not relocated: `_can_carry_marker` returns True for this message
(non-empty content), so the breakpoint budget already counted it. Hermes'
accounting believes the marker landed.
Measured with the real functions, native layout, same history, only
`anthropic_content_blocks` differing:
normal path 4 breakpoints assistant text block carries cache_control
replay path 3 breakpoints assistant carries NONE
Under the static-prefix layout only two conversation-tier breakpoints exist, so
this halves them. Nothing is logged and the request succeeds with identical
model output, which is why it survives: it recurs on every request for the life
of any Claude thinking+tools session, since the flag rides the message in
`agent.messages`.
#56195 fixed the complementary shape -- blank assistant content, where the
marker lands top-level -- and its regression test pins `"content": ""`. The
non-empty case is the one a Claude 4.5/4.6 tool turn normally produces (a
sentence of preamble before the call) and was never covered.
Harvest an in-`content` marker next to the existing top-level lookup and hand
it to the same `_apply_assistant_cache_control_to_last_cacheable_block` helper.
Extends the cherry-picked /context command (PR #52184) and prompt-size
attribution helpers (PR #66656) into one visual context view across
surfaces, and absorbs the per-component budget-visibility goal of the
/tokens proposal (PR #48470):
- agent/context_breakdown.py: pure renderers over the existing payload —
a 5x20 glyph block grid (1 cell ~= 1% of the model window), an
'Estimated usage by category' table with free space, and expanded
per-skill / per-toolset listings via compute_context_details(), which
reuses the prompt-size attribution mechanism (skills index-line bytes +
registry tool->toolset map) converted to the same chars/4 heuristic.
- cli.py: /context [all] renders grid + category table (+ expanded
listings) from the live agent and in-memory conversation history.
- gateway/slash_commands.py: /context appends the plain-text category
table (no grid — monospace not guaranteed on messaging platforms);
/context all adds the expanded listings. Fail-open: breakdown errors
never break the gauge.
- hermes_cli/commands.py: /context gains the 'all' subcommand; /version
demoted to /hermes version on Slack to keep the 50-slash cap.
- tests: renderer unit tests against synthetic payloads, registry test,
gateway /context + /context all + failure-degradation handler tests.
- docs: slash-commands reference + CLI guide entries.
Read-only and locally computed: no provider calls, no prompt-cache impact.
Co-authored-by: RemyFevry <29257684+RemyFevry@users.noreply.github.com>
Co-authored-by: joelbrilliant <joelbrilliant1@gmail.com>
Co-authored-by: CharlesMcquade <6466275+CharlesMcquade@users.noreply.github.com>
Follow-ups to the salvaged #23768 commit, which targeted a pre-79559214
codebase:
- agent/tool_executor.py + agent/agent_runtime_helpers.py: pass
multi_select at both current clarify dispatch points (the PR's
run_agent.py edits landed on dead code paths).
- tools/clarify_tool.py: replace the broad TypeError-retry in
_invoke_callback with inspect.signature detection, so a compatible
callback that raises TypeError internally is not invoked twice
(addresses hermes-sweeper review feedback on #23768).
- tests: cover single-invocation on internal TypeError, legacy 2-arg
callbacks, **kwargs callbacks, and registry handler multi_select
pass-through (schema arg → handler → callback).
* feat(billing): add payment_method union to the billing-state wire type
* feat(billing): carry the payment-method union through the gateway
NAS now sends a typed `paymentMethod` union on /api/billing/state alongside
the legacy `card` field. The gateway parses payloads field-by-field, so the
new field was dropped on the floor before reaching TUI/Desktop.
Parse it into PaymentMethodInfo and re-emit it as snake_case `payment_method`,
matching the translation the rest of this payload already does. The payment
method id is deliberately not carried through — clients have no use for it.
No client rendering changes: `card` stays populated for cards, so every
existing consumer behaves exactly as before and the new field is inert until
a surface opts into reading it.
* fix(billing): send only the fields each payment-method kind declares
The serializer emitted every key for every kind, so a Link method went out
carrying brand, last4 and wallet set to null. That contradicts the shared
type, where each kind declares its own fields: a client testing `'brand' in
pm` would read every Link method as a card, and one trusting the declared
non-null `brand` could crash on it.
Send each kind's own fields, and forward an unrecognized kind by name alone
so a client that predates it can still say something honest. The shared type
gains the matching fallback arm its own comment already promised.
Tests now follow a payload from the server response through to the client
wire for each kind, rather than checking parsing and serializing separately —
which is why the old expectation locked in the wrong shape without noticing.
* fix(billing): keep the payment-method kind narrowable
Typing the fallback arm's kind as `string & {}` borrowed a trick that only
works on unions of plain strings. On a union of objects it makes the
discriminant non-literal, so TypeScript stops narrowing on every arm — even
`if (pm.kind === 'card')` no longer gives you `brand`. The first client to
use this would have hit a compile error and reached for a cast.
An unrecognised kind now arrives as `unknown`, carrying the real name
alongside it, so every arm has a literal discriminant. A type-level test
pins this: it fails to compile if the discriminant stops narrowing.
The parser settles which kind it is, the way the card parser already does,
so the record cannot hold fields that do not belong to its kind and the
serializer no longer re-checks. The type comment also stops claiming `card`
is a safe signal — it is null for Link, so `!card` does not mean "nothing on
file".
The default max_iterations/agent.max_turns budget was set when long
agentic runs were rare; complex tasks now routinely exceed 90 tool
calls. Raise the default to 500 across every surface that hardcodes
the fallback: AIAgent constructor, DEFAULT_CONFIG, CLI resolution
chain, gateway env bridge, cron scheduler, and TUI gateway. Explicit
user config values are unaffected (deep-merge preserves them; no
_config_version bump needed).
Docs (en + zh-Hans), CLI help text, tips, and pinned tests updated
to match.
generate_title() sent the first 500 characters of the user turn to the
auxiliary model. On a /skill invocation those characters are the skill's own
opening prose, so the session got named after the skill instead of the request
— /work sessions came back as "Isolated Git Worktree Setup".
Route the turn through describe_skill_invocation() first, so the titler sees
what the user typed. Also keep only the first line of the response: a model
that ignores "return ONLY the title" and answers the prompt would otherwise
have a shell transcript stored as the title, truncated mid-command.
A /skill invocation expands into a message that embeds the whole skill body.
Anything that summarizes a user turn from its raw content reads the skill's
prose as if the user had written it.
describe_skill_invocation() sits next to the existing extractor and reuses its
markers, returning `/work — fix the title leak` for an invocation with an
instruction and `/work` for a bare one. It also exports the SQL LIKE pattern
and excerpt-joint sentinel that listing queries need to recognize scaffolding
before a row reaches Python.
2107b86024 flipped compression.in_place to True but left both explanatory
comments reading "Default False during rollout". The contradiction is
load-bearing: it is why two recent PRs (#71747, #48951) were built on the
premise that agent.session_id still rotates at every compaction.
Comment-only; no behavior change.
The compression lock was gated so that any holder on a Windows host was
assumed alive until the full TTL expired. A crashed holder therefore
stranded every other agent on the session for the whole lease window.
Probe liveness with psutil.pid_exists, which is safe on nt. Keep the
os.kill(pid, 0) path strictly for POSIX: on Windows signal 0 maps to
CTRL_C_EVENT (bpo-14484) and can kill the target's console group, so with
psutil absent the only safe answer stays 'assume alive' and let the TTL
run out.
Also carry the commit fence through cancelled compressions so an aborted
run leaves the lock reacquirable rather than half-held.
The background write guard decided ownership from `isinstance(usage_rec, dict)`,
so a local skill with NO usage record passed. That successful write called
bump_patch(), which created a `created_by: null` record — and the identical
write was refused from then on. "Allowed exactly once, then never" is a race
with our own bookkeeping, not a policy. Reproduced on main: patch #1 succeeds,
patch #2 with the same arguments is refused.
Option B from the issue. Option A (split `session_review` from
`scheduled_curator` and let the session fork patch user-owned skills it
consulted) would widen autonomous write permission onto skills the user owns
with no user present to consent — wrong direction for a no-user-present actor.
- skill_manager_tool: missing and explicit-null records now resolve
IDENTICALLY, both fail closed. The refusal names the reason and points at
`hermes curator adopt <name>`.
- background_review: both review prompts told the reviewer to patch any skill
consulted in the session and claimed pinned skills could be improved, while
enforcement refused both. Prompts now list pinned, external, and user-owned
skills as protected, and tell the reviewer to RECOMMEND adoption instead of
attempting a write that will be refused.
- skill_usage: document that `created_by` is a curator-management policy flag,
not a provenance claim, and add `is_curator_managed()` so call sites read as
the question they ask. Field name retained — it is on disk in every
`.usage.json` and renaming would strand those records.
- curator CLI: `hermes curator list-unmanaged` itemizes unmanaged skills with
the reason each is unmanaged (completes the #67139 spec).
Foreground writes are untouched: a user-directed edit to a user-owned skill
still works, including on pinned skills.
Sibling tests: 9 failures in test_skill_manager_tool.py were fixtures that
created record-less skills to exercise OTHER guards (consolidation-delete,
read-before-write) and relied on ownership falling through. Fixed at the
fixture, since the real curator only ever operates on managed sediment. One
test asserted the old "manually authored" wording; rewritten to assert the
behavior contract instead of the string.
Validation: 274 targeted tests + all 7 background-review files (60 tests) pass.
E2E on a temp HERMES_HOME (30 checks) covers the flip, foreground writes,
adoption unblocking, pin semantics, prompt/enforcement parity, and the new verb.
Each new test sabotage-verified: revert the fix, confirm it goes red.
Fixes#67140
The whole-prompt scan for "Current working directory:" matched USER project
content, not just Hermes' own host-info line. The prompt embeds AGENTS.md /
CLAUDE.md / .cursorrules in the context tier, which sits AFTER the host-info
block, and line_value() takes the LAST match — so any project file containing
a line starting with that label won the scan.
The comparison then ran runtime state against project prose, a mismatch that
never clears: the stored prompt was rejected on EVERY turn, rebuilding the
system prompt each message and destroying the prefix cache for the whole
session. Strictly worse than the staleness the check exists to catch, and
invisible to CI because no test encoded the invariant.
"last match wins" was only ever safe because Model/Provider/Platform live in
the volatile tier at the very END of the prompt. The cwd line is the one field
read from the STABLE tier near the start, so it needs an anchored read:
locate Hermes' own "User home directory:" line and take the working-directory
line that follows it.
Verified against the real builder on a real AIAgent, not a stub:
- stable cwd, clean project -> prompt REUSED (0 rebuilds)
- stable cwd, AGENTS.md naming the
field -> prompt REUSED (was: rebuild/turn)
- drifted cwd -> rebuilt
- drifted cwd, AGENTS.md naming the
NEW cwd -> rebuilt (prose can't mask drift)
Also re-verified no spurious rebuild across all five cwd-resolution paths
(pinned contextvar, TERMINAL_CWD, launch dir, symlinked workspace,
trailing-slash cwd).
The contributor's fixtures omitted the host-info anchor, so they silently
stopped exercising the cwd path once the read was anchored; reshaped them to
match real prompt structure via a _host_block() helper.
The progress-hook streaming from #71508 only activated when a
CompressionCommitFence was present (gateway session hygiene). CLI
/compress and in-loop auto-compression still used the plain
non-streaming summary call, where the SDK timeout is inactivity-based —
a byte-trickling provider that keeps the connection alive could outlive
auxiliary.compression.timeout indefinitely (the gap #69192/#41397 were
built to close).
Fenceless compression callers now install a no-op progress hook, which
routes their summary call onto the same streamed path: the configured
timeout acts on inactivity (slow models finish instead of being cut
off mid-generation), and a degenerate trickle stream is bounded by the
streamed total ceiling (max(600s, 4× the task timeout)) instead of
running forever. No config knob needed — the ceiling machinery ships
with the streaming layer and applies uniformly.
Supersedes the opt-in wall-clock deadline approaches in PR #69192
(@JabberELF) and PR #41397: same guarantee (bounded total compression
wall time even while bytes move) without a daemonized watchdog thread
or a new config surface, and without punishing slow-but-healthy models.
Some OpenAI-compatible endpoints — notably Tencent Copilot
(copilot.tencent.com) — only accept streaming chat requests; any
non-streaming call returns HTTP 400 (code 11101, 'Non-stream chat
request is currently not supported'). The main conversation loop already
streams, so interactive chat works, but every auxiliary task (title
generation, compression, web extraction) used the non-streaming path and
failed on each call.
_provider_requires_stream() detects stream-only endpoints
(copilot.tencent.com built in, plus user-configurable
auxiliary.stream_only_base_urls substring markers in config.yaml).
Matching sync auxiliary calls route through _create_with_progress
(force_stream=True) and async calls through the new
_acreate_with_stream, aggregating the chunk stream — including tool-call
deltas and reasoning deltas — into a complete response via the shared
_ChatStreamAccumulator.
Salvaged from PR #60686 by @kudi88 onto the progress-aware streaming
machinery from #71508, addressing both sweeper-review gaps: the async
path now consumes the stream with 'async for' (awaiting create() and
iterating synchronously raised on AsyncOpenAI streams), and tool-call
deltas are reassembled instead of dropped (MCP passes tools= through
call_llm). Under force_stream there is no silent non-streaming retry —
a stream-only provider rejects those by definition, so the original
error surfaces to the normal recovery chains.
The gateway's pre-agent session-hygiene compression killed the summary
call at a fixed 30s wall-clock deadline (compression.hygiene_timeout_seconds),
regardless of whether the summary model was hung or merely slow. A reasoning
model happily streaming a large summary was cut off mid-generation, the user
got '⚠️ Context compression timed out after 30.0s', and a 300s failure
cooldown left the session oversized — a doom loop for slow-but-healthy
auxiliary models.
Timeouts are now liveness-based instead of wall-clock-based:
- agent/auxiliary_client.py: new thread-local aux_progress_hook. When
installed (only by context compression today), the primary call_llm
attempt streams (stream=True) and aggregates chunks back into a complete
response, ticking the hook per chunk. The configured timeout then acts
per stream read (idle) instead of as a total budget. Providers that
reject streaming fall back to the plain non-streaming call; auth/payment/
rate-limit/transport errors propagate unchanged into the existing
recovery chains. Codex Responses (per SSE event) and Anthropic Messages
(per stream event, via the new create_anthropic_message on_stream_event
callback) tick the same hook from inside their wire adapters.
- agent/conversation_compression.py: CompressionCommitFence gains
touch_progress()/seconds_since_progress(); compress_context() installs
fence.touch_progress as the progress hook around the compress call.
- gateway/run.py: the hygiene wait loop treats hygiene_timeout_seconds as
an INACTIVITY budget — while the fence reports fresh progress the wait
extends, bounded by the new compression.hygiene_total_ceiling_seconds
(default 600s, clamped >= the idle budget) so a degenerate trickle
stream still dies. The timeout warning now says the summary model
produced no output, which is the only case that still triggers it.
- config/docs: hygiene_total_ceiling_seconds added to DEFAULT_CONFIG and
configuration.md; hygiene_timeout_seconds documented as inactivity-based.
Tests: tests/agent/test_aux_progress_streaming.py (hook plumbing, stream
aggregation incl. tool-call deltas and reasoning deltas, rejection
fallback, ceiling kill, fence progress surface); two new gateway tests
prove a slow-but-streaming worker survives past the fixed timeout
(sabotage-verified: fails with the old fixed deadline) and a
forever-trickling worker is still cut off at the ceiling.
Complements the task-granular background_work with the async-delegation
UNIT count (each dispatch/batch = 1), recovering the pool-slot semantics
active_count() gives. Together: background_work = real concurrent subagent
load (batch expanded), background_delegations = slot pressure to alert
against delegation.max_concurrent_children. Registered in metric_names,
documented, and covered by a task-vs-unit contract test.
The unquoted alternative in the directive pattern is `\S+`, so a ref built
by string interpolation truncates at the first space and strands the tail
as loose text next to a broken thumbnail. Composer images live in the app's
userData dir, which on macOS is `~/Library/Application Support/<App>/` — so
every pasted or dropped image hit this.
Adds format_reference_value next to REFERENCE_PATTERN, mirroring
formatRefValue in the desktop's directive-text.tsx, and covers the
round-trip through the parser.
A delegate_task fan-out batch occupies ONE async-pool slot by design, so
active_count() (unit/slot count) reports a 3-task batch as 1 — which
undercounts real concurrent subagent load on the background_work metric.
Add active_task_count() that expands a running batch to its child count
(N-task batch -> N, single -> 1) and switch the background_work reader to
it. active_count() is unchanged (capacity semantics preserved). Adds a
contract test for the unit-vs-task distinction and documents both counts.
active_agents counts foreground turns + cron + API runs but never the
backgrounded delegate_task subagents / terminal(background) processes /
kanban workers tracked only for scale-to-zero. Emit them as a distinct
content-free gauge so the fleet dashboard can show subagent/background
load per peer. Best-effort, sums async_delegation.active_count +
process_registry.count_running; 0 if a source is unavailable.
(cherry picked from commit 1b7ec684e253651a5aa5760b3bd527a3073ce207)
The cron health snapshot failure path logged only at DEBUG, so a cron
telemetry regression would silently drop all hermes.cron.* metrics while
gateway health metrics kept flowing. Promote to WARNING with the
exception *type* name only (no message text, preserving the no-raw-error
contract); keep the traceback on DEBUG.
(cherry picked from commit 1c9a3e737b7c57fa9886bc927f1623753a9a811b)
AST-driven pass over every Path.read_text()/write_text() without an
explicit encoding= across non-test code: 71 sites in 34 files
(skills_hub, hermes_cli/main+profiles+service_manager+container_boot,
mem0/hindsight/honcho plugins, achievements dashboard, release/CI
scripts, productivity+comfyui skill helpers, agent/*). Verified zero
positional-encoding collisions before insertion; per-file compile()
check after.
Adds a check-windows-footguns rule flagging bare single-line
read_text/write_text (multi-line forms stay covered by the AST guard
test from #38985). Together with the salvaged contributor commits this
retires the ~169-site bare file-I/O class (#37423's long tail).
Path.read_text() without an explicit encoding uses the platform's
default encoding. On Windows this is typically cp1252 or mbcs, which
causes UnicodeDecodeError or silent data corruption when reading
UTF-8 content (JSON files, user text, config with non-ASCII chars).
This is the read-side companion to the write_text() encoding fix.
Fixed the most critical locations that read JSON data, user content,
and config files across 14 files with 31 call sites.
Pattern: .read_text() → .read_text(encoding='utf-8')
json.loads(path.read_text()) → json.loads(path.read_text(encoding='utf-8'))
Several file-I/O call sites still use open() / Path.read_text() /
Path.write_text() without an explicit encoding, so they fall back to
the platform default. On Windows CN/JP/KR locales (GBK/CP932/CP949)
any non-ASCII byte in a config/state/user-content file raises
UnicodeDecodeError or UnicodeEncodeError and crashes the caller.
to the remaining hot paths:
- agent/copilot_acp_client.py: fs/read_text_file and fs/write_text_file
(Copilot's read_file / write_file tools,
directly reported in #18637 bug 2)
- agent/model_metadata.py: context-length YAML cache load + two
save sites (context probing is on the
call path of every model invocation)
- agent/nous_rate_guard.py: cross-session rate-limit JSON state
(read + atomic write via os.fdopen)
- cron/scheduler.py: user config.yaml read in run_job
- gateway/delivery.py: cron output writes for AI-generated
content, very likely non-ASCII
yaml.dump call sites also gain allow_unicode=True so the emitted
YAML preserves non-ASCII chars as-is instead of emitting \u escape
sequences.
Adds regression tests that monkeypatch builtins.open / Path.read_text
/ Path.write_text to simulate a GBK locale: each test raises
UnicodeDecodeError / UnicodeEncodeError unless the caller explicitly
passes encoding='utf-8'. Verified that the tests fail on main and
pass with this change, on Linux as well as on Windows.
Refs #18637
Widen the #70773 fix beyond the three in-request cleanup sites removed in
the cherry-picked commit: every remaining path that swaps out the shared
OpenAI client could still hard-close its pool from a thread that doesn't
own the in-flight sockets (credential rotation/refresh on the turn thread,
dead-connection cleanup, gateway cache eviction, transport recovery) —
the same FD-recycle corruption vector, just rarer.
Add AIAgent._retire_shared_openai_client(): shutdown(SHUT_RDWR) all pooled
sockets (FD-safe from any thread, unblocks in-flight readers) but never
call client.close() — FD release is deferred to GC, which cannot run until
every borrowing thread has unwound its SSL BIO. Refcounting is the
ownership handshake; with no borrowers the FDs are released immediately.
Wired into:
- _replace_primary_openai_client (rotation/refresh/dead-conn cleanup)
- try_recover_primary_transport (primary_recovery)
- release_clients (gateway cache_evict)
agent.close() keeps the hard close: full teardown is a real session
boundary where no request may be in flight.
Tests: new tests/run_agent/test_70773_shared_client_fd_corruption.py
covers the three watchdog/retry sites plus retire semantics; existing
close-assertions updated to pin retire-not-close.
The streaming stale watchdog was calling
_replace_primary_openai_client() from its polling thread, which closes
the shared client's connection pool. Worker threads from previous
stale-killed attempts may still be unwinding their SSL BIOs, causing
TLS application-data to overwrite SQLite file headers via FD reuse.
This is the same corruption vector documented in #67142 for Anthropic,
where the fix was to never close the shared client from a non-owner
thread. Apply the same pattern to the OpenAI-wire path:
- Stale stream watchdog: skip shared client replacement
- Mid-tool-retry cleanup: skip shared client replacement
- Stream retry cleanup: skip shared client replacement
The request-local client is already closed via _close_request_client_once.
The shared client is replaced lazily by _ensure_primary_openai_client
on the next request, which runs on the owning thread.
Closes#70773.
Follow-up to the cherry-picked #68258 base: the cross-session-stable
prefix (_cached_system_prompt_static) was only recorded on fresh
builds, so two paths silently degraded to the legacy single-breakpoint
layout (flagged in review of #68258/#69341/#69704):
- Session restore: gateway surfaces build a fresh AIAgent per turn and
restore the persisted prompt verbatim from the session DB; the static
prefix stayed None from turn 2 onward, flip-flopping the wire layout.
- Post-compression cached-prompt reuse: _invalidate_system_prompt()
clears the static prefix, and the keep-cached-prompt branch never
restored it.
Both sites now reconstruct the stable tier and adopt it ONLY when the
authoritative prompt string literally startswith() it — stable-tier
drift (skills edited, identity changed) falls back to the legacy layout
with the stored bytes untouched. Fail-open on any builder error. The
restore-path rebuild is gated on _use_prompt_caching so non-Anthropic
routes skip it entirely.
Refs #68191
Co-authored-by: JonthanaHanh <92574114+JonthanaHanh@users.noreply.github.com>
Co-authored-by: joaomarcos <joaomarcosdias444@gmail.com>
Co-authored-by: webtecnica <webtecnica@users.noreply.github.com>
The conversation loop normalizes message text right before the API call so
the request prefix is byte-identical across turns -- the stated reason is
KV cache reuse on local inference servers and better cache hit rates on
cloud providers. Cache breakpoints were injected *before* that pass, which
defeats it.
`_apply_cache_marker` rewrites a plain-string `content` into a
`[{"type": "text", ...}]` block. The normalization pass is guarded on
`isinstance(content, str)`, so every message that just got marked is
silently skipped by it and keeps its raw leading/trailing whitespace. A
message is only marked while it sits in the last-3 window, so:
turn N in the window -> marked, content "file1\nfile2\n"
turn N+1 rolled out -> plain, content "file1\nfile2"
The same logical message is sent with different bytes on consecutive
turns. The prefix stops matching at that position -- which is inside the
span the breakpoints were placed to protect -- so the reusable prefix
collapses back toward the system breakpoint on every turn. Tool results
carry a trailing newline almost by default (any shell command output), so
this is the common case, not an edge case.
Move the injection below every message mutation. Besides fixing the
whitespace divergence this stops breakpoints from being spent on messages
that the orphan sweep or the thinking-only drop is about to remove or
merge away -- a marker on a dropped message is a wasted breakpoint out of
the four available.
Nothing between the old and new call sites reads `cache_control`, and the
mutators now see the plain-string shapes they were written against.
The salvaged drift check compared durable rows to the in-memory snapshot
by content and ran in both modes. Two problems:
1. In-place compaction (the default) archives non-destructively — drift
cannot lose data there, and the strict-prefix content comparison
failed against seeded histories, aborting every in-place compaction
(5 test failures in test_in_place_compaction.py).
2. Content equality wedges on sessions with legal in-memory mutation of
past turns (multimodal compression, retry replacement) — the same
permanent-abort shape as #14694.
Now rotation-only and length-based: abort only when the durable parent
has MORE rows than the snapshot (a writer committed in the lease window).
Dead helper _durable_history_matches_snapshot removed.
When two consecutive compactions each failed to clear the threshold, the
anti-thrashing breaker blocked automatic compaction PERMANENTLY for the
life of the session: nothing decremented _ineffective_compression_count
(or _fallback_compression_streak) while blocked, so a session whose
middle region was briefly too small to compact never auto-compacted
again — it grew unbounded until the provider's hard context limit, and
only /new or /reset recovered it.
Recovery is a probation probe, not amnesty: after
_ANTI_THRASH_RECOVERY_SECONDS (300s) of continuous block the gate grants
exactly ONE attempt by dropping tripped counters to 1 strike (persisted,
so sibling agents on the same session row — gateway hygiene — unblock
too). An ineffective probe re-trips the guard on the next real-usage
verdict and the next recovery waits a full fresh window, so the worst
case in a truly incompressible session is one compaction attempt per
window — bounded, not thrash.
The recovery clock is armed lazily on the first BLOCKED evaluation and
is deliberately not durable: a restart that loads a durable tripped
counter (#69872) starts a full fresh window blocked, preserving the
restart-must-never-disarm contract (#54923).
Fixes#14694
Review follow-up for the dropped tool-call recovery (#69630): the
re-prompt pair was tagged _dropped_toolcall_nudge, but that marker was
not part of the ephemeral-scaffolding contract. _persist_session /
_flush_messages_to_session_db would therefore write the synthetic
'issue the actual tool call now' user message (and the narration-only
interim assistant turn) as real transcript rows — a resumed session
could replay the internal retry instruction as user-authored context
and prompt unsolicited tool use.
- Add _dropped_toolcall_nudge to _EPHEMERAL_SCAFFOLDING_FLAGS
(run_agent.py) so both SQLite and JSON persistence skip the pair.
- Add it to _SYNTHETIC_USER_FLAGS (conversation_compression.py) so the
compressor never treats the nudge as human intent.
- Flag the interim assistant half of the pair too, and include the
marker in the finalization scaffolding pop so a genuine turn end
strips the pair from the live transcript (mirrors the
_empty_recovery_synthetic pattern).
- Regression tests: flagged messages classify as ephemeral, the
returned transcript contains no scaffolding, and the turn tail stays
on the real assistant answer.
The initial fix guarded the no-tool-calls else branch, but that branch only
SETS final_response — the turn actually finalizes later, in a separate block
after final_msg is built. Runs that reached finalization via that path exited
without the guard ever running (observed live: a scheduled PR reviewer stalled
at tool_turns=1-2 with zero recovery nudges).
Move the recovery to the finalization chokepoint (right after final_msg is
built), so it catches every path that ends a turn. Single guard now:
- increments a consecutive-stall counter and re-prompts (bounded to 3),
- resets on any successful tool round, and
- resets on a genuine (non-mismatch) turn end,
so it guards each stall independently without capping the whole run and
without looping forever.
Verified live: the scheduled reviewer now recovers through the stalls and
submits real reviews — PR 57800 APPROVED, PR 54826 COMMENTED — one PR per run.
Some providers (observed: claude-opus-4.8 / claude-sonnet-4.5 on GitHub
Copilot, ~2026-07) return finish_reason="tool_calls" while the parsed
tool_calls array is empty — the model signalled it wanted to act but the
payload shipped no call. The conversation loop took the no-tool-calls
else branch, treated the turn's narration as the final answer, and exited
with the task unstarted.
On unattended multi-step jobs this is silent failure: a scheduled PR
reviewer, for example, would narrate "Let me verify the PR..." and stop
at tool_turns=0 every run, never submitting a review, while the job still
reported success.
Fix: in the no-tool-calls branch, detect the provider contract violation
(finish_reason == "tool_calls" with zero tool_calls) and re-prompt the
model to emit the call instead of exiting. Bounded to 3 consecutive
stalls; the budget resets after any successful tool round so it guards
each stall independently rather than capping the whole run. The narration
may live in content or only in the reasoning field (empty content) — the
guard keys on the finish_reason/tool_calls mismatch, so both are covered.
A genuine finish_reason="stop" text turn is unaffected.
Verified live: a scheduled reviewer that died at tool_turns=0 every run
now recovers through 20+ stalls per run and submits real reviews.
Tests: tests/run_agent/test_dropped_tool_call_recovery.py covers the
re-prompt, the empty-content case, the clean-stop control, and the
bounded-loop guard.
The last_activity field showed 'tool completed: read_file (120.7s)'
identically whether the tool succeeded or failed, making post-mortem
analysis of silent-hang reports (#69131) unnecessarily hard.
Append ' (error)' to the activity description when the tool result is
classified as a failure, in both the concurrent and sequential
execution paths.
(Salvaged from PR #69467; its conversation_loop.py activity-touch hunk
is the same fix as PR #69577 and lands in the preceding commit.)
After a tool call completes, the conversation loop does to
start the next API call. Between the last (from tool
completion in tool_executor.py) and the next one (at the start of the
next API call), there is a gap. If anything during this gap takes time
— context compression, slow provider prefill, or other post-tool
processing — the combined inactivity window can exceed the gateway's
inactivity_timeout (default 120s). The gateway kills the session and the
user sees 'agent never returns a final response', even though the tool
call itself succeeded in 0.1-0.2s.
Fix: call right before so the gateway
sees a fresh timestamp immediately after tool results are posted,
regardless of how long the follow-up API call takes.
Fixes#69559
Three durable ledgers used `with _connect() as conn:` where the sqlite3
connection context manager commits/rolls back but never closes, leaking the
db/-wal/-shm file descriptors on every call. On a long-running gateway this
exhausts RLIMIT_NOFILE and fails unrelated components with
`[Errno 24] Too many open files`. Same bug class as the cron execution ledger
(#69567 / PR #69594), which the connection helpers here are modeled on.
Fix: route every ledger operation through a `_transaction()` context manager
that guarantees `conn.close()` on exit. `_connect()` keeps its
schema-on-connect contract (several tests call it directly) and now self-closes
if schema init fails.
Adds per-module regression tests asserting every opened connection is closed,
including the no-op-update and exception-mid-transaction paths.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Follow-up to the cherry-picked #68633 commits, closing the final open
review point (egilewski): _relocated_replay_cache_control was applied
only inside `if replayed:`. When anthropic_content_blocks contained
only a blank cache-marked text block, `replayed` came out empty, the
function fell through to the main path's placeholder, and the cache
marker was lost; signed thinking + a blank marked text block likewise
returned with no cacheable carrier for the relocated marker.
The replay branch now appends the non-whitespace "(empty)" placeholder
when no cacheable (text/tool_use) block survives the blank filter and a
blank text block was dropped (or a marker needs a carrier) — so replay
stays schema-valid on Bedrock/strict endpoints and the breakpoint
survives on the placeholder.
Also reconciles the block-level tests from #69517 with the new
drop-then-fallback contract (blank blocks are dropped at the block
level; the message-level result is still always non-blank).
Refs #69512
Co-authored-by: ygd58 <buraysandro9@gmail.com>
Follow-up per independent review of #68633 (GPT-5.6-sol-xhigh in Codex,
reviewer egilewski) on this PR.
Two real bugs in the blank-text-block filtering added by that fix:
1. `effective = blocks or content` fell back to the RAW, unfiltered
`content` variable whenever every block was filtered out as blank --
which happens precisely when the entire message content WAS the
blank/whitespace payload the filter exists to remove (a sole blank
text block, a sole cache-marked blank block, or standalone
whitespace scalar content with no tool_calls). The fallback silently
restored the exact invalid content the filtering just stripped,
leaving the message provider-invalid.
Fixed: `effective = blocks if blocks else [{"type": "text", "text":
"(empty)"}]` -- never falls back to raw `content`. Also moved the
cache_control application (both the relocated-from-a-dropped-block
marker and the message-level marker) to run against `effective`
instead of the pre-fallback `blocks`, so a cache marker on a block
that was the ONLY content still lands on the (empty) placeholder
rather than being silently lost when `blocks` was empty at the
point it would otherwise have been applied.
2. The normal-path blank-text check used `(blk.get("text") or "").strip()`,
which is not type-safe for a truthy NON-string, non-None text value
(e.g. an int or dict from an invalid upstream payload) -- `or`
doesn't substitute for a truthy value, so `(7 or "").strip()` still
raises AttributeError. Now checks `isinstance(text, str)` first,
matching the replay path's `_sanitize_replay_block()`, which the
reviewer confirmed was already correctly type-safe.
Added regression tests for: sole blank list block, sole whitespace
scalar content, sole cache-marked blank block (marker relocation to
the placeholder), a truthy non-string (int) text value both mixed with
a surviving tool_use and as the sole content, and a dict-valued text
field. 7/7 new tests pass; 193/193 in the full
tests/agent/test_anthropic_adapter.py file; 23/23 in
tests/agent/test_prompt_caching.py (unaffected, confirmed).
Ports #63228 forward onto current main per teknium1's review.
Bedrock and strict Anthropic-compatible endpoints reject text blocks
where text is empty or whitespace-only with HTTP 400. The normal
list-content path extended blocks without filtering, and the
ordered-replay fast path (_sanitize_replay_block) returned blank text
blocks unfiltered.
Per review, fixes three gaps in the original port:
1. Type safety: the normal-path filter used blk.get('text', '').strip(),
which crashes with AttributeError when text is explicitly None (not
absent) -- .get()'s default only applies when the key is missing.
_convert_content_part_to_anthropic() can preserve None from an
invalid upstream input text block. Now uses
(blk.get('text') or '').strip() on both paths.
2. Cache marker loss: prompt_caching.py's _apply_cache_marker() sets
cache_control directly on content[-1] for list content. If that last
part happens to be blank text, dropping it without relocating
cache_control silently loses the breakpoint. Both the normal and
replay paths now capture a dropped block's cache_control and reapply
it to the new last surviving cacheable block via the existing
_apply_assistant_cache_control_to_last_cacheable_block() helper
(setdefault semantics, so it never clobbers a legitimately-placed
marker).
3. Scalar whitespace: the non-list content branch
(blocks.append({'type': 'text', 'text': str(content)})) accepted a
truthy whitespace-only string unfiltered. Now filtered the same way
as list-content blocks.
8/8 new tests pass in TestBlankTextBlockFiltering (including None-safety,
scalar-whitespace, and cache_control-relocation regressions on both
paths); 186/186 in the full tests/agent/test_anthropic_adapter.py file.
OpenAI documents GPT-5.5 / GPT-5.5 Pro as extended-cache-only: in-memory
prompt cache retention is not available for them, and only
prompt_cache_retention: "24h" is supported. Responses requests that omit
the field see near-zero cached_tokens even with a stable prompt_cache_key
and identical prefixes (observed on an OpenAI-compatible Responses relay:
0 cached across repeated identical calls before; 97% cache reads after).
Send the field for the gpt-5.5 model family (bare and namespaced ids like
openai.gpt-5.5) on OpenAI-compatible Responses routes, mirrored in the
auxiliary Codex adapter, and pass it through preflight normalization.
Skipped for xAI, GitHub/Copilot, and the chatgpt.com Codex backend, which
reject or ignore body-level cache fields.
A brew Python upgrade (original report) or an interrupted venv rebuild
(v0.19.0 report in the same thread) can leave certifi importable while
its bundled cacert.pem is missing or a dangling symlink. Every TLS
connection then fails — Feishu/Telegram/WeChat/DingTalk all down —
with an opaque 'Could not find a suitable TLS CA certificate bundle'
from deep inside httpx/requests.
The existing repair infrastructure only probed
`hasattr(certifi, 'contents')`, which PASSES in exactly this failure
state, so neither the early venv self-heal nor `hermes update`'s
import-probe repair ever classified certifi as broken. Extended, not
replaced:
- hermes_cli/_early_recovery.py: the in-process probe now also
validates that certifi.where() exists and is a plausible bundle
(>=1KiB), so the pre-import self-heal repairs it like any other
wiped core package.
- hermes_cli/main.py (_detect_broken_lazy_refresh_imports): the
subprocess probe script used by `hermes update`'s venv repair
applies the same bundle-file check inside the target venv.
- hermes_cli/doctor.py: `hermes doctor` already failed the cert
check; `hermes doctor --fix` now repairs it (pip force-reinstall
certifi + module-cache invalidation + re-verify), covering
brew/manual venvs where no update marker exists. Failures funnel
into the manual-action list with the exact command.
- agent/ssl_guard.py: the startup SSLConfigurationError hint now leads
with `hermes doctor --fix` instead of only the raw pip command.
Fixes#29866
When api_key_hint from a 401 response doesn't match any pool entry
(common with OAuth tokens where runtime_api_key rotates), the pool
rotated without marking anything exhausted and handed back a fresh
selection. Because nothing was ever marked, the pool could never reach
the "no available entries" state — the caller retried the same dead
token forever (~6 attempts/sec), starving the event loop so /stop was
never processed; only killing the gateway ended it.
Rebased onto the identity-tracking rework that landed on main
(73c4b5a045): the single-entry escape from that commit already stops
the most common OAuth case, so this fix bounds the REMAINING gap —
multi-entry pools ping-ponging A->B->A with an unmatched hint. Cap
consecutive no-mark rotations at one full lap of the available
entries, then return None so the error surfaces / fallback activates.
Deliberately does NOT mark innocent entries exhausted (the original
PR's approach): that would quarantine a healthy key for the full
cooldown TTL on a hint that provably matches nothing. No cooldown is
written by the escape, so healthy keys stay available next turn --
bounded without hammering.
The streak resets when a rotation identifies a real entry and on any
successful normal select(), so only genuinely consecutive unmatched
rotations trip the bound.
Fixes#70401
Hardening follow-up to the #69619 review fix. The previous regression
byte-pinned only the rescued pre-#69619 generation; older frozen entries
were covered solely by fragment assertions and a self-matching loop that
cannot detect a frozen entry mutating (the loop tests each entry against
itself).
- Pin all four _HISTORICAL_SUMMARY_PREFIXES generations as literals in
_FROZEN_PREFIX_GENERATIONS and assert order-sensitive tuple equality
plus detect/strip for each
- State the prepend-only contract explicitly on the tuple: never mutate
or reorder existing entries
Negative controls verified: mutating, dropping, or reordering a frozen
entry each fail the new test, while the legacy self-matching loop still
passes under mutation — confirming the closed coverage gap.
Address review on #69619: the previous commit mutated the newest frozen
entry in _HISTORICAL_SUMMARY_PREFIXES and never froze the live prefix it
retired (the generation with both the four-heading discard clause and
the tools-active clause). A summary persisted immediately before
upgrading was therefore treated as an ordinary message on
resume/re-compaction, keeping the old handoff text embedded in the body.
- Prepend the exact pre-change live prefix as a new frozen entry
(newest-first), leaving all existing frozen entries byte-identical
- Restore the Jul 2026 (#65848 class) frozen entry to its original
four-heading text
- Pin the retired generation as a literal in
test_summary_prefix_semantics.py so mutating or dropping it fails CI
- Make the #65848 tool-use regression position-agnostic (match the
pre-clause generation by content, not tuple index)
Verified byte-identity of both rescued generations against the parent
commit. 233 focused prefix/resume/compressor tests pass.
Remove three directive-heavy section headers from both the LLM
and deterministic summary templates that caused the agent to
resume stale tasks after context compression:
- Historical In-Progress State
- Historical Pending User Asks
- Historical Remaining Work
These sections read as actionable instructions even within a
REFERENCE-ONLY wrapper, hijacking the user's latest message.
The remaining sections are purely descriptive/past-tense.
Frozen prefix copies in _HISTORICAL_SUMMARY_PREFIXES updated
to match. Test 8/8 passed.
Anthropic released Claude Opus 5 (+ -fast variant) — both are live on
OpenRouter and the Nous Portal /models endpoint (verified against both
live APIs). Opus 4.8 entries are kept.
- hermes_cli/models.py: opus-5 + opus-5-fast in OPENROUTER_MODELS;
opus-5 in _PROVIDER_MODELS[nous] (Portal serves both, curated list
carries the base model like the rest of the Nous Anthropic block).
Ordering: below fable-5 flagship, above opus-4.8.
- agent/model_metadata.py: claude-opus-5 -> 1M context (matches live
OpenRouter metadata).
- agent/reasoning_timeouts.py: claude-opus-5 -> 240s stale-timeout
floor (same as the opus-4.x thinking family).
- website/static/api/model-catalog.json: regenerated via
scripts/build_model_catalog.py.
Both providers bill via official_models_api (live pricing), so no
_OFFICIAL_DOCS_PRICING snapshot entry is needed for these routes.
- Cut the leftover telemetry.* DEFAULT_CONFIG block (nothing reads it) and
the legacy telemetry-key fallback in policy.py.
- install_id: persist the minted UUID back to config.yaml on first use so
service.instance.id survives gateway restarts (fail-open when the write
is not possible); regression test covers the restart path.
- Remove the no-op gateway_health_export.redaction config keys. Redaction
is always-on by design and deliberately not configurable; status output
now says so.
- Collapse redaction to one unconditional secrets+PII scrub: drop the
none/pii content modes (they served the dropped trajectories plane) and
fold gateway_health.py's duplicate bearer/token/email/phone regex layer
into agent/monitoring/redaction.py.
Salvages the event-spine foundation from feat/telemetry-observability
(emitter, typed events, OTLP streaming, redaction — authorship preserved in
the preceding commits) and scopes it to the plane enterprise operators need
today: gateway Service Health Monitoring plus redacted Operational
Diagnostics, exported over OTLP.
Dropped from the salvaged branch, deliberately:
- run/model/tool trajectory capture (plugins/telemetry hooks, tel_spans)
- the local JSONL + state.db tel_* store (monitoring is egress, not storage)
- usage rollups/metrics, /insights integration, bulk export
- hermes telemetry CLI (replaced by hermes monitoring status)
Those planes — shared client usage metrics and enterprise trace telemetry —
are being designed on the NeMo Relay integration with distinct consent,
policy, and export boundaries; this keeps the monitoring plane content-free
and independently enableable.
Renames agent/telemetry -> agent/monitoring, config telemetry.* ->
monitoring.*, and pins the otlp extra at OpenTelemetry 1.39.1 (matching
uv.lock; 1.30.0 conflicts with mistralai>=2.4 on opentelemetry-api).
Self-review after the #51714 feedback found the reviewer's dead-table finding
was not isolated — the schema advertised far more than the code populates, and
our own tests hid it by hand-feeding fields production never sends. Make the
surface honest by subtraction.
Schema (10 tel_* tables -> 5):
- Delete tel_gateway_events, tel_cron_events, tel_skill_events,
tel_memory_events, tel_feedback_events — declared, never written, never read.
- Drop columns nothing populates: tel_runs.{profile_id,estimated_cost_usd,
cost_status}; tel_model_calls.{ttft_ms,estimated_cost_usd,cost_status,
cost_source,end_reason,retry_count}; tel_tool_calls.{backend,retry_count,
approval}; tel_spans.attrs_json. Cost duplicated the existing sessions
billing columns and was always NULL here.
- events.py / emitter _TABLE_COLUMNS / OTLP _span_attrs / rollup / preview
display all trimmed to match.
Correctness:
- end_reason no longer hardcodes "completed". Production finalize callers pass
`reason` (shutdown/session_expired/session_reset); _coarse_end_reason now
reads it and maps accordingly.
- Fix a latent bug the trim exposed: the model_call hook passed end_reason= to
ModelCallEvent, which the @_safe wrapper was silently swallowing — so
tel_model_calls dropped every row in real runs. Now writes correctly.
Tests:
- Stop hand-feeding estimated_cost_usd / turn_exit_reason that no production
call site sends. Finalize is now driven with the real `reason` kwarg, and
assertions cover only fields that are actually populated. This is what let
the model_call drop hide — the suite graded on a fictional contract.
Net: a smaller system that does what it says. Verified end-to-end over the real
dispatch path (runs + connected span tree + model/tool rows populate; dead
tables gone). 160 telemetry/state/insights tests green.
Addresses the review on #51714: the trace/span layer was declared but unwired —
tel_spans was never written, call rows had no timestamp, and nothing set parent
lineage, so the store was metrics-only and couldn't reconstruct a trace.
Wire the span layer (keeping the praised star-schema shape):
- New SpanEvent (span_id/trace_id/run_id/parent_span_id/name/kind/start_ns/end_ns)
mapped into tel_spans via the emitter's _TABLE_COLUMNS.
- The plugin mints a root span per run and, on each model/tool call, emits a
SpanEvent (timing + parent = the run's root) keyed by the SAME span_id as the
detail row, so tel_model_calls / tel_tool_calls JOIN to their span.
- Call hooks fire on completion, so end_ns = now and start_ns is reconstructed
from the measured latency/duration. The run's root span is emitted at finalize
with the true run start/end.
Result: tel_spans is a connected, single-trace_id, run -> calls tree a desktop
waterfall (or any reader) can render directly, ordered by start_ns. Existing
metrics rows (tel_runs/model_calls/tool_calls) are unchanged.
OTLP: spans now flow to the exporter with their trace/parent/timing attributes.
The exporter still emits one OTel span per event rather than reconstructing OTel
SpanContexts into a connected trace tree; that projection is left for a follow-up
and the module docstring now says so plainly instead of over-claiming.
Adds test_spans_trace.py (connected-tree + detail-row JOIN) over the real dispatch
path. Accurate (pre-hook) start times, real OTLP SpanContexts, and subagent
cross-run lineage remain follow-ups.
Aggregate metrics are derived from the local tel_* tables — they're a coarsened
view of local data, not an independent capture path. With telemetry.local=false
nothing is written, so an aggregate opt-in had nothing to aggregate, yet
may_upload_aggregate() returned True and `status` showed "Aggregate metrics: on".
The config could claim a state it couldn't fulfill.
Gate aggregate on local being enabled:
- may_upload_aggregate() now requires local_enabled AND allow_aggregate AND
consent_state == aggregate.
- `telemetry status` computes aggregate_enabled the same way and, when consent is
aggregate but local is off, prints "inert: local telemetry is off — nothing to
aggregate" instead of the opt-in hint.
Happy path is unchanged (local on + consent aggregate -> on). Adds policy and CLI
tests for the inert combo.
Rename the telemetry tiers away from the borrowed control-plane/data-plane
jargon to plain language, across code, CLI output, config, and docs:
- "local plane" -> "local telemetry"
- "aggregate plane" -> "aggregate metrics"
- "trajectories plane" -> "trajectories" / "telemetry.trajectories"
- "three planes with a hard wall" -> "three settings, isolated from each other"
User-facing `hermes telemetry status` now reads "Local telemetry: on" /
"Aggregate metrics: off" / "Content export: off (trajectories disabled)".
The OTLP resource attribute key telemetry.plane is renamed to telemetry.scope
(wire-level identifier; nothing consumes it yet).
No behavior change — wording only. Status renders identically apart from the
labels; tests updated to match the new strings.
policy.resolve() / TelemetryDecision was a read-only projection used only by
`hermes telemetry status` for display. The actual behavior gates already read
telemetry.* straight from config: the emitter (whether to write) and the plugin
loader (whether to auto-load) each call .get("local", True) on the loaded config,
never through policy.
Make config the single chokepoint the status command reads too: it now resolves
local/allow_aggregate/consent_state inline from the loaded config, the same way
the other gates do. policy.py keeps only what config can't express on its own —
the consent constants, ensure_install_id(), and may_upload_aggregate(config) as a
pure function (the gate a future uploader must consult). resolve() and the
TelemetryDecision dataclass are removed; policy.py drops 107 -> 70 lines.
No behavior change: status renders identically, and the default-on local plane is
still defaulted in DEFAULT_CONFIG plus a fail-safe .get(..., True) at each gate.
Add a built-in telemetry system that records what the agent does — workflows,
model calls, tool calls, errors — to the local machine, powers `/insights`, and
can export to an operator-chosen destination. Default-on locally; nothing leaves
the machine unless the user exports it or opts into the aggregate plane.
Three planes with a hard wall between them:
- local: full-fidelity observability (real model/provider/tool names), on by
default, never leaves the machine.
- aggregate: opt-in metadata, default off. No uploader ships — consent is
recorded via telemetry.consent_state, and `preview` shows what would be
produced, computed locally.
- trajectories: full message content, opt-in, exported only to the operator's
own destination.
Mechanism:
- Bundled `telemetry` plugin registers observational lifecycle hooks
(on_session_start / post_api_request / post_tool_call / on_session_finalize).
No core call sites are edited; hooks already carry the data.
- Fire-and-forget emitter: emit() returns in microseconds, never blocks or
raises into a model/tool call. A daemon thread writes events to an
append-only JSONL log and the tel_* tables in state.db (its own sqlite
connection, separate from SessionDB).
- tel_runs / tel_model_calls / tel_tool_calls live in the declarative
SCHEMA_SQL and are reconciled automatically; SCHEMA_VERSION 16 -> 17.
- metrics derives rollups for /usage and /insights; rollup builds per-run
summaries for `hermes telemetry preview`.
Consent is config, not a parallel command surface. The config file is the root
of trust: set telemetry.consent_state with `hermes config set`, or pin any
telemetry.* key (including allow_aggregate) via managed scope, which overrides
the user's value per key. `hermes telemetry` exposes only what config cannot:
status (report), preview (query), and export.
Export:
- exporter_bulk writes telemetry (and, when the trajectories plane is enabled,
session content) to ndjson/json.
- otlp_exporter streams spans to a configured OpenTelemetry Collector over
OTLP/HTTP. The SDK is an optional extra (hermes-agent[otlp]), lazily
installed via tools.lazy_deps on first use.
- Secrets are always redacted on every export path
(redact_sensitive_text(force=True)); content export is gated by the
trajectories plane, and PII scrubbing follows telemetry.content_redaction.
OTLP auth headers reference environment variable names, never inline values.
No outbound emission to Nous. The aggregate uploader is intentionally not built.
AST-driven pass over every subprocess.run/Popen/check_output/check_call/call
with text=True (or universal_newlines=True) and no explicit encoding=:
append encoding='utf-8', errors='replace' at the kwarg site. 136 call
sites across 28 files (cli.py, hermes_cli/main.py, tools_config.py,
environments, computer_use, gateway, scripts, skills helpers, agent/*).
Together with the salvaged #55339/#60741 commits this closes out issue
#53428's bug class; the salvaged #60751 linter rule in
check-windows-footguns.py now enforces it repo-wide (verified: 807 files
scanned, zero findings).
On Windows with Chinese locale (GBK), subprocess.run(text=True) without
explicit encoding causes UnicodeDecodeError crashes. This fix adds
encoding='utf-8', errors='replace' to all subprocess.run() and
subprocess.Popen() calls that use text=True across 76 non-test Python files.
Fixes#53428 (master tracker for Windows GBK locale crash).
Note: credential_pool.py and electron changes excluded per reviewer request —
those will be submitted as separate focused PRs.
Base-url+model dedup was meant for custom shim aliases, but it also
skipped first-class providers that share an inference host while using
different credentials. That stranded xai-oauth spending-limit failover
to the xai API-key provider when both used the same model slug.
Replace 3 duplicated entry_id resolution blocks (try/except +
entry_id_for_api_key + fallback to None) in agent_init.py,
chat_completion_helpers.py, and switch_model with a single
sync_credential_pool_entry_id(agent) function in agent_runtime_helpers.
Follow-up to #70323.
Track the selected credential by stable pool entry ID so token refreshes and shared cursor movement cannot detach failures from the entry that issued them. Stop unmatched single-entry pools from reporting a no-op rotation as successful recovery.
Co-authored-by: Maxim Esipov <maksesipov@gmail.com>
Registers `ar` in the supported-language set and alias table and ships
locales/ar.yaml at full key and placeholder parity with en.yaml, covering
approval prompts and gateway slash-command replies. Identifiers, commands,
paths, config keys, model/provider names, and {placeholder} tokens are kept
verbatim.
Co-authored-by: Da7-Tech <286182457+Da7-Tech@users.noreply.github.com>
profile_name was only written on the agent's initial lazy create
(e8b7ce8c1); every parented child row — compression rotation, TUI
/branch, desktop branch first-persist — was created without it. A
non-default profile's lineage therefore turned NULL on its first
compression or branch and aggregated as "default" in unified session
lists, completing the cross-profile session-jump.
Fix the class at the DB layer: _insert_session_row's parent backfill now
COALESCEs profile_name from the parent alongside cwd/git_* (#64709
pattern), so any parented child inherits its lineage's owning profile.
Stamp it explicitly at the three create sites as well — compression
rotation (mirroring _ensure_db_session), TUI session.branch, and the
TUI first-prompt row persist — so rows are self-describing even when the
parent row predates the profile_name column.
Flips the default fan-out cadence from per_iteration (advisors re-run on
every tool iteration, multiplying advisor spend by tool-loop depth) to
user_turn (advisors run once on the first message of each user turn; the
acting aggregator works the rest of the tool loop with that turn's
advice). Until per-mode benchmarks justify a costlier default, MoA
defaults to the cheapest, lowest-impact cadence (#67199).
One default for everyone — no split legacy/new-preset semantics; presets
that want per-step advising set fanout: per_iteration explicitly. All
three modes (user_turn / per_iteration / every_n:N) remain selectable;
every_n:1 still collapses to per_iteration (semantic identity), while
unparseable values now fall to user_turn (the default).
Docs updated with a default-change note; the per-iteration rerun test
pins its mode explicitly.
Co-authored-by: skyer-flyyy <188930297+skyer-flyyy@users.noreply.github.com>
The skill-authoring guide and curator prompt both reference
descriptions as the primary discovery mechanism but never mentioned
the 57-char system prompt truncation. Add explicit guidance:
- Authoring guide: frontmatter docs, template comment, size limits,
pitfall #3 with good/bad examples, verification checklist
- Curator prompt: parenthetical noting the 57-char window when
writing umbrella skill descriptions
When a skill is created or edited with a description longer than
SKILL_PROMPT_DESC_LIMIT (60 chars), the tool response now includes a
system_prompt_preview field showing exactly what the system prompt
skill index will display. This gives the agent immediate feedback to
self-correct truncated trigger phrases.
Also adds tool schema guidance about the 57-char window and fixes a
stale docstring in skill_commands.py that incorrectly claimed the
system prompt renders the full description.
The system prompt skill index truncates long descriptions to 57 chars,
but this limit was a hardcoded magic number. Extract it as a named
constant and factor the normalization logic into a shared private
helper so the extraction function and the new truncation predicate
cannot drift.
No behaviour change — pure refactor.
Verification follow-up for the #51226 salvage: the host call site guarded
select_context with hasattr(), but the ABC defines a default on every
engine, so the built-in ContextCompressor (and any non-implementing
engine) still paid per-request shallow copies of the conversation
history plus a hook call on every provider request. Identity-check the
bound method against ContextEngine.select_context and return the
request untouched — mirroring the existing base-method short-circuit in
_notify_context_engine_turn_complete — so the default path does zero
work, not just produces an identical result.
Adds two pins: the base no-op is never invoked (patched-to-raise base
stays silent), and ContextCompressor.__dict__ contains neither new verb.
Also registers the contributor email mapping for @chaos-xxl.
Addresses the hermes-sweeper review on #51226:
- _apply_context_engine_selection now passes shallow copies of the read-only
conversation_messages / incoming_message to the hook, so an engine mutating
them in place cannot corrupt persisted transcript state (enforces the
request-only contract, not just documents it). Adds a mutation-regression
test asserting persisted history + incoming message are untouched.
- on_turn_complete docstring: scope the coverage claim to the standard
finalization seam. Some abnormal early-return paths (content-policy block,
provider terminal failure) currently persist+return without finalization and
don't emit the hook; documented as best-effort with a shared-seam follow-up,
rather than over-promising a guaranteed callback for every early exit.
- _apply_context_engine_selection: reject an empty list. all([]) is True, so
a [] returned by a failing/buggy engine previously replaced a valid request
with an empty message list the downstream sanitizers can't restore; now it
falls open to the unmodified request (honors the fail-open contract).
Thanks @johnnykor82 for catching this on #41918's review.
- test: empty list keeps the original request (fail-open regression).
- docs: document select_context()/on_turn_complete() in the public
context-engine plugin guide (were still describing only the old contract).
- context_engine.py: document that select_context() runs before cache-control
and all request sanitizers, so (a) replacements still pass host validation
and (b) the no-op default keeps the request byte-stable (AGENTS.md prompt-
cache invariant). Note the hook is evaluated per provider request.
- tests: no-op path is byte-stable for cache-control; a role-unusual
replacement is passed through for the existing downstream sanitizers to
normalize (select_context does structural validation only).
The on_turn_complete() observation hook is the engine's post-turn signal,
so it should receive the completed turn's canonical token usage when the
host has it, not a hardcoded None. Per @johnnykor82's #41918 contract: the
engine uses prompt/completion + cache_read/write/reasoning buckets to judge
how large/expensive the selected context was before the next select_context().
- conversation_loop.py: stash the most recent provider response's usage_dict
(the same canonical shape fed to update_from_response) on the agent as
_last_turn_usage; reset to None at turn start so turns that never reach a
provider response (early failure / interrupt) forward None, not a stale
prior turn's usage.
- turn_finalizer.py: forward agent._last_turn_usage instead of usage=None.
- context_engine.py: document the usage param contract on the ABC hook.
- tests: cover both ends through the real finalize_turn path — completed turn
forwards the full canonical bucket set intact; no-response turn forwards None.
Co-authored-by: johnnykor82 <johnnykor82@users.noreply.github.com>
Adds the post-turn observation verb as the companion to select_context():
an optional, no-op-default on_turn_complete() called once after the
assistant/tool loop finishes, with the finalized transcript snapshot. Lets
an engine ingest/index/summarize the completed turn to inform the next
select_context(). Wired via _notify_context_engine_turn_complete() from
turn_finalizer.finalize_turn(); fail-open, base no-op short-circuited so
non-implementing engines (incl. the built-in compressor) pay nothing.
This is the request-assembly + observation pair from #41918; with this
commit the PR fully subsumes #41918's two hooks (prepare_request_messages
-> select_context, on_turn_complete) rather than only the selection half.
Co-authored-by: johnnykor82 <johnnykor82@users.noreply.github.com>
Adds an optional, no-op-default select_context() hook to the ContextEngine
ABC, called every turn after the request messages are assembled and before
provider dispatch — independent of should_compress(). Lets an engine select
or replace which context enters the prompt for a single request (retrieval,
topic routing, role/branch switching) without mutating persisted history,
removing the need to abuse should_compress()=True as a per-turn callback.
The host call site (_apply_context_engine_selection) is fail-open: a missing
hook, an exception, or an invalid return value leaves the assembled request
untouched. Additive and non-breaking: the built-in compressor and every
existing engine are unaffected.
Consolidates the per-turn request-assembly surface proposed across #41918,
Related: #36765#41918#24949#47109#50053#23837#25115#29370
Implements the agreed design (2026-07-23): org skills are FIRST-CLASS (bare
names) with three hard companions.
1. TOKEN-GATED RESOLUTION: _org/<org_id>/ mirrors resolve ONLY while marked
active. pull_org_skills (which runs only after the token's org_id+org_role
verified) writes _org/.active_org; discovery (iter_skill_index_files,
_find_skill_dir, snapshot manifest) prunes every other mirror. Leave the
org (verified personal token in maybe_pull_org_skills) => marker cleared
=> org skills stop resolving; offline => marker untouched (grace).
Snapshot manifest includes the marker so org switches invalidate the
prompt snapshot; _SKILLS_SNAPSHOT_VERSION bumped to 2.
2. FAIL-LOUD COLLISIONS: listing pass unified across snapshot/scan paths;
a personal/org name clash flags BOTH entries '[name collision — load via
category path]' — neither side silently wins (personal-wins = silent
divergence from the org set; org-wins = shadowed personal work).
skill_view's existing multi-candidate refusal already rejects the
ambiguous bare name.
3. PROVENANCE: org entries list under an org:<org_id> category with
'[org-shared: by <author>]' tags; skill_view prepends a load-time header
(org, author, as-of + read-only/fork-and-propose guidance) INTO the
content the model consumes, plus an org_provenance result field. Author
comes from the pull-time .org-provenance.json sidecar (HEAD commit author
— token-verified at push by the plane's author_mismatch guard, gg #166).
4. READ-ONLY MIRROR: skill_manage patch/edit/delete/write_file refuse org-
mirror targets with fork-and-propose guidance; org skills are curation-
exempt (is_curation_eligible False — the org HEAD owns them).
Tests: 11 new (tests/agent/test_org_skill_namespace.py) covering gating,
stale-mirror pruning, org-switch flip, snapshot provenance, listing labels,
both-sides collision flags, read-only guard, curation exemption; 448 green
across skills/prompt/sync suites. Live E2E (real modules, temp HERMES_HOME,
mock plane): merge -> pull -> marker+sidecar -> labeled listing -> exactly-2
collision flags -> load-time header -> edit refused -> marker cleared =>
org skills vanish, personal survive.
_render_tool_calls only handled dict-shaped entries; a SimpleNamespace-
shaped tool_call (SDK-style stream-stitched responses) rendered as
'[called tool: tool]', silently losing the function name and arguments
from the advisory view. Handle both shapes (including a namespace-shaped
nested function inside a dict entry).
One-hunk hardening salvaged from closed#59712.
Co-authored-by: SquabbyZ <601709253@qq.com>
Reference models may have a smaller context window than the aggregator
(e.g. kimi-k2.7-code @ 262K advising a glm-5.2 @ 1M conversation).
Without context-length protection, a reference whose window is exceeded
gets a hard HTTP 400 from the provider, which _run_reference's
try/except silently converts to a [failed: …] note — the MoA turn
silently degrades to fewer references (#60345).
Redesigned implementation of #60387:
- Estimate AFTER the advisory system prompt is prepended, so the
request that is actually sent is what gets budgeted.
- Reserve output headroom: the preset's reference_max_tokens when set,
else an 8192-token constant, plus a 10% estimator-error fraction.
- Trim on advisory-view boundaries (text-only user/assistant turns; no
tool-result frames to orphan), preserving the system prompt, the
user-first invariant after every pop (never assistant-first), and the
trailing synthetic user turn.
- Cache get_model_context_length per (provider, model) in a per-fan-out
dict shared across the worker threads, so a turn resolves each
window once instead of probing metadata sources
per-reference-per-iteration (failures are cached too).
Co-authored-by: webtecnica <75556242+webtecnica@users.noreply.github.com>
Follow-ups for salvaged #56344:
- A reference that completes between the interrupt check and the reap
keeps its REAL output and accounting (the provider call billed) instead
of being zeroed with a placeholder.
- A reference still in flight at interrupt time gets a placeholder in the
results, but its future now carries a done-callback that folds the
eventual real usage/cost into the facade's pending accounting
(late_accounting_sink -> _record_late_reference_accounting), so billed
spend is never silently dropped. Pending totals are folded (not
overwritten) and guarded by a lock since done-callbacks fire on
executor worker threads.
- Interrupted placeholder results are no longer written into the facade's
turn-scoped reference cache: a cache HIT never re-runs references, so
caching a partial snapshot would replay '[skipped: interrupted by
user]' notes for the rest of the turn. The cache is left empty and the
next create() re-runs the fan-out.
agent/tool_executor.py's concurrent tool batch checks agent._interrupt_requested
and aborts the wait early; agent/moa_loop.py's _run_references_parallel had
no equivalent, so a MoA-enabled turn blocked on ThreadPoolExecutor.result()
until every reference model finished or hit its own individual
auxiliary.moa_reference timeout -- there was no way for the user to abort a
live turn mid-fanout.
Thread an optional `agent` parameter through aggregate_moa_context ->
_run_references_parallel (used when MoA references run alongside the main
model) and MoAClient/MoAChatCompletions (used when the MoA preset itself is
the acting model), then poll concurrent.futures.wait() in
_REFERENCE_POLL_INTERVAL_S slices instead of blocking on future.result() per
reference, checking agent._interrupt_requested each cycle.
Deliberately scoped to interrupt/cancel only -- no new or changed timeout
value, so this doesn't overlap open PRs #53784/#53875 (which lower the
per-reference timeout default but don't add interrupt support). `agent` is
optional and defaults to None, so any caller that doesn't pass it keeps
today's uninterruptible blocking behavior unchanged.
Extends the all-references-failed short-circuit (#56975) to the
persistent `provider: moa` facade path: MoAChatCompletions.create()
previously attached 'use the reference responses below' guidance built
entirely from failure sentinels and called the aggregator with it. Now
an all-failed turn attaches either the sanitized unavailability notice
(loud policy) or nothing (silent policy), and the aggregator — which IS
the acting model — simply acts alone. Advisor accounting for the failed
fan-out is still recorded.
Co-authored-by: liuhao1024 <sunsky.lau@gmail.com>
When every MoA reference model returns a failure (HTTP error, timeout,
etc.) or is skipped by the recursion guard, the one-shot aggregator
synthesis call is now skipped entirely. Previously it would try to
synthesise a wall of failure sentinels, which could block for the full
provider timeout (observed ~6 min on SenseNova) before returning a
non-retryable error that left the session hanging.
The early return carries the sanitized unavailability notice (never raw
provider error text, per the failed-reference containment) so the main
agent loop can still act in single-model mode.
Salvaged from #56975, reworked atop the _is_failed_reference helpers.
Follow-ups for salvaged #53784:
- reference_timeout now defaults to None = no per-preset override, so the
reference fan-out inherits auxiliary.moa_reference.timeout (900s default)
via call_llm's own per-task timeout resolution. The PR's 30.0s default
would have cut off long-thinking advisors mid-response, and its 300s max
cap capped legitimate explicit values — both removed. Explicit per-preset
values are still honored as-is.
- _is_failed_reference also treats '[skipped: …]' recursion-guard notes as
internal sentinels, keeping them out of both aggregator prompts.
- Dashboard/desktop TS types updated to number | null; web_server validator
accepts null/empty as 'inherit'.
Keep MoA reference display events off the machine-readable -Q stdout
surface (platform=cli with tool_progress_mode=off) while preserving them
everywhere else. Extracts the relay into module-level helpers so the
policy is testable.
Salvaged from #67334.
Adds per-reference progress events and a phase-transition marker to the
MoA display pipeline so TUI / CLI / desktop surfaces can render a status
bar like `MOA: 2/3 refs done` and surface which phase (reference vs
aggregator) is currently active.
- `moa.progress` — fired once per reference completion with
`refs_done`, `refs_total`, and the source label
- `moa.phase` — fired on phase transitions (currently the single
`phase="aggregator"` transition once the fan-out
finishes)
Plumbed through the existing `reference_callback` →
`tool_progress_callback` → gateway path; no new UI surface. The legacy
`moa.reference` / `moa.aggregating` events are unchanged for backwards
compatibility.
AI-assisted fix by https://github.com/SquabbyZ/peaks-loop
Salvaged from PR #47971 (LSP subset). On Windows, .cmd-wrapped language
servers (e.g. pyright-langserver.CMD launched via cmd.exe /c) and the
npm/go/pip LSP auto-installers spawn without CREATE_NO_WINDOW, so a
console window flashes whenever the spawn happens under a console-less
parent — e.g. a VS Code/Zed extension host running the ACP adapter.
- agent/lsp/client.py::_spawn: pass creationflags=windows_hide_flags()
to the language-server asyncio subprocess (inert 0 on POSIX;
start_new_session is kept — it is POSIX-only and ignored on Windows).
- agent/lsp/install.py: same flags on the npm and go installer
subprocess.run calls. The pip path goes through
hermes_cli.tools_config._pip_install, which already hides its windows.
Adapted from the PR's hand-rolled _NO_WINDOW constant to the repo's
hermes_cli._subprocess_compat.windows_hide_flags() convention.
Extends the fanout enum with 'every_n:<N>' (N >= 2): advisors run on the
first iteration of each user turn and every Nth tool iteration after it;
off-cadence iterations REUSE the cached guidance from the last on-cadence
run via the same cache mechanism the user_turn fanout uses, so the
aggregator still gets advice on every step. The cadence counter is scoped
per user turn (resets on a new user message) and only advances when the
advisory state actually changes, so streaming retries never consume a
cadence slot. Mapping form {mode: every_n, n: N} normalizes to the
canonical string. Unknown/degenerate values fall back to per_iteration.
Addresses issue #63393 (advisor fan-out multiplies turn latency/cost by
the tool-iteration count). Redesigned from PR #63448: the submitted shape
skipped references entirely on off-cadence iterations (aggregator ran
advice-less); this version keeps the last advice in play, credited for
the idea and cadence framing.
Config-gated, default-off (default fanout remains per_iteration).
Co-authored-by: webtecnica <75556242+webtecnica@users.noreply.github.com>
Adds test_moa_gemini_aggregator_sanitize_uses_real_model: drives a full MoA
tool-call turn (virtual-provider mode) with a Gemini aggregator and asserts
the strict-API sanitize pass is invoked with the resolved aggregator model
(gemini-3-pro-preview), never the virtual preset name once a slot is
resolved — the exact path that stripped extra_content/thought_signature and
made Gemini aggregators 400 (#65092).
Writing the test surfaced a gap in the salvaged #66212 fix: in virtual-
provider MoA mode (provider=moa, no moa_config threaded through
run_conversation) the conversation-loop branch never fired because it only
consulted moa_config. Extend it to fall back to the facade's
last_aggregator_slot — the same source the handle_max_iterations fix uses —
so both MoA entry modes resolve the real aggregator model.
Also adds the contributors/emails mapping for the #15676 credit base.
Follow-up to the salvaged core of #53802: a naive MoAClient(preset) rebuild
restores a working facade but silently drops the reference_callback relay
wired in agent_init, so moa.reference / moa.aggregating display events stop
reaching every frontend for the rest of the session.
Introduce agent.moa_loop.build_moa_facade(agent, preset) as the single
construction point for the MoA facade and use it at:
- initial client construction (agent_init.py)
- turn-start fallback restore (restore_primary_runtime)
- transient transport recovery (try_recover_primary_transport — previously
fell through to _create_openai_client with MoA's empty client_kwargs and
died with 'api_key client option must be set')
- mid-session model switches (switch_model)
The relay reads agent.tool_progress_callback at emit time, so callbacks
attached after construction are picked up automatically.
Adds test_moa_restored_facade_still_emits_reference_events covering event
delivery through a restored facade.
When Hermes fails over from a non-Gemini provider (xAI, Anthropic, etc.) to
Gemini mid-conversation, the existing assistant tool_calls in history carry
no Gemini ``extra_content.google.thought_signature`` (the originating provider
never emits one). The native adapter's ``_translate_tool_call_to_gemini``
omitted ``thoughtSignature`` entirely in that case, so Gemini 3 thinking
models rejected every replayed turn with::
HTTP 400 INVALID_ARGUMENT
Function call is missing a thought_signature in functionCall parts.
Additional data, function call default_api:<tool_name>, position N.
The Cloud Code Assist sibling adapter already handles this exact case by
emitting a sentinel ``"skip_thought_signature_validator"`` (see
``agent/gemini_cloudcode_adapter.py:106``, originally added in #11270 and
documented as matching ``opencode-gemini-auth``'s approach). This change
mirrors that fallback in the native adapter so the two paths behave
identically when replaying cross-provider history.
Verified live against ``generativelanguage.googleapis.com/v1beta`` with
``gemini-3-pro-preview``: synthetic 2-turn conversation with no real
``thoughtSignature`` returns 400 without the sentinel and 200 with it.
Test added: ``test_build_native_request_emits_sentinel_for_cross_provider_tool_call``.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
When MoA mode is active with a Gemini model as the aggregator,
agent.model holds the virtual preset name (e.g. "closed"), not the
actual aggregator model name. The _sanitize_tool_calls_for_strict_api
call uses agent.model to decide whether to keep extra_content
(thought_signature) on tool_calls — since "closed" doesn't contain
"gemini", the thought_signature is stripped and the Gemini aggregator
rejects the next request with HTTP 400 (INVALID_ARGUMENT):
"Function call is missing a thought_signature in functionCall parts."
Fix: resolve the actual aggregator model name from moa_config
(conversation_loop) or last_aggregator_slot (chat_completion_helpers)
and pass it to _sanitize_tool_calls_for_strict_api so the
_model_consumes_thought_signature check sees the real Gemini model.
Closes#65092
Bare ContextCompressor.__new__ doubles (test_compress_focus,
cross_session_guard, image_tokens, pre_compress_memory_context) skip
__init__ and lack the attribute — the documented compression-path
test-double pitfall. Guard with getattr default 1 + int type pin
(bool excluded).
Follow-up fixes on top of the salvaged #22566 mechanism:
- N-collector now counts only REAL actionable user turns via
_is_actionable_user_turn + _is_synthetic_compression_user_turn —
the same filter pair _find_last_user_message_idx uses post-#69291.
The contributor's bare role=='user' + _is_context_summary_content
check let blank platform echoes and continuation/todo rows consume
N slots, silently degrading the guarantee.
- Default flipped 3 -> 1 (behavior-preserving): a default of 3 was
measured to change the tail cut on transcripts whose budget covers
only the last turn. min_tail_user_messages=1 delegates to the
existing single-user anchor; N>1 is opt-in, and the call site is
gated so the default path is byte-identical to main.
- Hardened config parse in agent_init (bool rejected, fractional
floats rejected, floor 1) matching the max_attempts parser shape.
- Wired the recurring external-PR config gaps: hermes_cli/config.py
DEFAULT_CONFIG + cli-config.yaml.example (PR only had cli.py).
- Regression tests: blank echoes / synthetic rows don't count toward
N; tool-call/result pairs never split by the N-boundary (no-orphan
both directions); N-guarantee wins over tail_token_budget and the
_MAX_TAIL_MESSAGE_FLOOR (floor is a minimum, not a cap); default
parity pin; DEFAULT_CONFIG pin.
Add _ensure_last_n_user_messages_in_tail to guarantee the last N user
messages survive compression in the uncompressed tail, with surrounding
assistant/tool context preserved.
- Add min_tail_user_messages parameter (default 3) to ContextCompressor
- New _ensure_last_n_user_messages_in_tail method generalizes single-user protection
- Skip context-summary handoff banners when counting user messages
- User messages are clean boundaries — skip _align_boundary_backward
- Wire through cli.py, agent_init.py, and gateway cache busting keys
Config:
compression:
min_tail_user_messages: 3
Co-Authored-By: Claude <noreply@anthropic.com>
_collect_ghosted_skill_names() covers both ghost-skill shapes in the
compressed middle window: rows already demoted to a [SKILL_PRUNED: ...]
marker AND raw skill_view bodies (> _SKILL_VIEW_PRUNE_MIN_CHARS) that
survived Phase-1 inside an earlier protected tail and then aged into the
compression window — the summarizer paraphrases those instructions away
too. Shared threshold constant between the emit site and the scan.
Pinned by a live-probe-shaped test (real compress(), mocked aux LLM).
Salvage rework of PR #44166 (@dolphin-creator) onto current main:
- ONE canonical prune marker: _skill_pruned_marker(name) builds
'[SKILL_PRUNED: ... reload with skill_view(name='X')]'; both emit
sites and the survival presence check use the same string, fixing the
original PR's defect where the emitted marker was '[SKILL_PRUNED:'
but the presence check looked for '[SKILL_PRUNED]' (re-injection
duplicated markers that had survived).
- Phase-1 prune (_prune_old_tool_results) now threads a protected-skill
set: skills whose skill_view call is within the last 10 messages, in
the protected tail, or named in a tail user message keep their full
bodies. Pass-4 pressure demotion deliberately overrides the guard so
the #61932 dead-end shape cannot return.
- P2 deterministic marker survival: skill names are extracted from the
summarizer INPUT (and the previous summary) before the aux LLM call
and any dropped canonical markers are re-injected afterward under a
'## Pruned Skills' section — routed through _redact_compaction_text,
appended to the summary body only (never in front of SUMMARY_PREFIX
or scaffolding start-of-content markers; classify_summary_content is
unaffected). Same treatment on the static fallback path, re-applied
after its size cap since truncation cuts exactly where markers land.
- Summarizer prompt gains a '## Pruned Skills' copy-verbatim section.
Fixes#32106.
Surgical reapply of the marker-alignment and dedup-guidance halves of
PR #44166 commits 52341f6ca3 / 3d8a31432d / ae07412e4b onto current main:
- the [SKILL_PRUNED: ...] marker embeds the exact reload call
skill_view(name='<skill>') so the model can act without guessing
- SKILLS_GUIDANCE Skill Safety Rule gains rule 4 (DEDUP): after one
reload, remaining markers for the same skill are historical artifacts
Fixes#32106 (part).
The Vertex AI provider (added same-day, commit c73e74386) was never added to
either of the two provider registries that agent/auxiliary_client.py and the
MoA slot-resolution chain depend on, breaking Vertex outside the main
conversation loop:
1. hermes_cli/auth.py::PROVIDER_REGISTRY had no "vertex" entry. The
plugin-auto-extend loop that normally fills gaps explicitly skips
non-api_key auth types (`if _pp.auth_type != "api_key": continue`), and
Vertex was never hand-declared like "bedrock" is. Because
resolve_provider_client() in agent/auxiliary_client.py gates everything
on `pconfig = PROVIDER_REGISTRY.get(provider)` and returns (None, None)
immediately when pconfig is None, its `elif pconfig.auth_type == "vertex"`
branch was permanently dead code — every auxiliary Vertex call (vision,
title generation, reflection, context compression, MoA reference/
aggregator slots) failed outright, not just a MoA-specific edge case.
2. hermes_cli/providers.py::HERMES_OVERLAYS also had no "vertex" entry, so
hermes_cli.providers.get_provider("vertex") returned None. This backs
_preserve_provider_with_base_url() in agent/auxiliary_client.py, which a
MoA slot's resolved (base_url, api_key) pair needs to keep its "vertex"
identity instead of silently collapsing to "custom" — losing the
identity _refresh_provider_credentials() needs to re-mint an expired
OAuth2 token (~1h lifetime) on a 401, and permanently breaking every
subsequent call in that MoA preset for the rest of the session.
Fix mirrors the existing "bedrock"/aws_sdk entries in both registries
exactly, plus adds a "vertex" branch to _refresh_provider_credentials() (it
had branches for openai-codex/nous/anthropic/xai-oauth but not vertex,
so a 401 fell through to `return False` without evicting the stale cached
client).
- hermes_cli/auth.py: hand-declared vertex ProviderConfig(auth_type="vertex")
in PROVIDER_REGISTRY, matching bedrock's shape.
- hermes_cli/providers.py: vertex HermesOverlay(auth_type="vertex") in
HERMES_OVERLAYS + "Google Vertex AI" label override.
- agent/auxiliary_client.py: vertex branch in _refresh_provider_credentials
that re-mints the token via get_vertex_config() and evicts the stale
cached client.
- 8 new regression tests across tests/hermes_cli/test_vertex_provider.py and
tests/agent/test_auxiliary_client.py: registry membership, end-to-end
resolve_provider_client("vertex", ...) building a working client (proving
the previously-dead branch is now reachable), and the 401-refresh/cache-
eviction path.
Follow-ups on top of the cherry-picked #27748 mechanism:
- move the cap constant to module level with full rationale comment
(class attribute aliases it so subclasses/tests can override)
- bound the iterative-update path too: the PREVIOUS SUMMARY block is
passed through _bound_summary_input so a pathological rehydrated
handoff cannot blow up the prompt (previous summary + new turns each
capped)
- extra regression tests: byte-identical small-input passthrough
(identity), direct bound+marker unit check, bound-after-per-message-
truncation shape (hundreds of under-_CONTENT_MAX turns), iterative
path bounded, marker vs classify_summary_content non-collision
- contributor email mapping for @robgfl45
The phase-1 tool-result prune only runs inside compress(), which fires
near 50% of the context window, so it never triggers on large-window
models; old tool outputs then ride in history and are re-sent every turn.
Add prune_tool_results_only(): the same no-LLM prune on a separate, low
proactive_prune_tokens trigger, run as an elif to the compression branch.
Opt-in (default 0), protects the recent tail by message count.
Add the method to the ContextEngine base as a no-op default so pluggable
engines inherit it safely (the post-tool-call path never AttributeErrors on
a non-built-in engine); the built-in compressor supplies the real prune.
Register both keys under the top-level compression config with defaults and
document them.
Follow-up to the #54805 cherry-pick (skill guard rules):
- turn_context.py: snapshot_preflight_display_tokens gets a
getattr+callable guard (SimpleNamespace compressor doubles / plugin
context engines lack the ContextCompressor-only method) and the
snapshot value is type-pinned to a real int (bool excluded) before
arming the rollback — MagicMock compressors return truthy Mocks.
- turn_finalizer.py: interrupted pinned 'is True', the
_turn_received_provider_response read pinned 'is not True' (MagicMock
auto-attrs are truthy), and the compressor rollback method call gets a
getattr+callable guard. Rollback stays display-only: it never touches
_ineffective_compression_count or any durable guard, and preserves the
-1 post-compaction sentinel.
Interrupted turns can seed a speculative display token count before the provider receives the request. Restore that display-only seed when interruption wins the race, while preserving completed post-compaction state and treating a successful provider response independently of optional usage metadata.
Constraint: #54776 remains reproducible on current main, while review #4702305384 identifies anti-thrashing rollback as stale and usage receipt as an unreliable response-completion signal.
Rejected: Restore anti-thrashing counters from a preflight snapshot | current main derives their verdict from real provider usage after a completed compaction boundary.
Confidence: high
Scope-risk: narrow
Directive: Keep interrupted preflight rollback display-only, and never infer provider completion from the presence of usage metadata.
Tested: ./.venv/bin/python -m pytest -q tests/run_agent/test_413_compression.py (29 passed); turn-finalizer/conversation-loop tests (31 passed); context-compressor targeted tests (12 passed); infinite-compaction targeted tests (3 passed); ruff; git diff --check.
Not-tested: End-to-end interactive interrupt through CLI or gateway transport.
A lock-loser compression pass returns its input unchanged, which the
automatic compression sites misread as 'cannot compress further': the
preflight loop armed the insufficient-progress blocker, the pre-API gate
burned a shared attempt, and a lock-contended 413/overflow retried into
the attempt cap and returned compression_exhausted — which the gateway
answers with a full session auto-reset (#9893/#35809). A temporary
concurrent-compression defer wiped the session.
Consume the landed #69870 lock-skip signal on every automatic path
(preflight in turn_context, pre-API pressure gate, 413 handler, overflow
handler, post-tool compaction): when a pass no-ops AND the type-pinned
lock-skip flag is set, refund the attempt (never count it toward the cap
or the insufficient-progress blocker), and when the turn cannot proceed
(provider already proved the request does not fit) end it with a soft
compression_deferred result — distinct from compression_exhausted — so
the gateway keeps the session intact and the next message retries after
the concurrent compressor finishes.
The new compression_skipped_due_to_lock() reader is type-pinned
(is True or isinstance(str)) per the MagicMock auto-attribute rule, and
compress_context() now also clears the signal at the very top of every
attempt (per-attempt state rule, #58629/#69853) so a stale value can
never make a later breaker/codex no-op look like lock contention.
Salvaged from PR #49874; rebuilt on main's #69870
_compression_skipped_due_to_lock signal instead of the PR's parallel
_compression_deferred_by_lock triple.
MoA reference_max_tokens is preset-level — one cap for all reference
models. When mixing a verbose model with a terse one, a single cap is
either too tight for the terse model or too loose for the verbose one.
Now each reference slot can optionally carry its own max_tokens:
reference_models:
- provider: openrouter
model: deepseek/deepseek-v4-pro
max_tokens: *** # per-slot cap, overrides preset-level
- provider: openai-codex
model: gpt-5.5
# no max_tokens → falls back to preset-level reference_max_tokens
_clean_slot (moa_config.py) preserves an optional max_tokens field on
the slot dict, coerced via _coerce_int_or_none. _run_reference
(moa_loop.py) reads slot-level max_tokens first, falling back to the
preset-level cap passed by the caller. Slots without the field are
unaffected — backward compatible.
Type hints on slot-handling functions updated from dict[str, str] to
dict[str, Any] to reflect the now-heterogeneous slot shape.
_build_call_kwargs omitted max_tokens for every provider except
anthropic-compat endpoints and NVIDIA NIM. Gemini's native
generateContent maps max_tokens -> maxOutputTokens and, when it is
omitted, applies a fixed 65,535-token ceiling (not "the model's full
budget"), so dropping the value made MoA's reference_max_tokens a
silent no-op for gemini advisors — they ran effectively uncapped
(observed ~2900 output tokens against a configured cap of 600),
inflating per-turn MoA latency.
Forward max_tokens for the gemini-native path (provider name or native
base_url). Gemini supports maxOutputTokens, so the cap is safe here;
providers that reject max_tokens (Copilot, GPT-5 max_completion_tokens,
ZAI vision) are unaffected — they still omit it as before.
Per review feedback from teknium1: reference_max_tokens is an advisors-only
contract. The aggregator is the acting model and must not be capped by the
reference budget. Changed _is_moa from startswith('moa_') to exact match on
'moa_reference'. Added regression test proving aggregator does NOT receive
max_tokens.
Copilot review pointed out that hardcoding kwargs['max_tokens'] would
400 on models requiring max_completion_tokens (GPT-5 family, Copilot).
The existing auxiliary_max_tokens_param() helper already selects the
correct parameter name per model — use it instead of hardcoding.
Test updated to parametrize expected_key so the Copilot gpt-5.5 case
correctly asserts max_completion_tokens instead of max_tokens.
Addresses Copilot review comments on both files.
PR #56756 added reference_max_tokens to cap MoA advisor output and cut
turn latency. The value is correctly threaded through five layers of MoA
code (moa_config → conversation_loop → aggregate_moa_context →
_run_references_parallel → _run_reference → call_llm(task='moa_reference',
max_tokens=800, ...)).
However, _build_call_kwargs() in auxiliary_client.py silently drops
max_tokens for all OpenAI-compatible providers (PR #34845, which fixed
endpoints and NVIDIA NIM keep it. This means reference_max_tokens never
reached the API for the vast majority of providers.
The bug affects every OpenAI-compatible MoA reference/aggregator slot:
Z.AI (coding plan), OpenRouter, OpenAI, GitHub Copilot, and local
providers. Only Anthropic-compat endpoints (MiniMax, /anthropic URLs)
worked — by coincidence, not MoA-aware design.
Fix: thread the 'task' parameter through all six _build_call_kwargs()
call sites. When task starts with 'moa_', max_tokens is always included
in the request kwargs regardless of provider. Non-MoA auxiliary tasks
(compression, titles, vision, etc.) keep PR #34845 behavior unchanged.
Verified end-to-end:
- Z.AI GLM-5.2 with max_tokens=50 → returned exactly 50 tokens
- Z.AI GLM-5.2 with max_tokens=20 → returned exactly 20 tokens
- Z.AI GLM-5.2 uncapped → returned 315 tokens
- 7 new regression tests covering 4 providers, Anthropic wire, non-MoA
tasks, and prefix-matching boundary
- 288 auxiliary_client tests pass (was 281, +7 new), 84 MoA tests pass
- Zero regressions
aggregate_moa_context's single max_tokens parameter was applied to
both the reference fan-out (_run_references_parallel) and the
aggregator's own synthesis call_llm. #53580 explicitly removed a
hardcoded cap from the aggregator call because it truncated long
aggregator syntheses; #56756 (reference_max_tokens, added to speed up
the advisor fan-out) reintroduced the same shared cap by passing it to
both calls, silently regressing #53580's fix.
Rename the parameter to reference_max_tokens (matching the caller's
own moa_config key) and stop forwarding it to the aggregator's
call_llm invocation, which now always runs uncapped as intended.
An assistant message with an empty or whitespace-only text content block —
produced by context compression or certain tool-call flows — is rejected by
the Anthropic Messages API with HTTP 400 "text content blocks must contain
non-whitespace text". Because the blank block is stored in session history and
replayed verbatim every turn, the session is permanently wedged behind the same
400.
The Bedrock adapter already guards this via _safe_text() (#9486); the native
Anthropic path never got the same treatment. _sanitize_replay_block() rebuilt
text blocks with the raw stored text, and the _convert_assistant_message()
guard only caught a fully empty block list, not a list still containing a
whitespace-only text block.
Add a _safe_text() helper mirroring the Bedrock one and apply it at both points:
the ordered-blocks replay path and a final in-place walk of the converted
content list. Both are self-healing — sessions that stored blank blocks recover
on the next API call. Only text blocks are coerced; thinking/tool_use/image
blocks are untouched.
Fixes#69512
Correction to the previous commit (PR #68402): the claim that api_server
never intercepts MEDIA: tags is inaccurate on current main.
_resolve_media_to_data_urls() (gateway/platforms/api_server.py) DOES
inline image MEDIA: tags (<=5MB, image extensions only) as base64 data
URLs on the four main endpoints (_handle_session_chat,
_handle_session_chat_stream, _handle_chat_completions, _handle_responses).
The real gaps elphamale's PR points at are narrower:
- the /v1/runs output path (_handle_runs) never calls the resolver;
- non-image filetypes are never resolved anywhere (_MEDIA_IMG_EXT is
image-only).
Reword the hint to teach both halves: images via MEDIA: work on the
chat/completions/responses endpoints; non-image files and anything on
the runs endpoint must fall back to plain file paths in the response
text. Update the test to pin the scoped guidance instead of a blanket
prohibition.
Every PLATFORM_HINTS entry for a messaging platform (Telegram, WhatsApp,
Discord, Slack, Signal, WebUI, desktop) teaches the model the MEDIA:/path
convention because an interception mechanism actually resolves it there
(native attachment delivery, or a validated/inlined data URL). The cli
entry, which has no such mechanism, explicitly tells the model NOT to use
it and to state the path in plain text instead.
The api_server entry had neither instruction. Its /v1/runs handler never
routes the final response through any MEDIA: resolver (confirmed against
source: none of the four call sites of the api_server module's media-tag
resolver are inside its runs-endpoint handler), so a MEDIA:/path tag there
renders as inert literal text in the API response — exposing a raw host
filesystem path to the caller with no delivery ever taking place. Nothing
platform-specific told the model not to use a convention it's correctly
taught for several sibling platforms in this same dict, so the general
cross-platform habit could surface here too, unlike cli where an explicit
prohibition already closes the gap.
Mirrors cli's prohibition, adapted for api_server's actual constraint: no
"state the path in plain text" fallback, since a typical API caller has no
filesystem access to the host at all. Points at "a registered file-delivery
tool" generically rather than naming any specific tool, since api_server
toolsets are deployment-defined.
The api_server adapter returned error code "invalid_api_key" for
API_SERVER_KEY authentication failures, which the Desktop error
classifier misidentified as a provider (OpenRouter/OpenAI) key
problem — showing "OpenRouter API key missing" when the real issue
was gateway auth.
Changes:
- gateway/platforms/api_server.py: return "gateway_auth_failed" code
with descriptive message for API_SERVER_KEY auth failures
- apps/desktop/src/store/notifications.ts: add "gateway_auth_failed"
handler before "invalid_api_key" to show correct error message
- agent/error_classifier.py: add "gateway_auth_failed" to auth patterns
- tests: update test_session_api.py to expect new error code
Fixes#39365
* fix(desktop): hide persisted agent-only history scaffolding
Filter verification-stop nudges and context-compaction handoffs at the
stored-history mapper boundary. Preserve a real reply when a compaction
handoff shares its stored message.
* test(desktop): build persisted E2E sessions through the real agent
Drive tui_gateway.entry over its stdio JSON-RPC transport against the mock
provider, wait for real completion events, and persist normal session history
through AIAgent and SessionDB. Migrate resume and hidden-history coverage,
including real compression and live verify-on-stop scaffolding, then remove
the unused direct SessionDB import scripts.
* fix(desktop): use the provisioned Python for real-session E2Es
Run the stdio gateway through uv's synced project environment outside the
Nix dev shell, while retaining the fully provisioned Nix Python when the
shell advertises HERMES_PYTHON_SRC_ROOT.
* fix(nix): expose the provisioned Python environment to uv
Mark the Nix-built Python environment active in the dev shell so the shared
E2E session builder can always run through `uv run --active --no-sync`.
* fix(timeline): persist typed display events
* fix(timeline): strip display-only fields from provider payloads, preserve through rewrites, fix /resume display history
Three review findings from PR #69771:
1. Provider payload leak: display_kind and display_metadata were forwarded
to the provider API as unknown message fields. Strict OpenAI-compatible
backends can reject the next request after a model switch or resumed
typed event. Strip both from the per-request api_msg copy in
conversation_loop alongside the existing api_content pop.
2. Rewrite/import data loss: _insert_message_rows preserved display_kind
but silently dropped display_metadata. After replace_messages,
archive_and_compact, or session import, async-delegation completion
events lost their task counts and fell back to generic display text.
Add display_metadata to the INSERT columns and bind tuple.
3. CLI /resume stale recap: startup --resume A set _resume_display_history
from A's lineage. A subsequent in-session /resume B loaded B only into
conversation_history via get_messages_as_conversation, leaving the stale
A display projection. _display_resumed_history preferentially read the
stale attribute, showing A's recap for B. Switch /resume to
get_resume_conversations and update _resume_display_history alongside
conversation_history.
Tests: 890 Python (5 files), 35 desktop TS — all green.
* feat(tui): render typed display events as ◈ markers in the Ink TUI
The TUI was not handling display_kind at all — model switch markers and
async delegation completions rendered as opaque user messages with the
full [System: ...] text, and hidden compaction handoffs were visible.
Wire display_kind through the full TUI chain:
- _history_to_messages (tui_gateway/server.py) forwards display_kind
and display_metadata to the gateway transcript payload.
- GatewayTranscriptMessage (gatewayTypes.ts) gains both fields.
- Msg.kind (types.ts) gains 'event' value.
- toTranscriptMessages (domain/messages.ts) maps:
- hidden → skip entirely
- model_switch → event "model changed"
- async_delegation_complete → event "N background agents finished"
(or "background agent work finished" without metadata)
- messageGroup (blockLayout.ts) routes event to its own group, with
SELF_SPACED + PAINTS_TRAILING_GAP so it owns its margins.
- messageLine.tsx renders event-kind as a dim ◈ marker with no gutter,
matching the CLI's ◈ event rendering.
- 4 new TUI tests for hidden/model_switch/async_delegation mapping.
TUI typecheck: clean. TUI lint: 0 errors (2 pre-existing warnings).
TUI tests: 9 passed (1 pre-existing failure on main, unrelated).
Claude-on-Bedrock already gets prompt caching via the AnthropicBedrock
SDK path. This adds it to the raw Converse API path used for non-Claude
models (Amazon Nova, and Claude when bearer-token auth forces Converse
routing, #28156) — a conservative model allowlist inserts cachePoint
blocks after tools, system, and the message before the newest turn, and
extracts cacheReadInputTokens/cacheWriteInputTokens into usage so cost
accounting picks them up through the existing Anthropic-style fallback.
Ref: relatorio-cache-performance-provedores-ia.md, P0 item 1.
'auto' is a sentinel meaning "inherit from main runtime / auto-detect",
not a literal model id -- already handled for cfg_model (config-derived)
in _resolve_task_provider_model, but not for the explicit `model` kwarg.
MoA reference/aggregator slots (agent/moa_loop.py's _slot_runtime) forward
a preset's `model:` field as this explicit argument rather than through
auxiliary.<task> config, so a MoA preset configured with `model: auto`
(a natural thing to try given the existing auxiliary.*.model: auto
convention) reached this function as the explicit `model` arg and took
the `model or cfg_model` branch, bypassing the cfg_model-only sentinel
check entirely -- sending the literal string "auto" to the wire as a
model id.
Normalize both the explicit `model` and `cfg_model` the same way, fixing
this at the single chokepoint every caller (MoA included) already goes
through, rather than patching moa_loop.py separately.
Follow-up to the #62614 salvage: try_refresh_matching (added by the
#69843 salvage after this PR's base) calls self.current() while already
holding the now-locking non-reentrant pool lock — a guaranteed deadlock
that git merges silently (no textual conflict). Use _current_unlocked()
and cover the method in the no-deadlock test.
Follow-up to review feedback:
- Acquire self._lock in the remaining public pool-state methods:
has_credentials, reset_statuses, remove_index, resolve_target, and
add_entry. All of them read or rebind self._entries (and the mutating
ones persist auth.json), so they now hold the same lock as select()
and the query methods. None are called from within the lock, so no
unlocked helpers are needed.
- Make the blocking test deterministic: an instrumented lock records the
acquire attempt, and the test first waits for the worker to actually
reach self._lock before asserting it blocks. Previously an unlocked
method could pass if the worker thread was scheduled late.
- Extend the lock test matrix to all nine public methods; the five newly
locked ones fail the test without this fix.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
`has_available()`, `peek()`, `current()` and `entries()` read (and, via
`_available_entries()`, mutate and persist) `self._entries` without holding
`self._lock`, while every other entry point — `select()`,
`mark_exhausted_and_rotate()`, `acquire_lease()`, `try_refresh_current()` —
guards the exact same access with the lock.
`_available_entries()` is not read-only: it prunes aged-out DEAD manual
entries (rebinding `self._entries` at the prune step) and calls `_persist()`
(writes auth.json). The gateway runs platform adapters in threads and cron
runs jobs in a ThreadPoolExecutor, so a status probe via `has_available()`
or `peek()` can race a concurrent `select()`/rotation: torn iteration of
`self._entries`, interleaved auth.json writes, or a lost token rotation.
Fix: take `self._lock` in all four query methods. Because the lock is
non-reentrant and `peek()` composes `current()` + `_available_entries()`,
add a lock-free `_current_unlocked()` helper and route the already-locked
internal callers (`_select_unlocked`, `mark_exhausted_and_rotate`,
`_try_refresh_current_unlocked`) through it to avoid self-deadlock.
Added regression tests: a no-deadlock check (peek re-entrancy) and a
lock-held-blocks-the-call check for each of the four methods.
Two related races in credential-pool cooldown state:
1. Lost update across processes: write_credential_pool merged only
entries missing from the caller's snapshot; for entries present on
both sides the caller's in-memory copy won wholesale. A process
holding a snapshot taken before another process marked a key
exhausted would, on its next persist (e.g. a round-robin rotation),
write the key back as healthy — erasing the cooldown so every
process resumes hammering a rate-limited key. Merge status fields by
last_status_at recency: adopt the on-disk status only when it is
strictly newer AND still binding (DEAD, or EXHAUSTED with an
unexpired cooldown), and never onto re-authed (token-changed)
entries, so legitimate expiry-clears and fresh logins are preserved.
2. Wrong-key quarantine: when mark_exhausted_and_rotate received an
api_key_hint that matched no entry, it fell through to
current()/_select_unlocked() — on a freshly loaded pool that selects
the NEXT healthy key and benches it for the full cooldown TTL,
punishing an innocent credential. When a hint is provided but
unmatched, rotate without marking anything instead of guessing.
Includes regression tests.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
A 402/429/401 is an API-key–level failure (account out of balance,
rate-limited, or key rejected), but the same key can back more than one
pool entry — e.g. an explicit pool entry plus a `model_config` entry
auto-seeded from `model.api_key`, both carrying the identical
`runtime_api_key`.
`mark_exhausted_and_rotate(api_key_hint=...)` only marked the *first*
matching entry, leaving the sibling OK. `_select_unlocked()` then kept
handing back the same depleted key, so the billing-recovery `continue`
loop in the conversation retry path never converged: the request hung
until the client disconnected (~2.5min observed against DeepSeek),
emitting only `response.created` with no 402 ever surfaced to the user.
Mark every entry sharing the failed key so the pool can reach the
"no available entries" state and let the error propagate immediately.
Adds a regression test covering two entries backed by the same key.
The dedup-reset calls assumed a full AIAgent; gateway/loop test doubles
built via object.__new__ lack _clear_context_overflow_warn and crashed
in build_turn_context (caught by test_api_content_sidecar on CI slice 3).
getattr-guard all four call sites per the established test-double pitfall
pattern (AGENTS.md #17).
Follow-up fixes for the #62625 salvage:
- Dedup-reset gap (sweeper review): when the block clears while the
context is STILL over threshold, execution enters the compression
branch — the PR's 'else' reset never ran, so the warning stayed
suppressed forever after the first block. _clear_context_overflow_warn()
now fires on every automatic compression path: turn-context preflight,
conversation_loop pre-API gate, and the post-tool loop-compaction gate.
- should_compress_info on current main: main refactored should_compress
into _automatic_compression_blocked()/_locally(); the tuple variant now
derives its reason from the same in-memory state via
_compression_block_reason(), keeping cooldown:<s>/ineffective shapes.
- ContextEngine.should_compress_info ABC default now actually returns
(should_compress(tokens), None) — the PR's default had a docstring but
no return (returned None, would crash tuple-unpacking call sites).
- Below-threshold guard: the turn-context persisted-cooldown branch and
the conversation_loop pre-API cooldown branch no longer warn when the
estimate is under threshold (should_compress_info returns a None
reason; the preflight pre-check is not a threshold guarantee). The
pre-API guard also honors compression.max_attempts instead of a
hardcoded 3, and no longer fabricates a cooldown reason.
- Noise-filter survival (#69550 composition): warning text is now a
template constant (CONTEXT_OVERFLOW_BLOCKED_WARNING_TEMPLATE) marked
FAILURE-CLASS, pinned un-swallowed in VISIBLE_COMPRESSION_MESSAGES and
in new tests that execute the real _TELEGRAM_NOISY_STATUS_RE +
_prepare_gateway_status_message.
- Contributor mapping for stanislav@local -> sl4m3.
- ContextEngine.should_compress_info() default impl so plugin engines
(e.g. _StubEngine) don't raise AttributeError at the call site.
- Centralise warning/reset in AIAgent._warn_context_overflow_blocked /
_clear_context_overflow_warn so turn-context and conversation-loop guards
share identical dedup logic and reset on the real compression boundary.
- Cover conversation_loop.py pre-API (~L1007) and loop-compaction (~L4774)
guards, not just the turn-context preflight.
- _FakeAgent mirrors the two helpers; test suite green (219 passed).
Fixes#62708
Previously, when a session crossed the compression threshold but compression
was skipped (summary-LLM cooldown, #11529, or anti-thrashing, #40803), the
model kept accumulating context until it hit the hard provider token limit and
silently stopped answering — with no signal to the user about why.
Changes:
- context_compressor.should_compress_info() returns a (should_compress, reason)
tuple. reason is 'cooldown:<seconds>' or 'ineffective' when compression is
needed but blocked. should_compress() keeps its bool contract so existing
callers (conversation_loop.py) and regression #29335 are unaffected.
- turn_context.build_turn_context() emits a deduped _emit_warning when the
context is over threshold but compression is blocked, advising /new or
/compress. Dedup keys on the block *kind* (cooldown/ineffective), not the
ticking countdown, so a cooldown doesn't re-fire the warning every turn.
- Adds tests/agent/test_turn_context_overflow_warning.py covering the tuple
shape, both block kinds, dedup, and re-fire-after-clear.
Extend the existing candidate-name resolver in _supports_vision_override
to accept 'vision' as an alias for 'supports_vision' on per-model config,
for both the providers.<name>.models dict and the legacy list-style
custom_providers form.
Per review feedback on #31912: this extends the current resolver rather
than replacing its candidate-name logic. Named custom providers resolve
to the runtime value provider='custom' while the config keeps the
user-declared name under model.provider; that lookup path is preserved.
Adds regression tests covering model.provider=my-vllm with runtime
provider='custom' for both config shapes.
Follow-up to the salvaged #57634 commits:
- agent/manual_compression_feedback.py: new describe_compression_lock_skip()
— single source of truth for lock-skip wording. A descriptive holder
string means another compressor CONFIRMED holds the lock ('already in
progress (holder: ...)'); True/None means acquisition failed without a
confirmed holder (hermes_state.try_acquire_compression_lock catches
sqlite3.Error internally and returns False), so the message says
'could not acquire ... the lock check failed' instead of falsely
claiming a concurrent compression is running.
- cli.py, gateway/slash_commands.py, tui_gateway/server.py (all three
in-process consumers: session.compress RPC, command.dispatch compress
branch, slash.exec mirror) now route through the shared helper.
- tui_gateway/server.py command.dispatch compress branch: catch
CompressionLockHeld explicitly — it previously fell into the generic
'compress failed' error handler.
- Deferred-notify contract (#69324): lock-skip discards the pending
context-engine notification (committed=False) in _compress_session_history
and the CLI path before returning.
- tests: lock-skip wording pins per surface, VISIBLE_COMPRESSION_MESSAGES
noise-filter carve-outs for both wordings, MagicMock signal opt-outs for
sibling tests added on main after the original PR.
Advisor review found a critical stale-signal leak: if auto-compress
sets _compression_skipped_due_to_lock during a lock-skip, a subsequent
successful manual /compress will see the stale signal, falsely report
'Compression already in progress', and discard the compression results.
Fix:
- compress_context clears _compression_skipped_due_to_lock = None at
entry so each call's outcome alone determines the signal.
- Unified gateway 'holder: unknown' drift to match CLI/TUI pattern
(omit holder clause when not a descriptive string).
- Added MagicMock opt-outs in 3 sibling test files broken by the new
signal check (test_compress_here, test_compress_focus,
test_compress_plugin_engine).
- Added stale-signal-leak invariant test proving the fix.
The anti-thrash guard (_ineffective_compression_count) was in-memory
only: a fresh compressor bound to a resumed, already-compacted session
started with compression_count=0 and a disarmed guard, so a
near-threshold session could legally re-compact once per process
restart, forever.
Persist the counter through the durable session-state channel,
mirroring the failure-cooldown (#54465) and fallback-streak (af7dceaf7)
pattern:
- hermes_state.py: sessions.compression_ineffective_count column
(declarative reconciliation adds it on existing DBs) +
get/set_compression_ineffective_count accessors.
- context_compressor.py: every strike/clear verdict routes through
_record_ineffective_compression_verdict() which writes through to the
session row (no-change verdicts skip the DB write);
bind_session_state() loads the persisted value; the compression
rotation boundary carries the counter onto the child row;
update_model()'s reset also clears the durable copy; the
ineffective-only fast path in _automatic_compression_blocked() is
removed because the counter is now durable and another agent's clear
must unblock a stale local snapshot.
- conversation_compression.py: _refresh_persisted_compression_guards
re-reads the counter alongside cooldown + fallback streak.
Reset semantics are unchanged: any real provider reading below the
threshold still clears the counter — and now clears it durably too.
Resolves the residual gap identified in #54923 by @lanyusea (the
second-threshold mechanism was superseded by persisting the existing
guard state).
Co-authored-by: lanyusea <lanyusea@gmail.com>
Follow-up to srojk34's explicit-provider unwrap (PR #56691):
- Extract _resolve_moa_aggregator() as the single preset->aggregator
resolver shared by _resolve_auto(), _resolve_task_provider_model(),
and resolve_provider_client() so preset lookup/validation can't drift.
- When the main provider is moa, the aggregator model is now the default
for every UNSET auxiliary model: _read_main_model_for_aux() substitutes
the preset's acting (aggregator) model wherever fallback chains
pre-filled from _read_main_model() (router prefill, custom-endpoint
fallback, named-custom default, external-process default,
_try_main_agent_model_fallback).
- Unwrap moa at the resolve_provider_client() chokepoint so direct
callers (vision auto-detect, plugin code) can't dead-end in the
unknown-provider branch, and unwrap the vision auto-detect main
provider before capability probes run against the preset name.
- Real-config tests: temp HERMES_HOME + actual config.yaml exercising
the genuine load_config()/resolve_moa_preset() boundary.
_resolve_task_provider_model() returned an explicit provider="moa" override
(from a caller-passed arg, or auxiliary.<task>.provider: moa in config.yaml)
verbatim, with no MoA-preset unwrap. Only the *implicit* "main provider is
moa" path inside _resolve_auto() unwraps to the aggregator slot (#53827) —
this function never goes through _resolve_auto() at all, so the explicit
case was never covered.
MoA is a virtual provider with no real HTTP endpoint: resolve_provider_client()
looks "moa" up in PROVIDER_REGISTRY (no such entry), falls to the
unknown-provider dead end, and call_llm surfaces a nonsensical "Provider
'moa' is set in config.yaml but no API key was found. Set the MOA_API_KEY
environment variable..." error for a provider that was never meant to be
reached over the wire.
Fix mirrors #53827's aggregator-resolution approach exactly: when either the
explicit `provider` arg or the config-derived `cfg_provider` is "moa",
resolve the named (or default) MoA preset via resolve_moa_preset() and
continue with its aggregator's real provider+model, dropping any explicit
base_url/api_key (the moa:// virtual endpoint and placeholder key belong to
the facade, not the aggregator's real provider). If the preset can't be
resolved (renamed/deleted), degrades gracefully to the pre-fix behavior
instead of raising harder.
- agent/auxiliary_client.py: _unwrap_moa_provider() helper + call sites for
both the explicit-arg and config-derived provider="moa" cases in
_resolve_task_provider_model(). Also tightened base_url/api_key parameter
types to Optional[str] (matching their actual None-accepting behavior),
which incidentally resolved 5 pre-existing ty diagnostics at call sites.
- 5 new regression tests in tests/agent/test_auxiliary_client.py: explicit
arg unwrap, config-derived unwrap, default-preset fallback when no model
is configured, graceful degradation on preset-resolution failure, and a
non-moa regression guard.
Six spawn sites reachable from the desktop GUI / TUI gateway lacked CREATE_NO_WINDOW, so a windowless parent (pythonw/Electron) flashed a conhost per spawn: cli.exec RPC, quick-commands exec dispatch, and shell.exec RPC in tui_gateway/server.py; the CLI REPL quick-commands exec in cli.py; and the per-session provider transports in agent/copilot_acp_client.py and agent/transports/codex_app_server.py (Popen, hide-only so PIPE stdio stays intact).
All use hermes_cli._subprocess_compat.windows_hide_flags() (no-op on POSIX), matching the pattern already used at three other sites in tui_gateway/server.py. Deliberately hide-only — no detach flags, no Electron changes (per the #54220 revert history).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Review follow-up: the auth path called pool.try_refresh_current() before
the hinted rotation, so a stale current() pointer could force-refresh a
different, healthy entry — consuming its single-use refresh token, or
(for non-OAuth entries, where a forced refresh marks the entry exhausted
outright) killing it entirely before api_key_hint was ever consulted.
Use try_refresh_matching(api_key_hint=...) to resolve and refresh the
entry that supplied the failing key under the pool lock, falling back to
the previous behavior when no key is known.
Adds a regression test with current() deliberately pointed at the
healthy entry: on the old code the healthy entry is exhausted by the
forced refresh and the pool ends up fully offline; with the fix the
failing entry is exhausted and recovery rotates to the healthy one.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
recover_with_credential_pool identified "which credential failed" via
pool.current(), a shared mutable pointer that is advanced by every
select() (round-robin rotation, concurrent turns, and other processes
reloading the pool reset it to None). By the time recovery ran, it
routinely pointed at a different, healthy entry — mark_exhausted_and_rotate
then stamped the failing request's error message and reset time onto that
innocent entry. With round_robin and one hard-capped key this
deterministically exhausted the healthy key too and took the entire pool
offline ("no available entries") from a single rate-limited credential.
mark_exhausted_and_rotate already supports api_key_hint for exactly this
(the auxiliary-client path passes it); the main conversation-loop path
never did. Pass agent.api_key — kept in sync with the entry in use by
_swap_credential — as the hint on all four rotation call sites, and make
the "already exhausted → rotate immediately" pre-check look up the failing
entry by key with the same fallback to current().
Adds regression tests that fail on the old attribution logic: a fresh
pool (current() is None) failing on key B must mark entry B, never
entry A.
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Follow-up for the salvaged #35191: the mid-turn pre-API pressure emit in
conversation_loop.py and the idle-resume emit in turn_context.py were not
routed through automatic_compaction_status_message, so an engine with
emit_automatic_compaction_status=False still leaked those lines. Both now
resolve through the hook (phases "pre_api" and "idle") while keeping the
#69550 template constants as the default wording. Suppression also skips the
#69546 structured 'compacted' terminal edge for compress-phase events that
opened no visible phase; failure warnings (_emit_warning) remain never
suppressible, pinned by test.
Follow-up hardening on the salvaged merge-into-trailing-turn fix:
- Merge only into REAL user tails (_is_real_user_message probe). Merging
into scaffolding tails (continuation marker, summary-as-user handoff)
would upgrade them to real-user evidence after SessionDB projection
strips the flags, breaking zero-user provenance (#69292 -
_is_synthetic_compression_user_turn keys on the TODO_INJECTION_HEADER
content marker, which merge-at-tail would bury mid-content).
- Strip a previously merged snapshot block before re-injection so
repeated boundaries refresh rather than accumulate todo state, and
refresh a bare stale snapshot row in place instead of stacking a
duplicate (empty/stale-skip semantics from #26981 by @YLChen-007).
- Scaffolding tails keep the flagged standalone append (pre-#53890
status quo; adjacent user rows are repaired downstream by
repair_message_sequence / _merge_consecutive_roles).
After context compression, the preserved todo list was unconditionally
appended as a standalone user message. When the compressed transcript
already ends with a user message (common case), this creates consecutive
user/user turns — a role-alternation violation some providers reject.
Fix: fold the snapshot into the trailing user message (blank-line separated)
when one exists with plain-string content. Falls back to append when the
tail is non-user, empty, or has structured (list) content.
Rebased on current upstream/main.
Closes#53890
After multiple in-place compactions, short tool-heavy sessions can leave
nearly every remaining message inside protect_last_n while those messages
are huge completed file/tool outputs. The middle compress window then
makes no material token progress and the turn dies with
"Cannot compress further" (#61932).
Cap the prune message floor at the same bound as tail-cut, and under
pressure demote bulky protected-tail tool bodies (keeping a short recent
floor) so preflight can reclaim headroom without wiping the active ask.
Follow-up for salvaged #58629: thread the previous flush baseline through
the idle-compaction caller in turn_context.py (the one caller the original
PR predates), and add regression coverage that every early-return path in
compress_context (breaker skip, no-progress, plus the completed-rotation
boundary) resets the per-attempt in-place outcome so a stale
_last_compaction_in_place from an earlier successful in-place compaction
can never baseline unflushed turns as persisted.
Relocates the #20424 wiring: the preflight region moved out of
run_agent.py into agent/turn_context.py (and through the
compression.max_attempts unification, #69315), so the contributor's
elif branch is reapplied at its current home as the else arm of the
threshold dispatch chain.
Integration contracts:
- Byte-identical default: the built-in ContextCompressor inherits
ContextEngine.should_compress_preflight() -> False, so the default
path performs no compression and touches no turn bookkeeping
(pinned by test_builtin_compressor_default_sub_threshold_path_unchanged).
- Attempt-cap: the engine gets exactly ONE compress() pass per turn,
mutually exclusive with the cap-bounded threshold multi-pass loop,
so turn-start passes stay within the resolved
compression.max_attempts budget in every case.
- No-op blocking (#64382 / 377244f7c): an engine pass that no-ops
(_compress_context returns the input list object) neither sets nor
clears preflight_compression_blocked and does not re-baseline the
flush history — a sub-threshold maintenance no-op proves nothing
about over-threshold compressibility.
- Engine exceptions are swallowed at debug level; cooldown/defer/
codex-native gates run before the hook is ever consulted.
Salvaged from #20424 by @Beandon13. Fixes#20316.
686f6c61
Gille
ethernet
Teknium
joaomarcos
Fangliquan
kshitijk4poor
TRON
carlotestor
Rod Boev
BB-light
spfcraze
JonthanaHanh
Doud-FR
x7peeps
Xipong
wz-heng
Israel Lot
brooklyn!
MaxFreedomPollard
Zeraphim
praneshnikhar
dstkwll
Kyle
knocksen
AKAZIK-py
Alex Fournier
rob-maron
Michael Jordan
Matt Ezell
Jasmine Naderi
Gabriel Stoltemberg
Soju06
Ben Barclay
Francesco Bonacci
victor-kyriazakos
zapabob
StellarisW
DonutsDelivery
Atakan
Siddharth Balyan
LunarNexus
PRATHAMESH75
Ivan Miao
kshitijk4poor
Koho Zheng
HexLab98
Aaron Weiker
Jeffrey Quesnelle
0xprincess
Jaaneek
kshitijk4poor
fangliquanflq
elcocoel
xxxigm
Tony Simons
Oleksii Kalentiev
embwl0x
Jeffrey Quesnelle
dsad
MorAlekss
webtecnica
kudi88
Victor Kyriazakos
AlexFucuson9
CoreyNoDream
Drexuxux
konsisumer
Anthony Ruiz
Jash Lee
Dhruv Raajeev
ygd58
akb4q
Stoltemberg
kshitij
flyingdoubleg
Alan Harman-Box
xue xinglong
srojk34
liuhao1024
robbyczgw-cla
sgtworkman
oppenheimor
SquabbyZ
hellofrommorgan
dyreckt
kosta
Dineth Hettiarachchi
Idris Almalki
Jerry
Jonathan Boisson
Cluster2
Kolektori
izumi0uu
root
Rain
aui
Janig88
Prathamesh Chaudhari
elphamale
annguyenNous
wjq990112
panding
solyanviktor-star
Blade
李航
Stanislav
happy5318
Ethan
trkim
Basil Al Shukaili
schattenan
Stephen Schoettler
Yingliang Zhang
giggling-ginger
Brett Bonner