Run the expensive-model warning for explicit startup `-m` / `--provider`
overrides before the chat loop starts, and fail closed for non-interactive
invocations that select an expensive or known-confusing model.
Also classify Nous paid-model 404s that say credits are required as billing
exhaustion so they fail fast with billing guidance.
Tested:
- scripts/run_tests.sh tests/hermes_cli/test_cli_startup_model_cost_guard.py tests/hermes_cli/test_model_cost_guard.py tests/agent/test_error_classifier.py -- --tb=short -q
Custom providers (custom:xxx) serve their own pricing; models.dev stores
OpenRouter prices for the same model ids. The cost guard fired on that
foreign pricing and blocked composer/CLI model switches on custom
providers with a wildly wrong warning (#54348).
expensive_model_warning now only trusts model_info/models.dev pricing
when the provider maps to a models.dev provider and the info's
provider_id matches, and only consults the pricing-entry lookup when
the billing route is known. Salvaged from #54422; the PR's desktop-hook
half predates the use-model-controls rewrite and is superseded by the
hook's existing rollback handling.
The command path (/model <name> --provider <p>) called
_confirm_expensive_model_switch() inline. That modal blocks its calling
thread on a response queue (see _prompt_text_input_modal); on the
prompt_toolkit main thread the TUI event loop freezes, the modal never
renders, and the switch silently cancels after the 120s timeout — the
user sees a frozen terminal and 'Model switch cancelled.' without ever
seeing the warning. The picker path already dispatched confirm+apply on
a worker thread; the command path now mirrors that contract.
Extract the inline confirm+apply block into
_confirm_and_apply_cli_model_switch() (preserving --once restore and
persist semantics) and dispatch it on a daemon thread when a TUI app is
present, keeping the synchronous path for non-interactive/test use.
Tests: new test_model_switch_confirm_thread.py pins (a) confirm runs off
the main thread when _app is present, and (b) the no-app path stays
synchronous. Existing _StubCLI helpers forward to the extracted method.
Plain type=completion events built in _run_process_watcher carried only
session_key (chat/thread routing) with no spawning-session stamp, so after
/new (or a session switch) a completion notification from the OLD session
was injected into the chat's NEW session. Main already solved this exact
class for async delegations via the _classify_completion_target pre-flight
(_USER_BOUNDARY_END_REASONS drop on user-closed sessions, deliver on
idle-ends, follow the compression-tip chain), but the gate only ran for
type=async_delegation events.
Kernel salvage of #16455:
- Stamp the spawning conversation's session-db id (HERMES_SESSION_ID via
session-scoped env) on the ProcessSession and the pending_watchers entry
at spawn time in tools/terminal_tool.py; persist it through the process
registry checkpoint/restore so recovered watchers keep the stamp.
- Thread the stamp into the completion_evt built by _run_process_watcher
(watcher entry first, ProcessSession fallback for recovered watchers).
- In _deliver_completion_notification, run the SAME pre-flight classifier
for stamped type=completion events: terminal -> drop with a log (output
stays available via process(action='log')), retry -> False so the
watcher re-polls, deliver -> proceed. The policy has exactly one owner
(_classify_completion_target); nothing is forked. Unstamped legacy
events keep today's deliver-always behavior, and the async-delegation
path is untouched.
Based on the session-boundary approach from #16455 by @Tosko4 (original PR
was over-scoped across adapters/slash-commands/cron; this lands the kernel
only).
Tests: completion from a /new-closed session is dropped; completion after
an idle-end still delivers; unstamped legacy event delivers; retry verdict
returns retryable False without adapter injection; async_delegation gate
unchanged; stamp survives checkpoint recovery.
The bare dimmed glyph had nothing to read against. Over a light document in
a light theme it is a pale mark on white, and every rest opacity tried
(0.35, then 0.45, then 0.75 behind a text halo) came back reported as the
button being gone.
Give the control its own substrate, which is what every shipped overlay
does: Apple's HIG puts controls on a material rather than directly on
content, Firefox picture-in-picture draws close/unpip as opaque chips, and
Discord's overlay adds a contrast layer over the game. Deriving contrast
from the backdrop is not available to us either way -- mix-blend-difference
composites against the page, and behind a transparent Electron window that
is nothing.
The chip now wears the composer bar's own tokens (fill, hairline, radius,
bottom shadow), so it inverts with the theme and with the OS appearance
under mode 'system'. It rests hidden and fades in while the bar, the band,
or the chip itself is hovered, with a hold on the way out so it survives the
reach across the gap -- reaching for the HUD is the motion that means "I
want the app". Hover rather than focus: the caret gate behind #81893 broke
the escape hatch exactly when it was needed.
The async-delegation watcher drained the completion queue as a batch but
then delivered each event as its own synthetic turn, flooding the session
when a fan-out of background subagents finished together. Builds on the
per-process completion batching salvaged from PR #71898 (thanks
@yuzilongleif-collab) which coalesces concurrent _run_process_watcher
completions behind a short per-route fan-in window.
This commit adds the async-delegation half: group the drained batch by
full routing key (session_key + parent_session_id + platform/chat/thread/
user) and inject ONE consolidated turn per group. Durable-ack handling
stays honest: sibling rows are claimed up front via claim_event_delivery;
rows another consumer owns are excluded from the consolidated text (no
double-delivery); sibling claims are acknowledged only after adapter
acceptance and released (still pending) on failure. Events for different
sessions never coalesce, and a single-event group rides the existing
per-event path unchanged (latency and text identical).
Tests: 3 same-tick events -> exactly one adapter.handle_message carrying
all 3 results with all 3 durable rows delivered; 2 sessions -> 2 turns;
single-event path unchanged; failed batch releases claims and retries;
foreign-claimed sibling excluded and left pending.
Async-delegation batch completions and background watch notifications
re-enter the gateway as synthetic MessageEvent(internal=True) turns via
_inject_watch_notification, but were persisted as bare role='user' rows —
indistinguishable from real user input in transcripts and the desktop UI.
Thread the event's internal flag through to persistence: when
event.internal is set, the turn's persisted user row is stamped
display_kind='internal_notification' (the existing DB-only presentation
sidecar used by auto_continue / model_switch rows). Wired through
_run_agent → _run_agent_inner → TurnContext → run_conversation's
persist_user_display_kind, and onto the three gateway-side fallback user
rows (transient failure, no-new-messages, pre-run crash), whose
append_to_transcript writer now forwards display_kind/display_metadata
to SessionDB.append_message.
Invariants preserved: role stays 'user' (alternation untouched), no new
injections, no past-context mutation, and display_kind is already popped
from every provider-bound copy in conversation_loop, so replayed sessions
never leak the marker to the API.
Regression tests: internal turn marked, real user turn unmarked, fallback
rows marked/unmarked per event, and a DB round-trip proving replay keeps
role/content intact while the provider copy drops the marker.
The terminal redactor is called without force=True on all three
user-facing sends in _run_process_watcher, so process output reaches the
platform raw when security.redact_secrets is disabled. The agent-notify
path was already covered; this covers the two direct adapter.send()
paths the sweeper identified.
Review: teknium1 (#73547)
_output passes through redact_terminal_output (force=False) but
not _redact_gateway_user_facing_secrets (force=True). When
security.redact_secrets is disabled or the command is not an
env-dump, the output can leak credentials (Authorization: Bearer,
PGPASSWORD=, etc.) into the session transcript and chat platforms.
Add the same redaction gate applied to _command so both paths
are protected by force=True + _GATEWAY_SECRET_PATTERNS.
Add TestNotificationRedaction class with two tests:
1. test_completion_notification_redacts_secret — verifies _move_to_finished
redacts API keys in completion notifications before enqueueing
2. test_watch_match_notification_redacts_secret — verifies _check_watch_patterns
redacts secrets in watch_match notifications before enqueueing
These tests cover the gap identified in #43025 where the explicit process
tool path (poll/log/wait) was redacted but the automatic notification
delivery path was not.
Apply _redact_process_result() to completion and watch_match
notifications before enqueuing them in the completion_queue.
Previously, the explicit process tool path (poll/log/wait) applied
redact_terminal_output() via _redact_process_result(), but the
automatic notification delivery path (notify_on_complete, watch_patterns)
only applied strip_ansi(). This meant API keys, tokens, and other
secrets from background process output were injected into the LLM
conversation unmasked.
The fix ensures both code paths apply the same redaction, matching
the foreground terminal tool behavior.
format_process_notification had no case for watch_overflow_tripped /
watch_overflow_released, so a watch-pattern notification flood surfaced
as '[IMPORTANT: Background process exited (exit code ?)]' — a phantom
exit notification for a process that never existed — while the actual
'watch flood, N notifications suppressed' summary in the event's
message field was silently dropped. The gateway delivery path was
worse: _drain_gateway_watch_events retained only watch_match and
watch_disabled, discarding overflow events entirely before formatting.
Route both event types through the message field in the shared
formatter and the gateway formatter, and retain them in the gateway
drain.
Adds hermes_cli/model_selection_guards.py: a single evaluation point that
runs every selection guard (cost + the new data-policy guard) and returns
the warnings that fired. All seven model-selection surfaces (CLI picker,
cli.py TUI modal, gateway typed /model, dashboard web_server, TUI gateway,
Telegram and Discord pickers) now call the registry instead of importing
model_cost_guard directly — so the data-training-tier warning from
PR #81416 fires everywhere at once, and future guards need zero surface
wiring.
Guard modules keep their public APIs; existing mock patch points
(hermes_cli.model_cost_guard.expensive_model_warning) remain valid.
muse-spark-1.2-contributor is heavily discounted BECAUSE Meta uses your
prompts and completions to train future models. Selecting it for the price
without realising the data trade-off is a footgun.
Add hermes_cli/model_data_policy_guard.py (mirrors model_cost_guard):
data_training_warning(model_id, provider, base_url) -> DataTrainingWarning|None,
driven by a vendor-agnostic rule table. The status is not machine-readable on
/v1/models or models.dev, so the v1 rule keys on the documented '-contributor'
model id (fires regardless of provider, so it also covers custom/gateway
routes). Message mirrors Meta's pricing-doc language and figures
(https://dev.meta.ai/docs/pricing-rate-limits/).
Wire it into the CLI model picker's confirm flow (auth.py) as a [y/N]
disclosure, chained after the expensive-model cost guard. Fires only on the
contributor tier; silent on muse-spark-1.1/1.2 and all other models.
The route dict in _persist_model_switch_to_session used filtering
(omits falsy values) while the top-level keys used (writes
explicit None to trigger deletion in _merge_model_config_json). This
asymmetry meant stale keys from a previous /model switch survived in
the nested gateway_runtime dict even after the fix in #85261 that
properly deleted them from the top-level keys.
Fix: build the route dict with and derive the top-level keys
from **route so both shapes always use identical deletion semantics.
Also filter None values in session_gateway_runtime's reader since
gateway_runtime is replaced as a whole dict (not deep-merged), so None
values written by the persist path survive in the nested dict.
Found by /simplify-code 3-reviewer review on #85261 (all 3 reviewers
converged on the route dict asymmetry as the verdict-relevant finding).
`_reap_unsupervised_gateway_orphans()` short-circuits on Linux hosts
with systemd via `supports_systemd_services()`, but returns `False` on
macOS — there is no systemd. This means the orphan reaper runs
unconditionally on macOS and treats the launchd-managed gateway as an
unsupervised orphan, SIGTERM-ing it.
When Hermes Desktop opens, `hermes serve` calls this function during
startup (web_server.py line ~251, gated on `HERMES_DESKTOP == "1"`).
The launchd gateway is killed, launchd restarts it via `KeepAlive: true`,
and the user sees a spurious gateway restart every time they reopen the
Desktop app.
Fix: exclude PIDs managed by launchd (`_get_service_pids()` already
returns launchd-managed PIDs on macOS) from the orphan scan, the same
way systemd PIDs are excluded on Linux.
Tested on macOS 26.5 (Tahoe) with Hermes Desktop 0.16+ and a
launchd-managed gateway (`ai.hermes.gateway` plist with `KeepAlive`).
Before the fix, quitting and reopening Hermes Desktop restarted the
gateway every time. After the fix, the gateway stays running across
Desktop quit/reopen cycles.
Two tests in TestReapUnsupervisedGatewayOrphansMacOS:
- test_macos_excludes_launchd_pid_from_kill: verifies a launchd-managed
PID is not SIGTERM'd while a real orphan is
- test_macos_no_orphans_when_only_launchd_gateway_running: verifies the
reaper returns False when the only gateway PID is launchd-managed
Both tests patch is_macos() to True and supports_systemd_services() to
False to simulate the macOS code path where the short-circuit does not
fire.
resolveAgentAvatar cached null permanently (window lifetime), so a bot
whose avatar reached the asset store moments after its first notice
rendered — freshly created bots, art backfills in flight — kept the 🤖
glyph until an app restart even though profiles.get_asset had the pfp
(user report: brand-new bots' notices never picked up their faces).
Hits stay cached for the window; misses re-probe after 30s. Same
dedupe/inflight behavior otherwise.
profiles.list's last_session.preview reused list_sessions_rich's
shared preview, which is the session's FIRST user message — right for
session lists (recognition), wrong for a messaging-style roster where
the line under each agent should track the conversation ('Hey, tell
me about yourself!' forever, per user report). Override with the
newest active user/assistant text (same query shape and lock
discipline as SessionDB.latest_message_row_id); best-effort, falls
back to the first-message preview on any failure.
E2E: live profiles.list now shows each bot's latest exchange.
Background process completions on messaging platforms now default to a
one-line status message (✅/❌ + command + duration; failures append a
short output tail) instead of dumping the raw output buffer into the
chat. New display.background_process_notifications mode 'concise' is
the default; 'all' keeps the old raw-dump behavior for anyone who wants
it. Config migration v35 moves users still on the old implicit default
'all' to 'concise' on their next update; explicit result/error/off
choices are preserved.
* fix(desktop): load the intact renderer bundle when an update tears one copy
index.html and the hashed chunks it names are one generation. A packaged app
ships that bundle twice (inside app.asar and, via asarUnpack, beside it in
app.asar.unpacked), so an update that replaces the app while its files are
locked can leave the two copies from different generations. resolveRendererIndex
took the first index.html that existed, so it could pick the torn one and the
window died on its first lazy import with "Failed to fetch dynamically imported
module" -- with no way out, because every relaunch reloaded the same copy.
Check each candidate's declared modules and prefer a complete generation; when
both are torn, log which files are missing and how to repair instead of leaving
the crash unexplained.
* fix(cli): rebuild the desktop app when its renderer bundle is half-replaced
The content stamp hashes the SOURCE tree, which an interrupted update leaves
intact, so `hermes desktop` reported "up to date" and skipped the rebuild that
would repair a torn bundle -- the app relaunched into the same crash and
reinstalling looked like the only option.
Treat a bundle whose index.html names missing chunks as stale regardless of the
stamp, and say so on the way into the rebuild.
The sending bot's chat showed inter-agent deliveries as raw terminal
tool rows (the hermes -p … chat … command + output transcript) —
plumbing, not conversation. When a terminal call matches the delivery
convention (-p <agent> chat … -q "Message from …", the shape from
#85855), it now renders as compact centered notices instead:
'Messaging <agent>…' while running, 'Messaged <agent>' on completion,
and — when the quiet run returns the recipient's reply — a 'Message
from <agent>' notice with the text behind a 'show message' expander.
Avatar resolution reuses the #85855 helper (now exported); glyph
fallback everywhere it can't resolve. Failed commands keep the real
terminal row (debuggable). Ordinary terminal calls untouched.
Sender + receiver now speak one visual language: the exchange is a
pair of timeline events on both ends (Grok-bots parity), with #85884
collapsing the receiving side's reply.
agent-delivery tests 6/6; thread suite 20 files / 117 tests green.
The recipient's reply to an inter-agent delivery rendered as a full
assistant message, so the receiving bot's chat read like a normal
human conversation. The exchange is an EVENT in that bot's timeline,
not conversation content: when the immediately preceding user message
is an inter-agent delivery (AGENT_MESSAGE_RE, shipped in #85855), the
reply now renders as a compact centered 'Replied to <sender>' notice
with the full text behind a 'show reply' expander — mirroring the
delivery notice above it. Never collapses while streaming (progress
stays visible); ordinary assistant messages untouched.
Thread suite 19 files / 111 tests green.
The desktop docs had no HUD mode section at all. Add one covering the
key interaction (long-press the composer to drag the bar), plus resize,
snap-to-pointer, and exit — all sourced from the actual implementation
in composer-drag.ts, resize-handle.ts, and keybinds/actions.ts.
A provider-confirmed rearm (#85846) resets the shared attempt budget, but
an earlier insufficient-progress verdict left _preflight_compression_blocked
armed, keeping the pre-API gate dark for the rest of the turn — a later
pressure spike could still grow unchecked until the provider overflow
handler fired. Clear the blocker and the stale pressure reading inside the
provider-confirmed rearm branch: the prompt is proven back below the
threshold, so the old verdict describes a request shape that no longer
exists.
Builds on @h-mascot's #84995, whose commit is preserved on this branch;
his rearm condition was superseded by #85846's latch-verified variant, but
the blocker-clear half was correct and is kept.
* fix(cli): --in accepts Git Bash / MSYS paths on Windows
Under Git Bash, 'hermes chat --in ~' reaches the CLI as /c/Users/<user>
(the shell expands ~ to an MSYS POSIX path; MSYS2 argument conversion
is disabled for native executables), and the isdir check failed with
'--in directory not found: /c/Users/...'. Route the value through the
existing _msys_to_windows_path translator (MSYS + Cygwin + WSL drive
spellings; no-op elsewhere) before expanduser/abspath.
Hit live: Bot Mode's agent-messaging protocol delivers with --in ~, so
every bot-to-bot send from a Git Bash-driven agent failed on Windows.
Tests pin both the translation cases and (source-level) the call site
actually using it.
* test: assert the MSYS translation, not platform abspath
The prior assertion ran os.path.abspath on the translated Windows path,
which on the Linux CI runner (posixpath) treats 'C:\Users\alice' as
relative and prepends the runner cwd. Pin the translation output and
ntpath absoluteness instead — same contract, platform-independent.
Add data-sessions-mode ('flat' | 'projects' | 'project' | 'archived' |
'search') and data-sessions-project (entered project id) to the sidebar
sessions wrapper so custom UIs can target project mode without relying
on internal class names.
Windows contributors' tools default to CRLF; without repo-level
normalization an edit becomes a whole-file phantom diff (583-line
'change' observed today from one 2-line edit), string-match patch
tooling breaks on invisible \r, and review is polluted. Extend the
existing LF rules (shell/Docker) to every source/text extension:
normalize at check-in AND check out as LF so working trees match the
index on all platforms. *.ps1 stays CRLF (PowerShell 5.1 tooling).
git add --renormalize: exactly one tracked file had mixed endings
(tests/tools/test_windows_agent_loop_papercuts.py) — normalized here,
so no phantom diffs land on anyone's next commit.
* feat(desktop): render agent-to-agent messages as attributed cards, not user bubbles
Bot-to-bot deliveries arrive on the user role (alternation requires
it) but are not the human speaking — they rendered as if the user
typed them. Detect the delivery prefix ('Message from 🤖 <sender>: …',
emoji-less, and the legacy bracket form) and render an attributed
inter-agent card: left-aligned, robot + sender header, 'agent message'
label, body through the same minimal markdown pipeline. Anchored regex
cannot fire mid-prose. Same pattern as ProcessNotificationNote.
Presentational only; content/roles/caching untouched.
* reshape: inter-agent card -> Grok-style compact timeline notice
Per maintainer screenshot: the delivery renders as a subtle centered
'🤖 Message from <sender>' notice (ProcessNotificationNote's shape),
with the delivered text behind a 'show message' expander instead of a
full-width card. The recipient's reply remains a normal assistant
message below it.
* feat: sender avatar on the inter-agent notice
The delivery prefix may carry the sender's profile handle —
'Message from 🤖 <Display Name> (@<handle>): …'. The notice resolves
it through profiles.list (has_avatar) + profiles.get_asset and renders
the sender's actual avatar in place of the 🤖 glyph, with module-level
memoization (one resolution per sender per window) and inflight
de-dup. Glyph fallback covers handle-less prefixes, older gateways
without profiles.*, avatarless profiles, and failures. 'hermes'
resolves to the primary profile by convention. Tests 6/6.
* lint: sort the $gateway import (perfectionist/sort-imports)
A marathon tool turn burned all compression_attempts on *successful*
pre-API compactions; the gate then went permanently dark and the context
grew unchecked until the provider rejected the request terminally
("Context length exceeded: max compression attempts (3) reached", session
f087963205f9, 2026-08-01). The budget now refunds at loop top when the
assembled request is back under threshold * 0.8 AND the compressor's own
should_compress() agrees the pressure is gone.
Anti-thrash intent of the cap is preserved (#11529): no-progress passes
never reach the refund margin, divergent-signal cases (should_compress
still True) keep the budget burnt, and the insufficient-progress blocker
is untouched.
Validation: new behavioral suite (7) + all 130 compression/context tests
via scripts/run_tests_hermetic.py.
Rückbau: Commit revertieren; kein Zustand, keine Migration.
(cherry picked from commit 041b489d56)
Picking a colon-less completion row (agent profile mentions from
complete.path, e.g. '@mr-tester') fell through serialize()'s typed-
reference branch: classify() marks colon-less entries type 'simple'
with insertId = text, so the serializer minted '@simple:`@mr-tester`'
— which rendered as a weird chip in the composer AND broke downstream
@mention routing (the backtick-quoted form no longer parses as a bare
mention). Colon-less rawText now inserts verbatim, matching what @diff
and @staged already did via the empty-insertId path.
Test: mention rows serialize to plain @name and commit to the editor
without an @simple: wrapper (directive-label 5/5).
Closing a pane contributed by a plugin used to disable the entire plugin,
unloading every one of its contributions. For a plugin that owns several
independent panes (e.g. Bot Mode's Cronjobs pane alongside its Bots roster
and composer middleware), closing one pane silently killed the rest.
Now: closing one pane of a multi-pane plugin dismisses only that pane; the
plugin stays enabled and its other panes/commands/middleware keep working.
Reset layout restores dismissed contributed panes. A single-pane plugin
keeps the existing symmetric behavior (Close disables the plugin, with
Settings -> Plugins as the recovery path).
Adds regression coverage for both cases.
main's check:lint (tsc) is red: 071d27d1c3 added a required update()
member to ComposerActionsScope, but the use-composer-actions.test.ts
scope double was never extended. TS2741 at line 294.
Covers the advisory memory and disk blocks added to GET /api/status in
#84965 (thresholds, staleness handling, fail-safe degradation) and the
dashboard resource-pressure banner (trigger precedence, boot-scoped
dismissals).
Stop freezing the xAI/xAI-OAuth catalog at import so /model and setup
pick up new Grok IDs after the models.dev cache refreshes. Put xai and
xai-oauth on the shared picker-time models.dev merge path and pin
grok-4.6 as the default headline model.
lkz-de
Teknium
Dustin Persek
Turgut Kural
dsad
Brooklyn Nicholson
yuzilongleif-collab
yuzilongleif-collab
handnewb
handnewb
spfcraze
Lidang-Jiang
fangliquanflq
Beto de Paola
kshitij
Shedrack Eze
hermes-seaeye[bot]
Henry Mascot
Gille
ai-ag2026
unsupportedpastels
Jaaneek