Add db_connections_open (gauge) and db_connections_established (counter), driven by SQLAlchemy connection-lifecycle events so they report real connections under every pool class — including NullPool, where the pool-object collector (db_pool_connections) reads zero. Under NullPool the establishment rate approximates request rate.
DBConnectionTracker mirrors DBQueryInflightTracker: a ConnectionRecord.info marker makes each physical connection increment once and decrement at most once (no leak, no negative). Registered per-process in the API and deriver; zero-init via the pre-resolved labeled children, matching db_queries_in_flight_gauge.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
* telemetry: materialize dropped-event counter children at 0
A labeled Prometheus counter exports no series until its first labels()
call, so telemetry_events_dropped stayed invisible until an event was
actually dropped — impossible to alert on or graph, and "no drops" was
indistinguishable from "metric missing / scrape broken".
Pre-create the (namespace, reason) children at 0 on emitter start, for
each reason the emitter can emit, so the metric is always present.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* telemetry: generalize counter zero-init to all bounded-label counters
Extends #927 (which zero-inited telemetry_events_dropped) to every counter
whose label domain is bounded and known at startup, so metrics are present in
Prometheus before their first event — a missing series then signals a broken
scrape rather than "nothing happened yet".
- add initialize_bounded_metrics(instance_type) on PrometheusMetrics; call it
per-process from main.py (api) and deriver/__main__.py (deriver).
- extract a shared _touch() helper; refactor initialize_telemetry_dropped_metrics
onto it (that one stays per-emitter in start() — it's prefix-dependent).
- explicit ALL_EVENT_TYPES / HIGH_VOLUME_EVENT_TYPES registry in telemetry.events,
drift-guarded by tests that walk BaseEvent subclasses.
- only VALID (task_type, token_type, component) tuples for deriver_tokens (the
cartesian product would fabricate impossible always-0 series); only high-volume
event types for sampled_out; high-cardinality labels (endpoint, workspace_name)
left open.
- gauges: zero-init embed_now_tasks_in_flight + telemetry_buffer_size; add a new
message_embeddings_pending backlog gauge, set each reconciliation cycle and
zero-inited at deriver startup (Rajat's pending/in-flight ask).
- backfills the tests #927 shipped without.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* review: task-aware deriver combos + fail-soft gauge zero-init
I1: _DERIVER_TOKEN_COMBOS was factored task-independently, materializing the
impossible (ingestion, input, previous_summary) series — previous_summary is
summary-only. Make combos task-aware (_DERIVER_TOKEN_COMBOS_BY_TASK) so no
always-0 impossible series is fabricated, matching the PR's own goal. Tests
tightened to assert the ingestion/previous_summary series is absent.
I2: the three gauge .set(0) zero-inits were bare while the counter inits go
through the fail-soft _touch. Add _set_gauge_zero() so a gauge init can't
propagate an exception into process startup either.
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* test(telemetry): isolate zero-init namespaces, add deriver-to-api guard
Global-REGISTRY assertions used a fixed "test" namespace, which several
other suites also pin, so another test's materialized children could
satisfy a presence assertion or break an absence one. Each test now runs
under a unique namespace resolved from settings at read time.
Adds the inverse per-process isolation test: deriver-only init must not
materialize API-only series (dialectic tokens, embed_now).
Co-Authored-By: Claude <noreply@anthropic.com>
* review: per-replica backlog gauge, drop duplicated constants and .meta refs
Addresses Vineeth's review on #927.
Blocking:
- message_embeddings_pending is a DB-global count, so drive it from
ReconcilerScheduler._scheduler_loop (runs on every replica, every
interval) instead of run_vector_reconciliation_cycle (runs off the
queue behind work-unit dedup, so one replica per cycle). Combined with
the zero-init, the old placement made every replica that never won the
work unit export a confident permanent 0. Help string now names the
owner so dashboards don't reach for sum().
- guard initialize_telemetry_dropped_metrics on METRICS.ENABLED,
matching its sibling initializer.
- drop the duplicate REASONING_LEVELS; import the one in src/config.
Non-blocking:
- walk BaseSpecialist recursively via a shared utils.types.walk_subclasses
(replaces the direct-children-only __subclasses__() and the test's
private copy of the same helper).
- derive the specialist assertion from the subclasses instead of
hardcoding two names — the hardcoded pair kept passing after
CardRefreshSpecialist landed, leaving it uncovered.
- inline the zero-init rationale and the multi-instance bucket taxonomy;
removes both pointers to a .meta design doc that is not in the repo.
Tests: new tests/reconciler/test_pending_backlog_gauge.py pins both
halves of the relocation (verified it fails when reverted).
* review: fix inert test guard, stale comments, and the REASONING_LEVELS drift claim
Second review pass on the branch. Findings, most severe first:
- tests/reconciler/test_pending_backlog_gauge.py: the _try_enqueue_task stub
was patched onto the class but declared without `self`, so calling it
raised TypeError — which _scheduler_loop swallows. The guard was inert and
the test passed for the wrong reason. Fixed the arity.
- metrics.py still commented that the backlog gauge is "set live each
reconciliation cycle". That is the exact claim the previous commit
overturned; it now contradicted the help string, the bucket-3 docstring
and sync_vectors.py.
- metrics.py claimed REASONING_LEVELS is "derived from the config Literal so
it never drifts", but config.py hand-listed it, so the earlier dedup had
quietly traded away the guarantee the original get_args() call provided.
Made it true instead: config.REASONING_LEVELS = list(get_args(...)), which
keeps the dedup and restores the invariant.
- dropped _set_gauge_zero: all three gauges it zeroed already have identical
fail-soft setters, so it was a second way to do one thing. Using the
setters also makes _handle_metric_error name the actual gauge.
- record_pending_embeddings_backlog's docstring oversold the covering index
as making the COUNT "negligible". The index makes cost proportional to the
pending backlog, not to the table — which is worst precisely when the
backlog matters. Stated honestly.
- _scheduler_loop's docstring said it only enqueues; it also refreshes the
gauge, at a cadence set by the shortest task interval.
- comment reconciliation: stripped #927 / "the generalization" temporal
anchoring, a CardRefreshSpecialist change-narration clause, and
reviewer-directed phrasing from the test file; disambiguated the
src/utils/summarizer.py path.
- CLAUDE.md had no Prometheus section at all, so the new "add a BaseEvent
subclass -> update ALL_EVENT_TYPES" obligation and the never-sum() rule
for non-additive gauges were undiscoverable from the architecture doc.
Verified: ruff + basedpyright clean (0 errors), tests/telemetry + reconciler
+ dialectic + llm 497 passed, full suite 1768 passed with only the 4
pre-existing test_document failures (OpenAI key required, reproduced on
clean origin/main). Re-confirmed the relocation guard fails when reverted.
* fix: silence the two basedpyright warnings inherited from main
CI runs `uv run basedpyright` bare, and basedpyright exits non-zero on any
warning — so these two have been failing the staticanalysis job on every
branch cut from current main, not just this one:
- src/vector_store/__init__.py:209 implicit string concatenation (#496)
- tests/test_cache_redaction.py:5 private import (#869)
Both predate this branch and are unrelated to the telemetry work; fixed
here only because they block this PR from going green. Verified: clean
origin/main also reports "0 errors, 2 warnings" and exits 1.
basedpyright now 0 errors, 0 warnings, exit 0.
* docs(telemetry): make the bucket-3 aggregation rule precise
The multi-instance taxonomy said a service-scoped non-additive metric has
"no aggregation correct once they disagree", then immediately mandated that
every instance refresh on its own timer. Those undercut each other: staggered
timers ALWAYS disagree slightly, so as written the rule reads as "ensure they
don't", which is unachievable, and it leaves the reader unsure whether max()
and avg() survived the fix.
The actual rule is bounded disagreement plus a scale-preserving aggregator.
Instances are N witnesses to one fact, not N parts of one whole, so sum() can
never be correct (it scales with replica count) while max()/avg()/quantiles
are correct precisely because the per-instance timer bounds the spread.
Wording only; no behavior change. The gauge help string already said
"max() or avg(), never sum()" — this makes the normative docstring agree
with it. Surfaced walking Vineeth's comment 3668208059 for comprehension.
* refactor(bench): import REASONING_LEVELS from config instead of re-listing
Third copy of the constant, missed when ee781c0/694e07f deduped the other
two. This one re-declared the ReasoningLevel Literal as well as the list,
so the type alias could diverge from config's with nothing to catch it —
and the list was hand-written, the variant that typechecks clean while
missing a member.
No import barrier justified it: this module already imports from src, as do
seven of its siblings in tests/bench. Concrete effect of the drift was that
a newly added sixth reasoning level would be rejected by the bench CLI's
argparse choices=.
src.config.REASONING_LEVELS is now the single definition repo-wide.
* test(telemetry): pin the METRICS.ENABLED guard on the per-emitter initializer
initialize_telemetry_dropped_metrics gained a METRICS.ENABLED guard in
ee781c0, addressing Vineeth's asymmetry comment, but nothing asserted it —
it had only the enabled half of the pair its sibling has. Deleting the guard
left the suite green, so the fix closed the asymmetry in the guards and
reproduced it one level up in the tests.
Mirrors test_init_noop_when_metrics_disabled. Verified live rather than
assumed: deleting the two guard lines turns this test red.
Uses a unique namespace, without which the absence assertion would be
satisfied by the enabled test's children rather than by the guard.
* docs(telemetry): fold zero-init why-prose behind # region ai markers
Comment/docstring-only pass over the changed files, per the groudon
comment-marker standard: the terse human-facing "what" stays visible, and
load-bearing "why" (the zero-init / absent-series-means-broken-scrape
rationale, gotchas, receipts) folds into # region ai / # ai: blocks.
Behavior-preserving: AST-identical modulo docstrings/comments vs the
pre-pass merge; ruff, ruff format --check, and basedpyright all clean.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
---------
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
* feat(db): add connection retry, adaptive deriver polling, and pool metrics
Add resilience and visibility for DB connection handling under transaction-
pooler (Supavisor) saturation, where client-connection limits get exhausted
across many tenants.
- get_db/tracked_db now force an eager pool checkout with bounded exponential
backoff (tenacity), retrying SQLAlchemy TimeoutError + OperationalError so
transient pooler rejections degrade gracefully instead of 500ing. Toggle via
DB_CONNECTION_RETRY_ENABLED (+ delay/backoff knobs); ~10s default budget.
- Deriver polling backs off when idle or erroring (base -> max, x2 each cycle)
and snaps back to base on claimed work, cutting steady-state query load.
Toggle via DERIVER_POLLING_BACKOFF_ENABLED (+ max/multiplier).
- Add scrape-time db_pool_connections Prometheus gauge (checked_out/checked_in/
size/overflow, labeled api|deriver), registered in both the API lifespan and
the deriver metrics server.
- Make SqlalchemyIntegration explicit in both Sentry inits; wrap connection
acquisition in a db.pool.acquire span and capture live pool stats on
retry-exhaustion.
* feat(db): add acquisition counter and in-flight query gauge
Build on the pool-connection metrics with two signals that turn detection
into diagnosis under transaction-pooler saturation:
- db_connection_acquisitions{outcome=ok|retried|exhausted}: counts how often
connection checkout retries through pooler rejection — the alertable early
warning before requests start failing.
- db_queries_in_flight: statements actually executing on the wire (via
SQLAlchemy cursor-execute events, drift-proof across query errors). Pairs
with checked_out: the gap reveals connections held but parked (the "idle in
transaction during an external call" antipattern). Labeled namespace +
instance_type only; gated on METRICS.ENABLED for zero overhead when off.
Add DB-free unit tests for retry outcomes, polling backoff, and in-flight
gauge drift handling.
* fix: address CodeRabbit review on PR #758
- db: roll back the session on a retryable checkout failure before
retrying — a failed autobegin can leave it pending-rollback, making the
next db.connection() raise instead of re-checking-out cleanly. Cheap
Python-side cleanup when no connection was bound.
- metrics: guard DBPoolCollector.collect() so a pool-read/import hiccup
can't raise and abort the whole /metrics scrape (Prometheus drops ALL
metrics if any collector raises) — log and fall back to empty.
* fix(db): lazy retrying session + review fixes for connection backoff
Address Codex/CodeRabbit review on PR #758.
- Replace eager checkout with HonchoAsyncSession: a lazy AsyncSession that
checks out its connection (with retry) on the first DB-touching call, not at
construction. Request handlers doing non-DB work (embedding/file/LLM) before
their first query no longer pin a connection across it, while the API path
still gets checkout retry. Only the checkout is retried — the statement runs
once via super(), so writes are never duplicated. Tracing's set_config moves
into the same lazy acquire hook.
- Roll the session back on a retryable checkout failure before retrying, so a
failed autobegin can't leave it pending-rollback.
- Lower default POOL_TIMEOUT to 5s and validate it stays under the retry budget
for pooled (non-null) POOL_CLASS; update config.toml.example and v2/v3 docs.
- Clamp pool overflow gauge to >= 0 (was negative before the pool fills).
- Remove double-sleep in the deriver idle poll (true backoff cap, not 2x);
make in-flight instrumentation registration idempotent.
- Tests: HonchoAsyncSession lazy/idempotent acquire, statement-runs-once,
tracing, commit/rollback flag reset, get_db no-acquire-at-entry, polling-loop
single-sleep, and the POOL_TIMEOUT/retry-budget validator.
* fix(db): cover all DB-touching session methods; clear flag on close/reset
Address Codex follow-up review on PR #758 (polish, no behavior-critical bug).
- HonchoAsyncSession: wrap get/get_one/stream/stream_scalars/delete in addition
to execute/scalar/scalars/flush/merge/refresh/commit, so the "lazy checkout
with retry on first DB use" guarantee has no holes. connection() stays
unwrapped (acquire_connection_with_retry calls it — wrapping would recurse).
- Reset the acquired flag on close()/reset() too, so a session reused after
close/reset re-acquires (and re-wraps retry) on its next DB use.
- Fix stale comments: connection retry now applies lazily to the request path
via HonchoAsyncSession (config.py), and the FakeSession helper note.
- Tests: close/reset flag reset, and get/delete route through acquisition.
* feat(embedding): add dimensions_mode for OpenAI dimensions= forwarding
Add EMBEDDING_MODEL_CONFIG__DIMENSIONS_MODE (auto|always|never) controlling
whether the dimensions= parameter is forwarded on OpenAI embeddings.create
calls. auto (default) sends it when the operator explicitly set
EMBEDDING_VECTOR_DIMENSIONS and the configured model is not on the
known-rejecting allowlist (currently text-embedding-ada-002).
The provenance check (was VECTOR_DIMENSIONS explicitly set?) lives as
EmbeddingSettings.resolve_send_dimensions() because it needs access to
model_fields_set, which the standalone resolver does not have. The
resolved boolean is passed into _EmbeddingClient at construction time;
the client never inspects mode or provenance.
Also pins cloudevents <2.0 — 2.0.0 reorganized the package and dropped
cloudevents.conversion and cloudevents.http, which src/telemetry/emitter.py
imports. The original `>=1.12.0` constraint allowed the broken 2.0 resolve.
With the pin, the imports resolve cleanly and the basedpyright warning
cascade (37+ warnings about unknown types) disappears.
Drive-by cleanups (all unnecessary cast/ignore comments flagged by
basedpyright after the cloudevents downgrade):
- vector_store/lancedb.py, tests/conftest.py, and
tests/deriver/test_vector_reconciliation.py — drop dead pyright ignores
- sdks/python/src/honcho/http/{async_,}client.py — drop unnecessary
cast(datetime, ...) (parsedate_to_datetime already returns datetime)
- vector_store/turbopuffer.py — cast(Any, rows) for the upsert_rows
TypedDict that the SDK exposes but our row builder doesn't satisfy
- tests/test_datetime_parsing.py — ignore reportArgumentType on the
test that deliberately passes wrong types to assert raises
* feat(models): honor EMBEDDING_VECTOR_DIMENSIONS in pgvector columns
* feat(startup): atomic swap dim-vs-MIGRATED guard for runtime schema validator
Add src/startup/embedding_validator.py that introspects the actual pgvector
column dim at boot and refuses to start if it does not match
EMBEDDING_VECTOR_DIMENSIONS. Runs after the DB pool is up and before the
embedding client is constructed, in both src/main.py (FastAPI lifespan) and
src/deriver/__main__.py.
Implementation details:
- Schema-qualified pg_attribute join through pg_class/pg_namespace respects
DB.SCHEMA rather than relying on search_path
- Bounded retry (3 attempts, 1s backoff) for transient introspection failure,
then fail-closed with "could not validate embedding schema" — uncertainty
is not a green light to serve traffic
- External-store sampler (turbopuffer, lancedb) enumerates workspaces from
the application DB and probes their lazy-created namespaces; current
per-namespace probe is a no-op stub since the SDKs do not expose
uniform dim introspection — full enumeration is left to
`configure_embeddings --report` in Phase 3
Atomic guard swap: deletes the old dim-vs-MIGRATED config validator (which
forbade non-1536 pgvector unless MIGRATED=True) in the same commit as the
new runtime validator. There is no release window where non-1536 pgvector
can start unprotected. The 9 dual-write branches that use VECTOR_STORE.MIGRATED
remain untouched and load-bearing for legacy-tenant backend swaps.
VECTOR_STORE_DIMENSIONS deprecation: drop the "must match" raise; in
propagate_namespace, check model_fields_set and emit logger.warning +
DeprecationWarning (DeprecationWarning alone is filtered by Python's default
config and would not reach operators). Always overwrite with
EMBEDDING.VECTOR_DIMENSIONS regardless.
Test changes:
- tests/test_models_vector_dim.py: Phase 1's VECTOR_STORE_TYPE=lancedb +
MIGRATED=true escape hatches removed; the test now passes on plain
EMBEDDING_VECTOR_DIMENSIONS=768
- tests/llm/test_model_config.py: the two tests asserting the old guards
replaced with tests for the new deprecation + acceptance behavior
- tests/startup/test_embedding_validator.py: 10 new tests — dim assertion
logic (pass/mismatch/missing/unbounded/non-public-schema), fail-closed
retry budget, real-test-DB pass, real-DB ALTER-then-validate, deprecation
warning capture, non-1536 + pgvector + MIGRATED=false at config time
* feat(scripts): add configure_embeddings bootstrap CLI
Adds scripts/configure_embeddings.py alongside the other one-off scripts
(provision_db, migrate_db, generate_jwt_secret, etc.). Invoked as
`uv run python scripts/configure_embeddings.py` — same convention as the
existing scripts in that directory, including the sys.path shim that
lets src.* imports resolve when run directly.
Bootstrap step for self-hosted installs at a non-default
EMBEDDING_VECTOR_DIMENSIONS — runs between `alembic upgrade head` and
starting the API/deriver.
pgvector ALTER safety (single transaction):
- LOCK TABLE {schema}.documents, {schema}.message_embeddings IN ACCESS
EXCLUSIVE MODE — closes the TOCTOU window between population check
and ALTER
- COUNT(*) WHERE embedding IS NOT NULL on both tables; refuse with a
non-zero exit if either is populated (ALTER ... USING NULL would
silently wipe those vectors)
- Snapshot HNSW index DDL from pg_indexes; drop, ALTER, recreate from
the captured DDL so operator-set HNSW params (m, ef_construction)
survive the round trip
External vector stores (turbopuffer, lancedb) are never created or
modified — namespaces are per-workspace and lazy-created on first write.
The --report mode enumerates workspaces and collections from the
application DB, derives the expected namespaces via
get_vector_namespace(), and prints a per-namespace status table.
CLI modes (mutually exclusive):
- (default) interactive: print plan, prompt to confirm
- --dry-run: print plan and exit 0 without touching the DB
- --yes: apply without prompt
- --report: print external-store namespace inventory and exit
Also updates src/startup/embedding_validator.py error-message paths and
docs/v3/contributing/configuration.mdx invocations to point at the new
script location.
Tests cover plan no-op, plan needs-alter, plan raises on missing column,
ALTER + HNSW round-trip, refuse-when-populated (monkeypatched count to
avoid wiring the full workspace/peer/collection/document FK chain just
to land one vector row), and idempotency.
* docs: add changing-embeddings operations page
Document the supported way to change EMBEDDING_VECTOR_DIMENSIONS or
EMBEDDING_MODEL_CONFIG__MODEL on a Honcho deployment: provision a new
deployment at the desired configuration, replay source data out of
band, cut over at the application layer.
The page explains the asymmetry:
- Dimension is machine-enforced as immutable. The startup validator
introspects pg_attribute and crashes the API/deriver on mismatch.
- Model is operator-owned. There is no persistent metadata recording
which model produced each vector, so a same-dim model swap is
silently undetectable — flagged with a Warning callout.
Also documents the truncation edge case (text-embedding-3-large truncated to 1536 with EMBEDDING_VECTOR_DIMENSIONS left at default)
and the DIMENSIONS_MODE=always mitigation, plus a pointer that
storage-backend swap (VECTOR_STORE_MIGRATED + reconciler) is a distinct operation unaffected by this work.
Registers the page in docs/docs.json under the Self-Hosting nav group
and cross-links from configuration.mdx.
* fix(embedding): correct turbopuffer regex + tighten DIMENSIONS_MODE docs
- Turbopuffer attribute type for a vector column is `[N]f32` / `[N]f16` /
`[N]i8`, not `f32_vector(N)` as the earlier probe assumed. The earlier
regex returned None for the real SDK format, so existing Turbopuffer
namespaces would have been reported as "missing" instead of validated
for mismatch. Regex switched to `\[(\d+)\]` which is the
vendor-stable shape. Test cases rewritten to lock the actual format.
- docs/v3/contributing/configuration.mdx had a contradictory pair of
bullets: 223 said explicit 1536 makes `auto` forward dimensions=, 224
said `auto` would skip the parameter because 1536 is the default.
Operators reading both would (rightly) conclude they need `always`
even when `auto` would work. Rewrote both bullets so:
- `auto` is provenance-driven (explicit-set, not non-default-value).
- `always` is positioned as defense-in-depth for config layers that
might strip explicit default-valued envs, not the only path for
same-as-default truncation.
* fix(embedding): address PR #678 review comments
CodeRabbit + Rajat review feedback. All actionable items addressed
except two false-positives (responded on PR).
Bug fixes:
- deriver telemetry leak: validator was called outside try/finally so
shutdown_telemetry() did not run on validation failure. Moved inside.
- _emit_report printed "no effect with pgvector" unconditionally,
including from implicit post-apply calls. Added is_report_mode flag;
only print on explicit --report.
- LanceDB and Turbopuffer probes returned None when the namespace
existed but its schema was malformed (no vector field / unparseable
type string), silently bucketing real corruption as "missing"
(lazy-create) and letting it pass the startup validator. Now raise
VectorStoreError with actionable diagnostics; None remains valid only
for "namespace does not exist."
- Startup validator only sampled message namespaces; added a parallel
Collection-row sample so document namespaces are probed too, with the
same dim assertion. Mirrors the --report path.
Hygiene:
- StartupValidationError now subclasses HonchoException so existing
exception handlers recognize it. ValidationException is @final and
has 422 request-validation semantics that would be misleading here.
- scripts/configure_embeddings.py main() no longer spins up two event
loops. engine.dispose() moved into a try/finally inside _async_main
so cleanup runs in the same loop as the pipeline.
- Replaced hand-rolled retry loop with tenacity.AsyncRetrying; same
fail-closed semantics, less code, before_sleep_log for visibility.
- Added _validate_identifier() defense-in-depth: DB.SCHEMA and HNSW
index names are regex-checked against [A-Za-z_][A-Za-z0-9_]* before
SQL interpolation. Operator config + DB catalog are not user input
under the current threat model, but the constraint is cheap to gate.
Test + docs:
- test_app_settings_accepts_non_1536_with_any_vector_store_configuration
now actually exercises turbopuffer (was missing); supplies a dummy
TURBOPUFFER_API_KEY to satisfy the model_validator.
- changing-embeddings.mdx: hyphenated "out-of-band" per reviewer style.
* fix: modify conftest to fix ci
* fix: ci tests for typescript server
* fix: Add JSON repair for truncated LLM responses across all providers and Gemini thinking budget support
LengthFinishReasonError from OpenAI-compatible providers (custom, openai, groq) was crashing the deriver
with 14k+ occurrences in production. The vLLM path already had repair logic but it was gated on
provider=="vllm", unreachable when routing through litellm as a custom provider.
- Extract shared _repair_response_model_json() helper for all providers
- Catch LengthFinishReasonError in OpenAI/custom parse() path and repair truncated JSON
- Add repair fallback to Anthropic and Gemini response_model paths
- Add repair fallback to Groq response_model path
- Pass thinking_budget_tokens to Gemini 2.5 models via thinking_config
- Add 14 tests covering repair paths for all providers and Gemini thinking budget
Fixes HONCHO-YC
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
* feat: live llm integration tests
* feat: Consistent Model Config Protocol
* fix: migrate the remaining app callers off the legacy llm_settings path
* fix: Docs and regression tests
* fix: refactor llm runtime path to model-config-only API
* fix: refactor config to nested model-config source of truth
* fix: refactor llm streaming and tool dispatch through backends
* fix: cut over llm config to nested model_config only
* fix: collapse vllm and custom into openai_compatible transport
* feat: refactor llm config to explicit transports and bare model ids
* feat: (embed) Add configurability for embedding model
* fix: tests for embedding provider
* fix: Address Review Comments
* fix: (llm) remove Groq backend and per-vendor base URLs
* chore: move llm tests
* fix: (llm) address review findings — config regressions, backend bugs, dead code
* fix: address backend end silly errors
* chore: (docs) update configuration and self-hosting guides
* chore: fix tests
* fix: address code rabbit comments
* fix: add validation to the dream settings
* fix: further address code rabbit comments
* fix: Address Code Rabbit Comments
* fix: Another round of code rabbit
* fix: Address Code Rabbit Nits
* fix: tests
* refactor: rename thinking validator to reflect transport scope
_validate_anthropic_thinking_minimum only enforces the >=1024 rule for
Anthropic and no-ops for other transports, so the name was misleading
now that it's shared across ConfiguredModelSettings, FallbackModelSettings,
and ModelConfig. Renamed to _validate_thinking_constraints with a docstring
clarifying per-transport behavior. No logic change.
* fix(config): drop transport-specific thinking params when env override changes transport
_fill_defaults_for_nested_field previously preserved the default MODEL_CONFIG's
thinking_budget_tokens/thinking_effort across a transport override. This leaked
Gemini-family defaults (e.g. thinking_budget_tokens=1024) into OpenAI-transport
overrides, and the OpenAI backend then correctly rejected the unsupported param
at call time (OpenAI uses reasoning.effort, not a token budget).
The helper now strips thinking_budget_tokens and thinking_effort from the
default dict when the env override supplies a transport different from the
default's. Explicit thinking params in the override are preserved.
* fix(config): apply thinking-param strip to dialectic level merge too
DialecticSettings._merge_level_defaults does its own inline MODEL_CONFIG
merge (parallel to _fill_defaults_for_nested_field), so the previous fix
missed dialectic-level overrides. E.g. flipping
DIALECTIC_LEVELS__minimal__MODEL_CONFIG__TRANSPORT from gemini (default)
to openai still leaked the default thinking_budget_tokens=0 into the
openai config, which the OpenAI backend then rejected at call time.
The level-merge path now applies the same 'strip transport-specific
thinking params when transport changes' rule as the generic helper.
Added a regression test exercising the merge validator directly.
* refactor(llm): wire ModelConfig knobs through, prune clients.py migration leftovers
Three connected fixes to finish carving the LLM stack out of src/utils/clients.py
and into src/llm/:
1. Propagate ModelConfig tuning knobs into backend calls.
honcho_llm_call_inner built extra_params from only {json_mode, verbosity},
silently dropping top_p, top_k, frequency_penalty, presence_penalty, seed,
and operator-supplied provider_params from any ModelConfig. Thread the
selected config through ProviderSelection and merge
build_config_extra_params(selected_config) into extra_params; per-call
kwargs still win over provider_params defaults. Makes
_build_config_extra_params public as build_config_extra_params so
clients.py and request_builder.py share one translation. Adds
TestModelConfigExtraParamsPropagation covering OpenAI/Anthropic knob
propagation, provider_params passthrough, and per-call override
precedence.
2. Drop dead extract_openai_* duplicates in clients.py.
extract_openai_reasoning_content, extract_openai_reasoning_details, and
extract_openai_cache_tokens had no callers outside their own definitions
— the live implementations live in src/llm/backends/openai.py. -103
lines from clients.py.
3. Unify on ModelTransport, delete SupportedProviders.
The "google" vs "gemini" split forced a _provider_for_model_config
translation shim in two places. Replace all SupportedProviders usages
with ModelTransport, rename CLIENTS["google"] → CLIENTS["gemini"],
update provider branches + LLMError labels + reasoning-trace entries
accordingly. Trace JSONL now writes "provider": "gemini" instead of
"google" — consistent with the broader env-var rename cutover.
Also tidies up pre-existing basedpyright findings in tests/llm/test_model_config.py
(pydantic before-validator dict inputs + descriptor-proxy call).
ruff: clean. basedpyright: 0 errors, 0 warnings. Tests: 153/153 pass across
tests/utils/test_clients.py, tests/utils/test_length_finish_reason.py,
tests/llm/, tests/dialectic/, tests/deriver/.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* refactor(llm): finish the src/utils/clients.py → src/llm/ migration
honcho_llm_call_inner now delegates to request_builder.execute_completion
and execute_stream instead of re-implementing backend call scaffolding
inline. The new _effective_config_for_call helper carries per-call kwargs
(temperature, stop_seqs, thinking_budget_tokens, reasoning_effort) onto
the selected ModelConfig — or synthesizes a minimal config for the
test-only callers that pass provider+model directly. max_output_tokens
is zeroed on the effective config to preserve the current
"per-call max_tokens wins" semantic; honoring ModelConfig.max_output_tokens
is a separable correctness concern.
Side effect of routing through the new path: ConfiguredModelSettings'
thinking_budget_tokens validator now fires on synthesized configs.
test_anthropic_thinking_budget was asserting that a sub-1024 budget
propagated to Anthropic — bumped to 1024 to match what Anthropic actually
accepts.
Unified client construction. Promoted the cached client factories in
src/llm/__init__.py (get_anthropic_client, get_openai_client,
get_gemini_client, get_{anthropic,openai,gemini}_override_client) to
public API and added them to __all__. Promoted
credentials._default_transport_api_key → default_transport_api_key.
Deleted the duplicate _build_client and _default_credentials_for_provider
from clients.py; _client_for_model_config now falls through to the
public factories. CLIENTS dict and _get_backend_for_provider stay as the
mockable seam for the ~50 patch.dict(CLIENTS, {...}) test call sites.
Wired operator-configurable Gemini cached-content reuse end-to-end.
PromptCachePolicy moved from src/llm/caching.py into src/config.py so
ModelConfig can reference it as a field without a circular import;
caching.py re-exports the name for existing imports. Added
cache_policy: PromptCachePolicy | None on ConfiguredModelSettings,
FallbackModelSettings, ResolvedFallbackConfig, and ModelConfig.
resolve_model_config, _resolve_fallback_config, and
_select_model_config_for_attempt copy the field through.
honcho_llm_call_inner passes effective_config.cache_policy into
execute_completion / execute_stream, so operators opt in via
e.g. DERIVER_MODEL_CONFIG__CACHE_POLICY__MODE=gemini_cached_content
and the selection actually fires instead of sitting on a dead path.
New regression test test_cache_policy_reaches_gemini_backend asserts the
PromptCachePolicy object reaches the Gemini backend's extra_params.
ruff + basedpyright: clean. Tests: 154/154 pass across
tests/utils/test_clients.py, tests/utils/test_length_finish_reason.py,
tests/llm/, tests/dialectic/, tests/deriver/.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* refactor(llm): move all LLM orchestration into src/llm/ and delete clients.py
The 1624-line src/utils/clients.py has been carved up into focused modules
under src/llm/ and deleted. There is now one golden path for LLM
orchestration and no dual entrypoint.
New module layout:
src/llm/
__init__.py thin stable re-export surface
api.py public honcho_llm_call with retry + fallback + tool
loop delegation
executor.py honcho_llm_call_inner (single-call executor); bridges
to request_builder.execute_completion / execute_stream
tool_loop.py execute_tool_loop + stream_final_response, plus
assistant-tool-message and tool-result formatting
runtime.py AttemptPlan dataclass (replaces the loose
ProviderSelection NamedTuple), effective_config_for_call,
plan_attempt, per-retry temperature bump, attempt
ContextVar
registry.py single owner of CLIENTS dict + cached default and
override SDK-client factories + backend/history-adapter
selection + high-level get_backend(config)
conversation.py count_message_tokens, tool-aware message grouping,
truncate_messages_to_fit
types.py HonchoLLMCallResponse, HonchoLLMCallStreamChunk,
StreamingResponseWithMetadata, IterationData,
IterationCallback, ReasoningEffortType, VerbosityType,
ProviderClient
request_builder.py low-level request assembly (ModelConfig → backend
complete/stream); no longer owns credential resolution
credentials.py default_transport_api_key, resolve_credentials
caching.py gemini_cache_store; re-exports PromptCachePolicy
from src.config
backend.py Protocol + normalized result types
history_adapters.py provider-specific assistant/tool message shapes
structured_output.py
backends/ AnthropicBackend, OpenAIBackend, GeminiBackend
handle_streaming_response had no production callers; it is deleted. The
three tests that used it now drive honcho_llm_call_inner(stream=True,
client_override=...) directly, which exercises the same code path the
public API uses.
Dead credential passthrough removed. The ProviderBackend Protocol and
all three concrete backends no longer accept api_key / api_base — those
are baked into the underlying SDK client at registry construction time
and were being del'd everywhere they appeared. request_builder also
stops resolving and forwarding them.
Client construction is unified. The cached default-client factories
(get_anthropic_client, get_openai_client, get_gemini_client) and override
factories (get_*_override_client) are promoted to public API; the
module-level CLIENTS dict populates from them and remains the
patch.dict(CLIENTS, {...}) mocking seam tests rely on. Old duplicate
helpers (_build_client, _default_credentials_for_provider) are gone.
default_transport_api_key is promoted to public.
Application imports now come from src.llm (dreamer, dialectic, deriver,
summarizer, telemetry-adjacent tests). No code imports from
src.utils.clients anywhere in the repo.
ruff: clean. basedpyright: 0 errors, 0 warnings. Tests: 1013/1013 pass
across the entire non-infra test suite (excluding tests/unified,
tests/bench, tests/live_llm, tests/alembic).
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* fix(llm): sanitize tool schemas for Gemini's function_declarations validator
Gemini's native-transport function-declarations validator accepts a narrow
subset of JSON-Schema / OpenAPI: type, format, description, nullable, enum,
properties, required, items, minItems, maxItems, minimum, maximum, title.
Anything else — additionalProperties, allOf, if/then/else, $ref, anyOf,
oneOf, $defs, patternProperties — triggers an INVALID_ARGUMENT 400 at call
time.
Our agent tool schemas in src/utils/agent_tools.py use several of those
(additionalProperties: false, allOf + if/then conditionals) because they
were authored for OpenAI strict-mode + Anthropic, which need the richer
vocabulary. GeminiBackend._convert_tools was passing them straight through.
Add _sanitize_schema(): walks the parameters tree and drops unsupported
keywords while preserving semantics for the keywords that hold user data
(properties maps field-name → sub-schema; required / enum are lists of
literals; items is a single sub-schema). Other backends are untouched and
continue to receive the full strict schemas.
Regression tests:
- test_gemini_sanitize_schema_strips_unsupported_keywords: confirms
additionalProperties, allOf + if/then, and $defs are stripped at nested
levels while legitimate fields survive.
- test_gemini_convert_tools_sanitizes_parameters_schema: end-to-end
_convert_tools output has no forbidden keys.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
* fix: fix tool calling syntax for gemini
* refactor(llm): normalize defaults, widen OpenAI reasoning-model routing
* chore: fix test
* fix(llm): address post-migration review feedback
* fix(llm): gemini robustness + dreamer specialist ergonomics
* chore: addres review comments
* chore: (docs) unrelease changelog addition
* chore: (docs) merge commit changes
---------
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Co-authored-by: Erosika <eri@plasticlabs.ai>
* chore: Update versioning for release
* fix: remove db creation at start and sync migrations and models
* fix: Checkpoint changing metamessage schema
* chore: linter fixes
* fix: session cloning working
* Hybrid long-term memory (#92)
* Add TOM method switching
* Add system prompt and note on format
* Add persistence tweaks
* Specify format for each section of user representation
* Parse XML tags before saving representation metamessage
* Clean up
* Use Claude 3.5 Haiku and refine prompt
* Simplify message processing
* chore: update token limit on dialectic and model for deriver
* Add embedding-based long-term fact retrieval
* Fix bug preventing new documents from being created
* Use multiple queries + tweak prompt
* Fix collection name bug + add duplicate removal
* First implementation of on-demand user rep generation
* WIP debug on-demand user rep changes
* Fixed representations not being stored & deriver issue
* Some speed improvements
* Play with number of facts / queries
* WIP prompt caching for Claude
* WIP fix anthropic caching
* Anthropic prompt caching working but messages too short
* Use Cerebras for small inferences
* Make dialectic responses 1000 tokens max
* Make user representation generation model a constant
* Use llama 3.1 8b for query generation
* Update env template
* Add crud.get_or_create_protected_collection
* rabbit comments
* Fix linter issues
* Add Cerebras to stream router method
* Better handling of default-empty string args
* Change prints to debug logs
* Add error handling to TOM inference
* Handle missing/empty client in model responses
* Handle no messages case in get_chat_history
* Fix indent
* Add error handling to single_prompt methods
* Fix get_or_create_user_protected_collection
* Simplify openAI-compatible model client instantiation
* Remove health endpoint
* Remove LocalEmbeddingStore
* Change prints to debug logs
* Change sentry track
* Code review changes
* Add README to ToM module
* Switch to Groq
* Fix inconsistent openai compatible provider list in stream()
* Update env template to include Groq variables
* Add model_client tests
* fix: Fix unit tests
---------
Co-authored-by: Vineeth Voruganti <13438633+VVoruganti@users.noreply.github.com>
* add scoped API keys (#91)
* add AUTH_JWT_SECRET and ADMIN_KEY, use in security middleware (TODO granular keys)
* WIP: convert all API paths to use scoped keys
* add basic unit tests for API keys, ruff formatting
* MVP of route using JWT for payload
* add get_user_from_token
* add key table to postgres, use it to enable key revocation
* add key revocation pt 2 -- fix order of param checks
* finish convenience routes that assume params from JWT
* add tests for key API
* get_keys
* add secrets utility script, add key rotation, fill out tests
* add tiny cache as PoC
* nits, validations, etc
* only create keys table migration if necessary
* fix keys tests to always use auth
* tiny fix to make custom DATABASE_SCHEMA work
* review: add better docs, fix security issue with cache, clear db on rotation, and more
* remove rotation
* remove key database entirely
* Add `/all` path to get all apps (#94)
* add `/all` path for apps
* assert vector extension installed (need this for groudon)
* review
---------
Co-authored-by: Vineeth Voruganti <13438633+VVoruganti@users.noreply.github.com>
* add scoped API keys (#91)
* add AUTH_JWT_SECRET and ADMIN_KEY, use in security middleware (TODO granular keys)
* WIP: convert all API paths to use scoped keys
* add basic unit tests for API keys, ruff formatting
* MVP of route using JWT for payload
* add get_user_from_token
* add key table to postgres, use it to enable key revocation
* add key revocation pt 2 -- fix order of param checks
* finish convenience routes that assume params from JWT
* add tests for key API
* get_keys
* add secrets utility script, add key rotation, fill out tests
* add tiny cache as PoC
* nits, validations, etc
* only create keys table migration if necessary
* fix keys tests to always use auth
* tiny fix to make custom DATABASE_SCHEMA work
* review: add better docs, fix security issue with cache, clear db on rotation, and more
* remove rotation
* remove key database entirely
* Add `/all` path to get all apps (#94)
* add `/all` path for apps
* assert vector extension installed (need this for groudon)
* review
---------
Co-authored-by: Vineeth Voruganti <13438633+VVoruganti@users.noreply.github.com>
* chore: README and CHANGELOG updates
* add JWT expiry
* fix: Consolidate get methods with JWT token resolution
* chore: Add Annotation to Path, Query, and Body params
* chore: run ruff formatter
* chore: nits & add one exhaustive test of a query route
* fix: undo change to fly.toml
* fix: Langfuse tracing
* Consolidate Get Methods (#96)
* fix: Consolidate get methods with JWT token resolution
* chore: Add Annotation to Path, Query, and Body params
* chore: run ruff formatter
* chore: nits & add one exhaustive test of a query route
* fix: undo change to fly.toml
---------
Co-authored-by: dr-frmr <docterformer@protonmail.com>
* fix: dev-667 fix streaming endpoint
* fix: Anthropic Langfuse Tracing
* fix: add scripts folder to dockerfile
* fix: Remove redundant fields from pydantic schemas
* fix: Add deeper protection on reserved collection
* fix: Consolidate chat and stream methods
* docs: Update Mintlify API Reference and Changelog
* remove langchain guide, update architecture diagram
* honcho mcp server
* chore: Update .env template
* update discord, temporarily remove other guides
* Limit dialectic & deriver context usage with two-scale progressive summarization (#97)
* WIP two tiered summaries
* Move to process_item
* Save user rep metamessage even if no message_id
* Change number of messages per short summary
* Fix broken mock
* Remove prints
* chore: fix test
---------
Co-authored-by: Vineeth Voruganti <13438633+VVoruganti@users.noreply.github.com>
* feat: Add Gemini Support, link facts to message, use 8b for dialectic fact queries
* chore: Styling
* chore: coderabbit nitpicks
* keep dialectic guide
* Add streaming guide
* Remove TODO from dialectic guide
* Fix JS snippets that referred to honcho singleton as client
* Add App explanation to architecture page
---------
Co-authored-by: Dani Balcells <18307962+danibalcells@users.noreply.github.com>
Co-authored-by: doria <93405247+dr-frmr@users.noreply.github.com>
Co-authored-by: dr-frmr <docterformer@protonmail.com>
Co-authored-by: vintro <vince@plasticlabs.ai>
Co-authored-by: Daniel Balcells <dbalcells@gmail.com>
* fix(storage) Remove Prepared Statements to make compatible with Supavisor
* fix(deriver and dialectic) Use latest user level user_represenation
* fix(storage) Switch to text columns for best postgres practices
* fix(storage) Update docstrings and document query method
* chore(release) v0.0.14 Release State
* chore(release) CHANGELOG
* fix(deriver) Add more detailed profiling with sentry
* chore(release) CHANGELOG
* fix(deriver) increase retry limit on anthropic
* ready to test deriver
* prompts ready for testing
* compare entire list, don't loop through facts one by one
* parse xml, batch process ai query docs
* timing works
* session id is writing to csv now
* ignore timing logs
* infinite backrooms-inpsired prompts
* yousim/ib inspired prompts v0
* accept incoming changes on agent.py
* dialectic to sonnet for now
* remove timing, use py-spy
* cleanup from timing removal
* remove extra timing stuff
* Route for User LeveL Metamessages
* functional state
* chore: Clean up artifacts
* fix(deriver) update database calls for new schema and payload
* chore(docs) Update Honcho Version
---------
Co-authored-by: Vineeth Voruganti <13438633+VVoruganti@users.noreply.github.com>
* fix(storage) Remove location_id
* feat(deriver) Deriver can process multiple sessions at once and uses rich for prints
* fix(route) collection update error handling
* chore(docs) update openapi and version numbers
Phil
Phil
Vineeth Voruganti
Rajat Ahuja
doria
vintro