test_each_provider_has_expected_ids asserts the exact set of model ids per
provider against a snapshot. For the first-party catalogs that is what we
want: the snapshot exists so an upstream id typo fails before it silently
breaks every agent that references the old id.
openai_compatible is different. It is the zero-Python extension point —
the README's "Add an OpenAI-compatible provider" section and
examples/mistral.yaml.example both tell you to drop a YAML next to the
built-ins, and the loader globs the whole directory. So anyone who follows
those instructions gets a permanently red test on their fork, for having
done the documented thing. The _internal suffix carve-out only covers the
one filename we happen to gitignore.
Treat the snapshot as a floor for that provider: the built-in ids must all
still be present, extra ones are somebody else's provider. The rename this
guards against still drops an id, so it still fails, and the assertion
names the missing ids instead of printing two large sets to diff by eye.
Verified by copying examples/mistral.yaml.example into the catalog
directory: red before, green after, and still red when a built-in id is
renamed.
main has been red since 295969b: the model catalog gained claude-fable-5,
gpt-5.6-sol and the alibaba and zai providers, but the pieces that track
the catalog were not updated alongside them. Five tests across
test_model_registry_yaml.py and test_pricing.py have been failing since.
- EXPECTED_IDS picks up the two new hosted models and the two new
openai_compatible ones. The snapshot is deliberate — it exists to
catch an unintended rename — so it is updated, not relaxed.
- test_everything_set sets DASHSCOPE_API_KEY and ZAI_API_KEY alongside
the DEEPSEEK one it already set. Every openai_compatible catalog reads
its own key from the environment, so without them "everything set"
quietly excluded the two new providers.
- claude-fable-5 and gpt-5.6-sol had no rates, which
test_hosted_builtin_models_are_priced requires of every hosted model,
and which token_usage.cost and the quota ceilings both read. Priced on
the existing ladder: fable-5 above Opus 4.7 at 10/50 per million with
cached input at a tenth and cache writes at 1.25x, matching the other
Anthropic entries; sol alongside the gpt-5.5 flagship at 5/30.
Rename the unused *_cost_per_token capability fields to USD per 1M tokens,
add prompt-cache read/write rates, and ship list prices for the hosted
catalogs. The old per-token keys still load, scaled, with a warning.
docsgpt/pricing.py turns a call's token bins into a USD cost. Models with
no declared rate cost $0 unless QUOTA_UNPRICED_RATE_PER_MILLION is set.
Review follow-up. SCIM_ENABLED=true with no SCIM_TOKEN was only caught per
request, as a 503 from the SCIM routes. The auth group's model validator
now rejects that combination when Settings loads, the same way it rejects
AUTH_TYPE=oidc without its required settings, so the misconfiguration is
reported once at startup rather than on the first provisioning call. The
route-level guard stays as defence in depth.
The docs site failed to build: a bare "<= 1" in the prose of the
generated page is parsed by MDX as the start of a JSX tag ("Unexpected
character '=' before name"). Constraints are rendered as code spans now,
where MDX leaves them alone, and a test rejects any bare <, { or } outside
a code span so a future description cannot reintroduce the failure.
Verified with a local next build of the docs site.
Review follow-up. EMBEDDINGS_POOLING became Optional[Literal["cls", "mean"]],
but the group-base rule that maps "", "None" and whitespace to None only
matched Optional[str], so EMBEDDINGS_POOLING= in a .env file would have
failed validation. The rule now also covers an Optional Literal whose
choices are all strings, which lets the AUTH_TYPE validator drop its own
copy of that handling.
Review follow-up. The per-group secret validators normalised a hand-picked
list of API keys, which left other optional credentials and overrides
(OPEN_ROUTER_API_KEY, S3 and Daytona keys, ELASTIC_PASSWORD, the OIDC
trio, connector client ids, MICROSOFT_AUTHORITY, MCP_OAUTH_REDIRECT_URI)
holding the literal "None" or "" a .env file spells "unset" with, so
truthiness checks and fallbacks downstream saw a value. One rule on the
group base replaces those lists: every Optional[str] field maps "", "None"
and whitespace to None and strips real values. Plain str fields are left
alone. The OIDC required-settings check therefore also rejects those
spellings.
EMBEDDINGS_POOLING is Literal["cls", "mean"] with case-insensitive
parsing; its consumer silently ignored anything else.
Bounds added where the consumer rejects or misbehaves on the value:
SCHEDULE_RUN_OUTPUT_RETENTION_DAYS and MESSAGE_EVENTS_RETENTION_DAYS (the
cleanup repositories raise on <= 0), EMBEDDINGS_DELEGATE_TIMEOUT, the
remote-device idle/pairing/invocation TTLs and CELERY_VISIBILITY_TIMEOUT
(> 0), REMOTE_DEVICE_CMD_QUEUE_TTL_SECONDS (> 605, the documented drain
deadline), GRAPHRAG_MAX_CHUNKS_FOR_EXTRACTION (>= 0; negative would slice
the pending list from the end).
The generated reference now renders generic type arguments
(dict[str, int] rather than dict).
SAGEMAKER_REGION, SAGEMAKER_ACCESS_KEY and SAGEMAKER_SECRET_KEY survive
only as a fallback for the S3_* credentials. They carry
Field(deprecated=...) now, so any read emits a DeprecationWarning naming
the replacement and the generated reference shows the notice. The S3
store is the one sanctioned reader; it silences that warning locally
because it already logs its own operator-facing one when the fallback
is actually used.
DEFAULT_MAX_HISTORY was referenced nowhere. RETRIEVERS_ENABLED was read by
no code at all, while two docs pages described it as an enforced
allow-list; both the setting and those claims are removed.
The "AUTH_TYPE=oidc requires OIDC_ISSUER, OIDC_CLIENT_ID and
OIDC_FRONTEND_URL" check lived in app.py, so it only ran when the Flask
app was imported; a worker or script with the same misconfiguration
started fine. It is now a model validator on the auth group and runs
wherever Settings is loaded, with the same message.
DEPLOYMENT_TYPE, which app.py read straight from the environment to
decide whether a missing JWT_SECRET_KEY is fatal, is a documented
setting on the server group now, so it shows up in the reference like
every other variable the app reads.
Enum-like settings whose allowed values were only listed in a comment are
now Literal types, so a typo fails at startup with a message naming the
allowed values instead of falling through to a default with a warning
(or, for VECTOR_STORE, failing on first use):
AUTH_TYPE, VECTOR_STORE, STORAGE_TYPE, URL_STRATEGY, OCR_BACKEND,
OCR_ENGINE, SANDBOX_BACKEND, DOC_PARSER_ENGINE, TTS_PROVIDER, STT_PROVIDER
Each keeps a before-validator that strips and lower-cases the value, since
the registries that consume them already lower-cased at the use site, and
AUTH_TYPE maps the "None"/"none"/"" spellings a .env file carries to None
(it was the string "None" before, which only worked because nothing
compared against it). An empty TTS/STT provider still means "off".
LLM_PROVIDER stays a plain str because providers are plugin-extensible.
Containers are typed (dict[str, int], list[str], dict[str, Any]) instead
of bare dict/list, six fields that were Optional with a non-None default
are plain, and integer settings whose description already states a range
carry it as a constraint (ge=0 for "0 disables", ge=1 for counts that
cannot be zero, 0 < threshold <= 1).
The hand-maintained settings page documented 95 of 258 settings and
.env-template 42, and both drifted as fields were added. The field
descriptions now live on the model, so the reference is rendered from it:
python -m docsgpt.core.settings.reference --write
writes docs/content/Deploying/Settings-Reference.mdx, one section per
settings group with each field's type, default, constraints, aliases and
description. --check reports a stale page, and tests/core/test_settings.py
fails when the checked-in page no longer matches the definitions, so a
new setting cannot land undocumented.
test_settings.py also pins the composition contract: every group field is
a flat Settings attribute, no field is defined twice, every field has a
description, and the secret-normalising validator of every group is
applied (the case that a shared method name would silently drop).
The App Configuration page points at the reference instead of at
settings.py, and the reference is listed in the Deploying navigation.
Outside a checkout the data home was the working directory, so running
`docsgpt api` from another folder silently used different settings and data.
It is now ~/.docsgpt/server (/opt/docsgpt for root on Linux); DOCSGPT_HOME
and a checkout still take precedence. The API and worker commands create
the home and point out a .env left in the working directory.
- Ship tiktoken's cl100k_base inside the package and build the encoding
from it, so token counting never downloads anything.
- Default EMBEDDINGS_CACHE_DIR to <data home>/models instead of FastEmbed's
temp dir, and read tokenizer.json and repo metadata from that cache, so
a model downloads once and survives reboots.
- TTS_PROVIDER=none and STT_PROVIDER=none switch the speech features off:
the endpoints return 404, audio files fail to ingest with a clear
message, /api/config reports tts_available/stt_available, and the UI
hides the Speak and microphone buttons.
- Drop the Google Fonts Roboto import from the web UI.
- prefetch-models fills the cache the app reads; verify-offline checks the
packaged encoding.
- Docs: new Air-Gapped Deployment guide, settings and cache notes.
Both SDKs' 1.x/3.x majors run on httpx2 (the maintained fork of httpx by its
original author, published by Pydantic at github.com/pydantic/httpx2, version
line 2.x) instead of httpx, which is what makes these two bumps one change.
httpx2 is now a declared dependency because two modules import it directly.
Everything else in the 0.x -> 1.x / 2.x -> 3.x change lists is absent here:
no Text Completions, no `with_raw_response`, no raw `output_format` dicts, no
Bedrock client, and `requires-python` is already 3.12.
Three code changes, all forced by the bump:
The BYOM DNS-pinning client in `docsgpt/security/safe_url.py` is handed to
`OpenAI(http_client=...)`, and the SDK rejects an old-httpx client at
construction — which would have taken the SSRF guard offline. It is built on
httpx2 now, and its `sni_hostname` extension carries a `str` rather than
ascii bytes: httpcore passes the value straight to
`ssl.SSLContext.wrap_socket`, and the truststore backend httpx2 uses for the
default system trust store encodes it instead of accepting bytes. Bytes
therefore failed every real handshake while passing the existing tests, which
stub the transport out; the test now pins the type and says why.
The stream-retry error tuple in `docsgpt/llm/base.py` named `httpx`
exceptions only. The two libraries' exception classes are unrelated types, so
after the bump the retry silently stopped firing for openai and anthropic
while still working for google-genai and elevenlabs. It now covers both
stacks, with a parametrized test over each.
anthropic 1.x dropped temperature/top_p/top_k from `messages.create`'s
signature (passing one raises TypeError) without dropping them from the API,
so the provider forwards them through `extra_body`. The wire request is
unchanged and a model that rejects them 400s exactly as before.
- The image pins DOCSGPT_HOME=/app: it ships no checkout, so the data home
no longer depends on the working directory.
- api, worker, beat and migrate print the data home and env file they
resolved, so an API and a worker started from different directories show
it.
- The worker passes -Q only when asked; a bare worker consumes every
configured queue, which honours EMBEDDINGS_QUEUE and DOCUMENT_PARSE_QUEUE.
- The worker runs through celery.start and returns its exit code; click
usage errors print usage and exit 2 instead of a traceback.
- Windows: solo pool and no embedded scheduler (celery rejects -B there),
with a pointer to the new `docsgpt beat` command, which runs the
scheduler on its own.
- prefetch_models and verify_offline parse their arguments, so --help is
help rather than a model name.
- A DOCSGPT_ENV_FILE that is not a file raises instead of booting with
defaults.
- docsgpt api binds 127.0.0.1 by default, like gunicorn and uvicorn do;
--host 0.0.0.0 exposes it. The docs say so.
- The embedded Milvus and LanceDB defaults derive from the data home, so
they follow DOCSGPT_HOME like the faiss indexes and uploads do. A checkout
run from its root and the Docker image resolve to the same paths as before.
- The docs and the pyproject comment describe the CPU torch install as two
steps (torch and torchvision from the PyTorch CPU index first, then the
docling extra): pip picks the highest version across indexes, so
--extra-index-url only yields the CPU build while that index keeps pace
with PyPI.
- AGENTS.md separates DOCSGPT_HOME (moves the data home) from
DOCSGPT_ENV_FILE (selects the .env file); the docs example uses a password
placeholder.
pip install docsgpt (extras: docling, milvus) installs the backend with a
docsgpt command: api, worker, migrate, prefetch-models, verify-offline,
reembed. Second step of the PyPI work after the package rename.
- hatchling build; the version comes from docsgpt/version.py. The wheel is
the docsgpt package with the data it reads at runtime (prompts, model
catalogs, seed config, alembic.ini and migrations) and without the
Dockerfile, the exported requirements, the sample index and local runtime
data. The application import alias stays checkout-only. uv sync installs
the package editable now that [tool.uv] package = false is gone.
- docsgpt/cli.py: api (gunicorn + BoundedDrainUvicornWorker with the image's
flags, --reload for uvicorn), worker (Celery worker with beat embedded,
--no-beat/-Q/--concurrency/--pool, solo pool on macOS), migrate, and
argument pass-through to the maintenance scripts. --help imports no app.
- docsgpt/core/paths.py: runtime data lives in a data home (DOCSGPT_HOME,
else the checkout, else cwd); DOCSGPT_ENV_FILE overrides the env file.
Settings, the dotenv load, LocalStorage and the internal upload route use
it instead of "three directories above this file", which is site-packages
for an installed package. A checkout and the Docker image behave as before.
- [project] dependencies are compatible ranges so the package installs next
to other packages; uv.lock resolves to the same versions and the exported
requirements files are unchanged.
- package-build.yml builds and checks the wheel on PRs and installs it into
a clean venv; pypi-publish.yml publishes on a published release through
trusted publishing (environment pypi), or to TestPyPI on a manual run.
- Docs: Deploying -> Install with pip. AGENTS.md notes the package.
- The post-task reclaim skip recognises the legacy application.* embed name,
so query embeds queued by the previous release do not pay a full collect.
- The Azure compose file mounts host data on /app/{indexes,inputs,vectors},
where the process actually reads and writes; it mounted /app/application/...
before the rename and /app/docsgpt/... after it, and nothing wrote to either.
- The offline image check triggers on docsgpt/requirements*.txt again;
dependabot's pip entry points at docsgpt/.
- install_hint() and its docstring name docsgpt/requirements-<extra>.txt;
the test asserts the full path.
- application/vectors/ stays ignored: the compose files still mount it.
- The durability QA script quiets the docsgpt logger tree.
- Upgrade note: the three renamed source-sync entries start their timers
from the upgrade.
The backend import package is now docsgpt, the name it will carry on PyPI;
application was far too generic to install into anyone's site-packages.
git mv plus a mechanical rewrite of every import, dotted string and path
reference: 734 Python files, the compose files, Dockerfile, workflows, docs,
setup scripts, devcontainer, k8s manifests, vscode config, pytest and coverage
config, .gitignore. Behaviour is unchanged.
Kept for one release:
- A top-level application package whose meta-path finder resolves
application.x.y to the already-imported docsgpt.x.y object, so old imports
and entry points (celery -A application.app.celery,
uvicorn application.asgi:asgi_app) keep working with a FutureWarning.
- Celery registers every application.* task name as an alias of its
docsgpt.* task on start-up, so messages queued by the previous release still
run. The redbeat key prefix moves to redbeat:docsgpt:v2: so schedule entries
the previous release wrote are left unread instead of firing twice.
The backend image builds from the repository root (docker build -f
docsgpt/Dockerfile .) so it can ship the alias package; a root .dockerignore
allow-lists docsgpt/ and application/ and keeps caches, local data, .env
files, the sample index files and the Dockerfile out. Compose and the image
workflows point at the new context.
requirements.txt pinned torch and transformers in core although only docling
needs them, and on Linux torch pulls the CUDA 13 stack: 2.7 GB of the 3.0 GB
wheel download. Direct dependencies now live in pyproject.toml, uv.lock pins
everything, and application/requirements*.txt are exported from the lock by
scripts/export_requirements.sh (each file is the core set plus one extra).
The docling extra pins torch/torchvision/transformers itself and, on Linux,
resolves torch from the CPU-only PyTorch index (no nvidia packages). milvus
(pymilvus + milvus-lite, which pulls pyarrow) is the second extra.
application/core/optional_deps.py is the one place install hints come from;
the milvus store and the docling call sites use it so a missing extra fails
with the exact command to run.
Source access control
---------------------
`active_docs` is client-supplied and reached the retriever unchecked, and the
retriever queries `WHERE source_id = <id>` with no owner predicate — so any
caller could pass any source id to /stream or /api/answer and have another
tenant's documents quoted back, while /api/sources/<id>/search correctly
refused the same id. Gate it through `can_access`, the helper the guarded
endpoints already use, and filter `self.source` down to the authorized set.
Fails closed: no principal, or a check that errors, drops the source.
Three sibling paths had the same gap:
- workflow agent nodes: `AgentNodeConfig.sources` is written verbatim from
client JSON at save time and nothing validated it, so a node could name any
tenant's source. Gate against the workflow owner, so shared workflows keep
reading their owner's sources like shared agents do.
- /api/share: `_resolve_source_pg_id` resolved any id with no ownership
predicate and baked it into the agent the share creates; /api/search then
searched it. Authorize before attaching.
- search_service: re-resolve the ids stored on an agent row instead of
trusting them, so a row written by any future path with the same gap cannot
be read back.
Team grantees previously lost their source's retrieval config: the post-check
read was still owner-scoped, so it missed and fell back to defaults (an
`agentic_tool` source was bulk-prefetched for every grantee). Read unscoped
after `can_access` passes.
Retrieval
---------
`PGVectorStore._ensure_table_exists` created an IVFFlat index on the empty
table it had just created. IVFFlat computes centroids at build time, so those
centroids were random, and combined with the `source_id` post-filter a source
with hundreds of embedded chunks returned zero rows — retrieval reported no
documents, the model answered from memory, and nothing was logged. Stop
creating the index (exact search is correct and fast well past the sizes most
deployments reach); raise `ivfflat.probes` to sqrt(lists) where an index still
exists; and re-run a short indexed search exactly, since post-filtering means
no index setting can guarantee a full result. `graphrag` had the same
empty-table index with no fallback at all.
Also: bound `chunks` to 0-500 on both the request and agent paths (0 still
means "skip retrieval"), let a source's configured `retrieval.chunks` outrank
the request body, and cap ClassicRAG's per-source floor at
max(top_k, n_sources) so attaching sources cannot inflate the result set.
Silent failures
---------------
An empty retrieval was invisible to both the model and the client: the `source`
event was suppressed when the list was empty, so "searched and found nothing"
looked identical to "no source attached", and the prompt said nothing at all.
Emit the event always, and tell the model when a search ran and returned
nothing. A file that parses to nothing now fails ingest with a message naming
the cause instead of storing an embedding of the empty string. `score_threshold`
returns warnings when the active store or retriever cannot honour it.
Prompt structure
----------------
Retrieved documents move from the system prompt into the user turn, with the
injection guard restated next to them: they change every turn (defeating prefix
caching), they are third-party text that should not carry system authority, and
routing them through the query budget makes them truncatable rather than
silently crowding it out. Documents are shed lowest-ranked-first before the
question is touched.
The six chat presets (3 tones x 2 retrieval modes) differed only in their
Answering section; they are now composed from single-source fragments at load
time, not through Jinja inheritance, which would have opened a file-read
surface in the template sandbox and broken the tool-prefetch parser. Per-tool
guidance moves out of the prompt into tool schemas, so it travels with the tool
and cannot render when the tool is absent. A plain-text custom prompt is staged
as a persona value inside the skeleton instead of replacing it wholesale — it
used to silently lose the injection guard, platform block, memory and
attachments, and its braces are now inert.
Other fixes
-----------
- agents/base: an oversized system prompt drove the query budget negative and
dispatched a full-price request with an empty question; raise instead.
- llm/anthropic: migrate off the retired Text Completions API. It flattened
history to first+last message and ignored tools entirely. Adds the missing
Anthropic handler, without which every tool call was silently dropped.
- sources/upload: `sitemap` had no branch, so every sitemap ingest died on a
TypeError; `validate_url` now rejects a falsy URL cleanly.
- workflow nodes: retrieved documents never reached the node agent, so a
classic node with a source and an ordinary prompt answered "I have no
documents" while the run reported completed.
- parser/bulk: copy the metadata dict, or every chunk reports the last chunk's
token_count.
- crawler_loader: carry the page title, or citations render the whole chunk
body as the label.
Route opted-in models to /v1/responses while keeping Chat Completions as
the default. The flavor is selected per-model in the catalog YAML and
threaded through ModelCapabilities (with a new reasoning_effort field).
- Translate chat-shaped messages/tools/structured-output into the
Responses input / text.format / flattened-tool shapes at the provider
boundary; stored history and intra-turn tool round-trips need no
migration.
- Normalize the typed SSE event stream back into the existing handler
contract (content / thought / tool-call chunks) so the streaming
accumulator and OpenAI handler stay unchanged.
- Carry encrypted reasoning items across the in-turn tool loop
(provider-local, keyed by call_id) for stateless reasoning continuity.
- OPENAI_RESPONSES_STORE opt-in enables server-side state and
previous_response_id chaining across turns, persisted in message
metadata.
- Set gpt-5.5 to api_flavor: responses, reasoning_effort: medium.
- Remove AzureOpenAILLM and the azure_openai provider/catalog.
* feat: SSE notification system
Adds a per-user SSE pipe (GET /api/events) plus a per-message
chat-stream reconnect endpoint (GET /api/messages/<id>/events).
Backend substrate:
- application/events/ — durable journal (Redis Streams) + live
pub/sub for user-scoped events, with publish_user_event() as
the worker-side entrypoint.
- application/streaming/ — broadcast_channel for pub/sub fanout
and event_replay for the per-message snapshot+tail path.
- application/storage/db/repositories/message_events.py +
alembic 0007 — Postgres journal for chat-stream events.
- application/worker.py — ingest/reingest/remote/connector/
attachment/mcp_oauth tasks publish queued/progress/completed/
failed envelopes alongside their existing status updates.
Frontend client:
- frontend/src/events/ — connect/reconnect, Last-Event-ID cursor,
backoff with jitter. Each tab runs its own connection; no
cross-tab dedup (future work).
- frontend/src/notifications/ — recentEvents ring, cursor
tracking, tool-approval toast.
- frontend/src/upload/uploadSlice.ts — extraReducers for
source.ingest.* and attachment.* events.
Coverage: 132 SSE tests across events substrate, replay, journal,
routes, and worker publishes.
* refactor(attachments): remove polling, SSE-only
frontend/src/components/MessageInput.tsx no longer runs a 2s
setInterval against getTaskStatus for every processing
attachment. The attachment.* SSE reducers in uploadSlice.ts are
now the sole driver of attachment state transitions.
* feat(connector): consume source.ingest.* SSE, remove polling
frontend/src/components/ConnectorTree.tsx now mirrors FileTree's
slice-walking pattern: it watches notifications.recentEvents
for source.ingest.{completed,failed} envelopes matching the
sync's source id, and no longer polls /task_status every 2s.
* refactor(source-ingest): remove polling, SSE-only
frontend/src/upload/Upload.tsx and
frontend/src/components/FileTree.tsx no longer run getTaskStatus
polling fallbacks. The source.ingest.* SSE reducers in
uploadSlice.ts and FileTree's slice walk are now the sole
drivers of upload/reingest state transitions.
* refactor(mcp-oauth): carry authorization_url in SSE, remove polling
application/worker.py::mcp_oauth now publishes
authorization_url on the mcp.oauth.awaiting_redirect envelope.
frontend/src/modals/MCPServerModal.tsx consumes it from SSE
instead of polling /oauth_status/<task_id> every 1s.
The URL is generated inside DocsGPTOAuth.redirect_handler when
the FastMCP client triggers OAuth. The worker now plumbs a
publish callback through tool_config -> MCPTool -> DocsGPTOAuth
so the awaiting_redirect publish fires from inside the handler
at the exact point the URL becomes known. The legacy Redis
mcp_oauth_status setex writes and the GET
/api/mcp_server/oauth_status/<task_id> endpoint are kept as
belt-and-suspenders; nothing in the frontend reads them now.
* feat(source-ingest): plumb limited flag through SSE for token-cap UX
application/worker.py::ingest_worker and remote_worker now publish
``limited: bool`` on the source.ingest.completed envelope.
uploadSlice routes ``payload.limited === true`` to a failed status
with a ``tokenLimitReached`` flag, and UploadToast surfaces the
translated tokenLimit i18n string. No worker code path sets
limited=true today; this is a forward-looking contract so when
token-cap detection lands, the UX is already wired.
* refactor(mcp-oauth): read status from SSE journal, drop polling endpoint
MCPOAuthManager.get_oauth_status now walks the per-user SSE Streams
journal (user:{user_id}:stream) for the latest mcp.oauth.* envelope
matching the task id, returning the status string derived from the
event type suffix and the payload fields. The worker is the single
source of truth — its publish_user_event calls write the same
record the SSE client receives live.
Removed:
- /api/mcp_server/oauth_status/<task_id> route in
application/api/user/tools/mcp.py
- mcp_oauth_status worker function and mcp_oauth_status_task Celery
wrapper
- All mcp_oauth_status:{task_id} Redis setex writes (4 in mcp_oauth,
2 in DocsGPTOAuth.redirect_handler / callback_handler)
- The update_status closure in mcp_oauth that wrote the polling
payload
Tests updated:
- get_oauth_status now takes (task_id, user_id); new coverage walks
a fake xrevrange response for the completed envelope, the no-match
case, and a Redis-down case
- Removed TestMCPOAuthStatus route tests and TestMcpOauthStatusTask
celery-wrapper test
- Removed the two oauth_status methods from the integration runner
mcp_oauth:auth_url/state/code/error Redis keys remain — they are
the OAuth flow's own state (not the dropped polling payload).
* chore(mcp-oauth): delete orphaned getMCPOAuthStatus client
The /api/mcp_server/oauth_status/<task_id> endpoint was removed in
the prior commit; the corresponding userService method and the
MCP_OAUTH_STATUS endpoint constant had no remaining callers in the
frontend, so they're deleted along with it.
* fix(events): drop live publish when journal write fails
application/events/publisher.py returned an envelope to live
pubsub subscribers even when the XADD to the durable journal
failed. The envelope had no ``id`` field, which bypassed the SSE
route's dedup floor and broke ``Last-Event-ID`` semantics for any
reconnecting client.
Best-effort delivery means dropping consistently, not delivering
inconsistent state. Now: if the journal write fails the publisher
returns None and skips the live publish entirely.
* fix(notifications): dedupe sseEventReceived against immediate dupes
Snapshot replay + live tail can both deliver the same id when the
live pubsub frame and the replay XRANGE overlap. The route's own
dedup floor catches the common case, but consumers walking
``recentEvents`` (FileTree, ConnectorTree, MCPServerModal,
ToolApprovalToast) would otherwise act on the same envelope
twice when a duplicate slipped through.
Belt-and-suspenders: short-circuit when the most recent id in
the ring matches the incoming one.
* fix(events): skip replay budget INCR when no snapshot work possible
_allow_replay incremented the per-user counter on every
/api/events GET, including no-op connects from a fresh client
with no cursor against an empty backlog. React StrictMode dev
double-mounts plus a few tabs trivially tripped the default
30-per-60s budget on idle reconnects.
XLEN pre-check: when last_event_id is None and the user stream
is empty, the connect can't do snapshot work — return True
without INCR. Cursor-bearing connects still INCR unconditionally
(probing the cursor's relationship to stream contents would
require a redundant XRANGE).
* fix(streaming): tighten journal contract + recover from seq collisions
Two related fixes to application/streaming/message_journal.py.
1. record_event now rejects non-dict payloads at the gate. The
live path (base.py::_emit) wrapped non-dicts as
{"value": payload}; the replay path in event_replay synthesized
{"type": event_type}. A reconnecting client would receive a
different envelope than the one originally streamed. Now both
paths see byte-identical envelopes because non-dicts can't be
journaled at all. The corresponding event_replay fallback is
replaced with a warn-and-skip for any legacy rows.
2. record_event handles IntegrityError on (message_id, sequence_no)
collisions by reading latest_sequence_no and retrying once with
latest+1. The most likely cause is a stale seq seed on a
continuation retry where the route read MAX(seq) from a
separate connection before another writer committed past it.
Previously the error was swallowed and the event silently
dropped from the journal; now it lands at the next available
seq. The live pubsub publish uses the materialised seq so the
journal row and the live frame agree.
* perf(streaming): batch message_events INSERTs per stream
complete_stream previously opened a fresh db_session() per yielded
event, doing one Postgres INSERT + commit per chunk on the WSGI
thread. Streaming answers emit ~100s of answer chunks per response,
so the route was paying ~100 PG roundtrips per stream serialized on
commit latency.
New BatchedJournalWriter in application/streaming/message_journal.py
accumulates rows per stream and flushes on three triggers:
- size: buffer reaches 16 entries
- time: 100ms elapsed since the last flush
- lifecycle: close() at end-of-stream
Live pubsub publishes still fire synchronously per record(), so
subscribers see events in real time — only the durable journal write
is amortized. On bulk INSERT IntegrityError the writer falls back to
per-row record() with the existing seq+1 retry so a single colliding
seq doesn't drop the rest of the batch.
complete_stream wires journal_writer.close() into every exit path
(happy end, tool-approval-paused end, GeneratorExit, error handler)
so the terminal event is committed before the generator returns —
otherwise a reconnecting client could snapshot up to the last flush
boundary and live-tail waiting for an end that's still in memory.
Repository gets bulk_record() — one SQLAlchemy executemany INSERT
for the bulk path. All-or-nothing on collision (Postgres aborts the
whole batch); the writer's per-row fallback handles recovery.
* chore(upload): drop dead UploadTask.lastEventAt field
The lastEventAt field on UploadTask had no remaining consumers — the
matching Attachment.lastEventAt was cleaned up earlier. Remove the
field declaration and the slice write site.
* chore(frontend): drop orphaned getTaskStatus client
After the polling-removal sweep no caller in frontend/src/ references
userService.getTaskStatus or endpoints.USER.TASK_STATUS. The backend
route /api/task_status itself stays — agents, webhooks, e2e specs,
and the public docs still depend on it.
* docs(repo): remove stale planning docs from repo root
notification-channel-design.md, plan.md, and reminder-tool-design.md
were leftover Claude planning artifacts from the SSE substrate work
that landed accidentally. CLAUDE.md prohibits creating planning docs
unless asked — delete them.
* docs(message-events): clarify repo vs wrapper payload contract
MessageEventsRepository.record accepts any JSONB-compatible value; the
streaming wrapper record_event tightens this to dicts only because the
live and replay paths reconstruct non-dict payloads differently. Spell
the split out so the next reader of the repo method doesn't assume the
wrapper's contract applies here.
* refactor(events): raise on malformed stream id instead of lex fallback
stream_id_compare's lex-fallback branch was a footgun: a malformed id
that sorts lex-greater than a real one would pin live-tail dedup
forever, dropping every subsequent legitimate event silently. Both
current callers in application/api/events/routes.py pre-validate
inputs against _STREAM_ID_RE before calling, so changing the function
to raise ValueError is a no-op on the happy path and turns the future-
caller footgun into a loud failure.
* test(tasks): cover cleanup_message_events task body
Adds skipped-when-no-POSTGRES_URI and happy-path coverage for the
Celery janitor. The skipped path returns the documented short-circuit
shape without touching the repo. The happy path seeds a backdated
row, runs the task against the pg_conn fixture, and asserts the
retention window's row is deleted while in-window rows survive.
Mirrors the TestCleanupPendingToolState pattern.
* fix(notifications): treat /c/new as no current conversation
useMatch('/c/:conversationId') treats the literal URL /c/new as a
real conversation id, so the toast suppression check confused
'user is on /c/new' with 'user is on the conversation needing
approval'. Explicit guard: when the matched id is 'new', fall
through to the no-match case so approval toasts still surface.
* docs(events): enumerate publish_user_event None-return paths
The function returns Optional[str] today, with None conflating five
distinct outcomes (missing args / push disabled / unserialisable /
Redis down / XADD failed). Every current call site is fire-and-
forget and ignores the return, so the right move is to document the
five cases rather than promote to an enum return — keeps the API
small while making the diagnostic surface (logs) obvious. If a
future caller needs to react differently per reason, promote then.
* refactor(sources): move source-id derivation out of worker module
application/api/user/sources/upload.py imported _derive_source_id
from application.worker — pulling the entire Celery worker module
into the API process at import time just for a two-line helper.
Move DOCSGPT_INGEST_NAMESPACE and the derivation function to a
new application/storage/db/source_ids.py module that both layers
can import without that dependency edge. worker.py re-exports the
old names (_derive_source_id, DOCSGPT_INGEST_NAMESPACE) for
backward-compatible imports from tests and any other in-tree
callers; new code should import from the new module directly.
* fix(cache): enable Redis health_check_interval to surface half-open TCP
Without health_check_interval, a half-open TCP socket (NAT silently
dropped state, ELB idle-close) can leave pubsub.get_message hanging
past the SSE generator's keepalive cadence — the kernel never
surfaces the dead socket because no payload is in flight. Setting
health_check_interval=10 makes redis-py ping every 10s when
otherwise idle, so the next get_message after the dead window
raises and the SSE loop falls into its reconnect path instead of
silently freezing on the user.
* chore(events): rename attachment.processing.progress to attachment.progress
The event-type taxonomy was inconsistent: source ingest emits
source.ingest.progress (three segments) while attachments emitted
attachment.processing.progress (four segments). Drops the
.processing. infix for parity. Worker publish sites, the slice
reducer's match, and the worker tests all flip together.
No external consumers — the event type is purely internal between
the publisher and the in-tab slice; safe to rename in one commit.
* feat: events cleanup
* fix: better docs
* fix: e2e tests
Pavel