mirror of
https://github.com/bytedance/deer-flow.git
synced 2026-10-28 06:00:38 +00:00
b99ac916b6
* fix(auth): count login lockouts in the database across Gateway replicas POST /api/v1/auth/login/local throttled failed logins per client IP in a module-level dict, so with N Gateway replicas behind one load balancer an attacker got N x auth.local.max_login_attempts guesses and a lockout on one replica was invisible to the others. Move the counter behind a LoginThrottleStore (deerflow.persistence. login_throttle): MemoryLoginThrottleStore keeps the historical per-process semantics and 10k tracked-IP bound; SqlLoginThrottleStore shares one login_throttle row per IP through the application database, incrementing and starting the lock in a single dialect-native upsert, applying check decisions with compare-and-set predicates in a separate write transaction, and sweeping served locks plus 24h-idle counters in bounded batches. Both stores pin the existing contract: the lock starts when the count reaches max_login_attempts, the committed duration is honored when the policy shrinks mid-lock, a served sentence is never resurrected, tightening keeps the count, success resets the IP, and the 429 status and message are unchanged. New startup-only selector auth.local.throttle_storage (auto | memory | db, default auto = the database whenever database.backend is sqlite or postgres). The Gateway installs the store once at startup after the persistence engine; deps._validate_login_throttle_storage warns when a multi-process deployment keeps memory. Migration 0034_login_throttle creates the table; config_version 55 -> 56 in config.example.yaml and the Helm chart. Docs: README, CONFIGURATION.md, AUTH_DESIGN.md, config and migrations AGENTS.md, CHANGELOG. * docs(changelog): link the login throttle entry to #6501 * fix(auth): re-check expiry on the row the login-throttle sweep deletes The bounded cleanup in SqlLoginThrottleStore.record_failure deleted `WHERE ip IN (SELECT ip ... WHERE <served or stale> LIMIT 200)`. On PostgreSQL READ COMMITTED the expiry predicate applied only inside the snapshot-bound subquery: when the DELETE had to wait on a row lock held by a concurrent check() that was extending the same lock under a raised lockout_seconds, PostgreSQL re-evaluated only the outer WHERE on the updated tuple, IP membership still held for the old candidate, and the now-active lock was deleted; the next attempt from that IP got a clean record and reached password verification (review of #6501, P2). Repeat the served/stale predicate on the DELETE target as well (SqlLoginThrottleStore.sweep_statement), keeping the bounded candidate subquery, so the re-evaluation sees the updated row's expiry. The other writes already carry their full compare-and-set predicate on the target row: _discard and the check() UPDATE both use _matches (ip, fail_count, locked_at, lock_duration_seconds), so a tuple changed while they waited no longer matches and the stale decision is skipped. _discard keeps the equality compare-and-set rather than adding an expiry condition: its early-release branch deletes a lock whose committed sentence has not elapsed under the lowered live policy, which an expiry predicate would wrongly block, and the equality predicate already pins every input of the decision. reset stays unconditional by design. Tests: a dialect-compile test asserts the DELETE's own WHERE carries the expiry predicate for postgresql and sqlite (red on the previous HEAD), and a PostgreSQL-only test (skipped without TEST_POSTGRES_URI) reproduces the update-versus-sweep interleaving with two engines, waiting on pg_stat_activity for the sweep to block before the extender commits. Docs: sql.py concurrency contract and the migrations guide's 0034 bullet. * fix(auth): keep an active login lock's clock and fail closed on explicit db Address the "changes requested" review on #6501. Lock clock (confirmed defect): record_failure re-stamped locked_at = now for every failure on an already-locked row, so a sustained attacker's sentence slid forward with each attempt. Both stores now apply one rule in one atomic decision: new_count = fail_count + 1; below max_attempts the IP counts with no lock (a failure after the operator raised the threshold clears an existing lock); at or over it, a failure during an *active* lock keeps locked_at and the committed lock_duration unchanged, while a never-locked or served row starts anew with now / lockout_seconds (so the failure that reaches a raised threshold starts a fresh lock rather than keeping a stale locked_at). The SQL upsert expresses it as nested case() on the row's own values; the memory store mirrors it. The sweeps in both stores now leave the row being recorded to that decision instead of deleting a served lock or stale counter first, which split the contract (SQL would have produced a fresh (1, NULL, NULL) row). base.py states the rule. Explicit db downgrade: resolve_login_throttle_store raises SystemExit — the failure type of the other startup validations in app.gateway.deps — when throttle_storage='db' resolves to the shared table but no engine is initialised, so langgraph_runtime refuses to start instead of silently running every replica on per-process counters; the lazy bare-app path applies the same rule and installs nothing. 'auto' keeps warning and falling back. 'db' on a memory database still degrades with a warning (there is no shared table to insist on). Non-blocking items: _key() logs a warning when a client key is truncated to 255 characters (never the key itself); the sql.py docstring covers wall-clock locked_at across replicas (run NTP; skew shifts evaluation by that much) and why STALE_COUNTER_SECONDS=24h / SWEEP_BATCH_SIZE=200 were chosen; a store-level blocking-I/O anchor covers the SQL reset path; the _validate_login_throttle_storage warning was verified to render 'throttle_storage=memory' for the StrEnum (the gate test now passes the enum member and asserts that text). reset stays unconditional: a peer failure that commits after a reset leaves fail_count=1, one genuine failure after the cleared history, which a new test documents. Docs: CHANGELOG entry, config.example.yaml, CONFIGURATION.md, AUTH_DESIGN.md and the throttle_storage field description. * fix(auth): re-evaluate a login-throttle check whose compare-and-set missed SqlLoginThrottleStore.check applied its decision (commit the live duration, or clear a served / released lock) with a compare-and-set predicate on the snapshot it had read, but ignored the affected-row count. When a peer changed the row between the read and the write — an admitted wrong password incrementing fail_count on another replica, or a successful login clearing the row — the UPDATE or DELETE matched zero rows, yet check reported the decision as applied: it could return an extended 60s sentence while the row kept its 1s duration, so the next check cleared the lock and let password verification through; a live duration decrease could be dropped the same way; and a row a peer had just cleared was still reported as locked (#6501 review, P2). Every compare-and-set write in check now checks result.rowcount. On a miss the row is re-read and re-evaluated from the top, up to CHECK_CAS_ATTEMPTS (3) times, applying a fresh compare-and-set decision on the new snapshot and never clobbering a reset or replacement: a row that is gone is reported as allowed, a fresh peer lock is reported as locked, and an incremented count gets the extension or decrease committed on it. Exhausting the bound fails closed: nothing is written and the answer is the last snapshot's lock under the live policy, logged at debug. The read stays in its own short transaction and each write in a separate one. The evaluation is now one shared pure function, evaluate_check in base.py (CheckDecision), used by both stores so the contract cannot diverge. The memory store keeps a single pass: its compare-and-set cannot miss because there is no await between the read and the write on the loop thread, which its docstring now states. Tests (red first on df26c2acf) use a session proxy that runs the peer mutation on a second engine right after the snapshot read ends its transaction: the reviewer's raised-duration interleaving (59.5s reported and committed, the increment survives, a later check still locked), the mirrored decrease, early release on the fresh row, a reset landing between read and write (0.0, no row recreated), and the bound-exhausted fail-closed case. * docs(changelog): resolve merge markers left in the Unreleased section * test(migrations): pin 0035_login_throttle as the head in the 0034 watermark test Upstream #6477 added test_migration_0034_run_event_seq_watermark.py asserting 0034 is the single head; this branch chains 0035_login_throttle after it, so the test now pins the current head and walks the chain like its siblings. * perf(auth): one session per login-throttle check, one transaction per failure Two non-blocking performance suggestions from the #6501 review. Single session per check. The router probed the row with store.get() and then called store.check(), which re-read it: a recorded IP cost two connection-pool checkouts and two SELECTs per request on the unauthenticated login path, letting a lockout flood compete with legitimate traffic for Postgres connections. LoginThrottleStore.check now takes the policy as an async callable (LoginThrottlePolicy) and resolves it lazily inside the store: the probe, the policy resolution and the compare-and-set loop share one session, a clean IP returns 0.0 without ever calling the policy (the config.yaml read is still skipped for the common case), and a recorded IP resolves it exactly once, also across compare-and-set retries. The decision is made on a snapshot read after the resolution: the policy read is a yield point, and a decision that needs no write has no compare-and-set to catch a row a peer changed meanwhile — so one session, but still a read after the yield. The memory store takes the same signature and re-reads its record after the await for the same reason. The router passes its asyncio.to_thread policy read as the callable and drops the separate probe; get stays on the protocol for tests and anchors. No short-TTL verdict cache: it would mask a successful login's reset for the TTL window, and the single session already halves the hot path. Sweep in the upsert's transaction, plus an index. record_failure ran the housekeeping DELETE in its own transaction and the upsert in a second one, two commits (two WAL flushes on PostgreSQL) per failed login; the sweep now rides in the upsert's transaction with one commit, the keep exclusion already protecting the row being recorded. The new ix_login_throttle_updated_at (ORM __table_args__ and the unreleased 0035 migration, created after the table and dropped before it, each guarded separately so a table left by an interrupted attempt still gets its index) serves the sweep's stale-counter predicate; the served-lock predicate scans a table the sweep itself keeps small. Tests (red first on a137fbc47): a clean IP never resolves the policy and a recorded IP resolves it exactly once (both stores), the SQL store opens exactly one session per check and commits once per record_failure (counting wrappers), the migration test asserts the index after upgrade, after completing a partial upgrade that lacked it, and gone after downgrade while compare_metadata stays empty; the router-level "mutated during policy resolution" tests now inject their mutation through the policy callable. * docs(changelog): note config_version 57 in the login throttle entry
1194 lines
56 KiB
Python
1194 lines
56 KiB
Python
"""Centralized accessors for singleton objects stored on ``app.state``.
|
||
|
||
**Getters** (used by routers): raise 503 when a required dependency is
|
||
missing, except ``get_store`` which returns ``None``.
|
||
|
||
``AppConfig`` is intentionally *not* cached on ``app.state``. Routers and the
|
||
run path resolve it through :func:`deerflow.config.app_config.get_app_config`,
|
||
which performs mtime-based hot reload, so edits to ``config.yaml`` take
|
||
effect on the next request without a process restart. The engines created in
|
||
:func:`langgraph_runtime` (stream bridge, persistence, checkpointer, store,
|
||
run-event store) accept a ``startup_config`` snapshot — they are
|
||
restart-required by design and stay bound to that snapshot to keep the live
|
||
process consistent with itself.
|
||
|
||
Initialization is handled directly in ``app.py`` via :class:`AsyncExitStack`.
|
||
"""
|
||
|
||
from __future__ import annotations
|
||
|
||
import asyncio
|
||
import logging
|
||
import os
|
||
from collections.abc import AsyncGenerator, Callable
|
||
from contextlib import AsyncExitStack, asynccontextmanager
|
||
from pathlib import Path
|
||
from typing import TYPE_CHECKING, Any, TypeVar, cast
|
||
|
||
from fastapi import FastAPI, HTTPException, Request
|
||
from langgraph.types import Checkpointer
|
||
|
||
from deerflow.community.browser_automation.session import browser_multi_worker_error
|
||
from deerflow.config.app_config import AppConfig, get_app_config
|
||
from deerflow.config.deployment_config import multi_instance_declaration
|
||
from deerflow.persistence.feedback import FeedbackRepository
|
||
from deerflow.runtime import ORPHAN_RECOVERY_STOP_REASON, STARTUP_ORPHAN_RECOVERY_ERROR, RunContext, RunManager, StreamBridge
|
||
from deerflow.runtime.events.store.base import RunEventStore
|
||
from deerflow.runtime.runs.store.base import RunStore
|
||
from deerflow.utils.file_io import await_drained
|
||
|
||
logger = logging.getLogger(__name__)
|
||
|
||
# Upper bound (seconds) for draining in-flight runs during shutdown, before the
|
||
# AsyncExitStack tears down the checkpointer (and its connection pool). Kept
|
||
# local to avoid an app -> deps -> app import cycle. This is a *separate* budget
|
||
# from ``app.gateway.app._SHUTDOWN_HOOK_TIMEOUT_SECONDS`` (currently also 5.0s,
|
||
# which bounds channel-service stop): the two govern independent teardown steps
|
||
# and may diverge, but both count toward the lifespan shutdown window — revisit
|
||
# them together if their sum must stay within the server's graceful-shutdown
|
||
# timeout.
|
||
_RUN_DRAIN_TIMEOUT_SECONDS = 5.0
|
||
|
||
|
||
def _browser_tools_enabled_in_config(config: AppConfig) -> bool:
|
||
"""Return whether process-local agentic browser sessions are configured."""
|
||
get_tool_config = getattr(config, "get_tool_config", None)
|
||
if callable(get_tool_config):
|
||
return get_tool_config("browser_navigate") is not None
|
||
return any(getattr(tool, "name", None) == "browser_navigate" for tool in (getattr(config, "tools", None) or []))
|
||
|
||
|
||
# ``GATEWAY_WORKERS`` is the knob this project documents and the one
|
||
# ``docker/docker-compose.yaml`` forwards as ``--workers``. ``backend/Dockerfile``
|
||
# and ``scripts/serve.sh`` launch uvicorn with no worker count at all, and uvicorn
|
||
# then takes the count from ``WEB_CONCURRENCY``. Both must be read, otherwise a
|
||
# multi-process deployment starts with the safety gates inert.
|
||
# A blank value means "unset", as the compose command's ``${GATEWAY_WORKERS:-1}`` treats it.
|
||
_WORKER_COUNT_ENV_VARS = ("GATEWAY_WORKERS", "WEB_CONCURRENCY")
|
||
|
||
|
||
def _gateway_worker_count() -> tuple[int, str]:
|
||
"""Return the Gateway worker-process count and the environment variable that set it.
|
||
|
||
The name belongs in the refusal text: an operator who only ever set
|
||
``WEB_CONCURRENCY`` cannot act on a message about ``GATEWAY_WORKERS``.
|
||
"""
|
||
for name in _WORKER_COUNT_ENV_VARS:
|
||
raw = os.environ.get(name)
|
||
if not raw or not raw.strip():
|
||
continue
|
||
try:
|
||
return int(raw), name
|
||
except (TypeError, ValueError):
|
||
# Uvicorn rejects a non-numeric count itself, so this is not the place
|
||
# to fail the launch -- but an unparsable documented knob must not hide
|
||
# the count uvicorn takes from ``WEB_CONCURRENCY`` on the launches that
|
||
# pass no ``--workers``. Keep looking; report 1 only after the loop.
|
||
continue
|
||
return 1, _WORKER_COUNT_ENV_VARS[0]
|
||
|
||
|
||
def _multi_process_signal(config: AppConfig) -> tuple[str, str] | None:
|
||
"""Return ``(reason, rollback)`` when this process must assume peer Gateway processes.
|
||
|
||
``reason`` names the knob that established the topology (``GATEWAY_WORKERS=2``,
|
||
``DEER_FLOW_MULTI_INSTANCE=1`` or ``deployment.multi_instance=true``) and
|
||
``rollback`` is the single-instance remediation a refusal message offers.
|
||
When the worker count and a declaration are both active, ``rollback``
|
||
withdraws both: resetting only the worker count would bounce the operator
|
||
through a second refusal on the declaration at the next start.
|
||
|
||
The worker-count variables only see one process tree, so a Kubernetes
|
||
Deployment with ``replicas > 1`` and one worker per Pod is invisible to
|
||
them; the explicit declaration exists for exactly that topology.
|
||
"""
|
||
workers, worker_env = _gateway_worker_count()
|
||
declaration = multi_instance_declaration(config)
|
||
if workers > 1:
|
||
rollback = f"Set {worker_env}=1"
|
||
if declaration is not None:
|
||
rollback = f"{rollback} and {declaration.rollback}"
|
||
return f"{worker_env}={workers}", rollback
|
||
if declaration is None:
|
||
return None
|
||
step = declaration.rollback
|
||
return declaration.knob, f"{step[0].upper()}{step[1:]} and run a single Gateway instance"
|
||
|
||
|
||
def _stream_bridge_is_cross_process(config: AppConfig) -> bool:
|
||
"""Return whether live run events can reach SSE clients on every instance.
|
||
|
||
Mirrors ``runtime/stream_bridge/async_provider.py::_resolve_config``: the
|
||
config.yaml section wins, and an omitted section falls back to the
|
||
``DEER_FLOW_STREAM_BRIDGE_REDIS_URL`` variable the Docker and Helm
|
||
deployments inject.
|
||
"""
|
||
bridge = getattr(config, "stream_bridge", None)
|
||
if bridge is not None:
|
||
return getattr(bridge, "type", None) == "redis"
|
||
return bool((os.environ.get("DEER_FLOW_STREAM_BRIDGE_REDIS_URL") or "").strip())
|
||
|
||
|
||
def _enforce_postgres_for_multi_worker(config: AppConfig) -> None:
|
||
"""Refuse unsafe multi-process configurations before persistence starts.
|
||
|
||
A deployment counts as multi-process when ``GATEWAY_WORKERS`` /
|
||
``WEB_CONCURRENCY`` is above 1, or when the operator declares peers with
|
||
``deployment.multi_instance: true`` / ``DEER_FLOW_MULTI_INSTANCE=1`` (the
|
||
worker count cannot see other Pods, and a Pod that starts next to a peer
|
||
without these prerequisites writes the peer's live runs off as orphans).
|
||
Multi-instance scheduler recovery also needs the durable run ownership
|
||
contract even when each Pod runs a single Gateway worker.
|
||
|
||
1. The background scheduler must be disabled for ordinary multi-worker
|
||
mode. ``scheduler.multi_instance`` opts into the lease-aware path.
|
||
2. Process-local browser sessions must be disabled. Browser tools keep
|
||
Chromium and Playwright objects in one worker's memory, while ordinary
|
||
uvicorn dispatch provides no thread-id affinity.
|
||
3. The DB backend must be Postgres — SQLite write-locks cannot support
|
||
concurrent multi-process access.
|
||
4. ``run_events.backend`` must be ``db``. Memory and JSONL stores are
|
||
process-local, so workers cannot enforce a shared singleton receipt.
|
||
5. ``run_ownership.heartbeat_enabled`` must be True — without heartbeat,
|
||
every run has a NULL lease, so reconciliation treats all inflight
|
||
runs as orphans and Worker B would kill Worker A's live runs on
|
||
every rolling update or scale-up.
|
||
6. The stream bridge must be Redis. The memory bridge is process-local,
|
||
so SSE streams, reconnects and ``/wait`` only work on the owner.
|
||
7. ``sandbox.ownership.type`` must not be an explicit ``memory``: an
|
||
in-process ownership store cannot see peers, so reconciliation would
|
||
adopt and idle-destroy another instance's live containers (#4206).
|
||
|
||
This gate runs once at startup before any persistence engine is
|
||
initialised so the error message is clear and the process exits
|
||
immediately.
|
||
"""
|
||
scheduler = getattr(config, "scheduler", None)
|
||
multi_instance_requested = bool(getattr(scheduler, "multi_instance", False))
|
||
multi_instance_scheduler = bool(getattr(scheduler, "enabled", False) and multi_instance_requested)
|
||
|
||
backend = getattr(config.database, "backend", None)
|
||
run_events_backend = getattr(getattr(config, "run_events", None), "backend", None)
|
||
run_ownership = getattr(config, "run_ownership", None)
|
||
|
||
if multi_instance_requested and backend != "postgres":
|
||
raise SystemExit(f"scheduler.multi_instance=true requires database.backend='postgres'. database.backend is '{backend}'. Set scheduler.multi_instance=false or configure Postgres.")
|
||
if multi_instance_requested and run_events_backend != "db":
|
||
raise SystemExit(f"scheduler.multi_instance=true requires run_events.backend='db'. run_events.backend is '{run_events_backend}'. Set scheduler.multi_instance=false or configure run_events.backend: db.")
|
||
if multi_instance_requested and (run_ownership is None or not run_ownership.heartbeat_enabled):
|
||
raise SystemExit("scheduler.multi_instance=true requires run_ownership.heartbeat_enabled=true so peer runs retain a valid lease. Set scheduler.multi_instance=false or enable run ownership heartbeats.")
|
||
|
||
signal = _multi_process_signal(config)
|
||
if signal is None:
|
||
return
|
||
reason, rollback = signal
|
||
|
||
if config.scheduler.enabled and not multi_instance_scheduler:
|
||
raise SystemExit(f"{reason} cannot run with scheduler.enabled=true because each worker starts its own scheduler. {rollback}, scheduler.multi_instance=true, or scheduler.enabled=false.")
|
||
|
||
if _browser_tools_enabled_in_config(config):
|
||
workers, _worker_env = _gateway_worker_count()
|
||
if workers > 1 and multi_instance_declaration(config) is None:
|
||
raise SystemExit(browser_multi_worker_error(workers))
|
||
raise SystemExit(f"{reason} cannot enable agentic browser tools: browser sessions are process-local and a request can land on any instance. {rollback} or disable the browser_navigate tool.")
|
||
|
||
if backend != "postgres":
|
||
raise SystemExit(f"{reason} requires database.backend='postgres', but database.backend is '{backend}'. SQLite cannot support concurrent multi-process access. {rollback} or switch to Postgres.")
|
||
|
||
if run_events_backend != "db":
|
||
raise SystemExit(
|
||
f"{reason} requires run_events.backend='db', but run_events.backend is '{run_events_backend}'. "
|
||
"Memory and JSONL event stores are process-local, so delivery receipt singleton guarantees cannot hold across workers. "
|
||
f"{rollback} or configure run_events.backend: db."
|
||
)
|
||
|
||
if run_ownership is None or not run_ownership.heartbeat_enabled:
|
||
raise SystemExit(
|
||
f"{reason} requires run_ownership.heartbeat_enabled=true. "
|
||
"Without heartbeat, every run has a NULL lease, so reconciliation "
|
||
"treats all inflight runs as orphans — a starting instance would kill its "
|
||
"peers' live runs on every rolling update or scale-up. "
|
||
"Set run_ownership.heartbeat_enabled=true in config.yaml."
|
||
)
|
||
|
||
if not _stream_bridge_is_cross_process(config):
|
||
raise SystemExit(
|
||
f"{reason} requires stream_bridge.type='redis' (or DEER_FLOW_STREAM_BRIDGE_REDIS_URL), "
|
||
"but the stream bridge is the process-local memory bridge: SSE streams, reconnects and /wait "
|
||
f"only work on the instance that owns the run. {rollback} or configure the redis stream bridge."
|
||
)
|
||
|
||
ownership = getattr(getattr(config, "sandbox", None), "ownership", None)
|
||
if ownership is not None and getattr(ownership, "type", None) == "memory":
|
||
raise SystemExit(
|
||
f"{reason} cannot run with sandbox.ownership.type='memory': an in-process ownership store cannot see "
|
||
"peer instances, so startup reconciliation would adopt and idle-destroy containers another instance is "
|
||
"using (#4206). Set sandbox.ownership.type='redis', or omit the section so it is inferred from the redis stream bridge."
|
||
)
|
||
|
||
|
||
def _validate_agent_storage(config: AppConfig) -> None:
|
||
"""Fail fast on an agent-storage backend the database cannot support.
|
||
|
||
``agent_storage.backend: db`` needs a durable, shared SQL database — a
|
||
``memory`` database is per-process, so custom-agent and managed-subagent
|
||
definitions would silently diverge across nodes (and there is no SQL URL
|
||
to open). Mirrors deermem's create_storage fail-fast and the multi-worker
|
||
gate above.
|
||
|
||
Also warns when a multi-worker Postgres deployment leaves agent storage on
|
||
``file``: custom agents created on one node's local disk are invisible to
|
||
the others, exactly the divergence the db backend exists to fix.
|
||
"""
|
||
agent_storage = getattr(config, "agent_storage", None)
|
||
backend = getattr(agent_storage, "backend", "file")
|
||
db_backend = getattr(getattr(config, "database", None), "backend", None)
|
||
if backend == "db" and db_backend not in ("sqlite", "postgres"):
|
||
raise SystemExit(
|
||
f"agent_storage.backend='db' requires database.backend to be 'sqlite' or 'postgres', "
|
||
f"but database.backend is '{db_backend}'. A 'memory' database is per-process and cannot "
|
||
"share agent definitions across nodes. Set database.backend, or use agent_storage.backend='file'."
|
||
)
|
||
signal = _multi_process_signal(config)
|
||
if signal is not None and db_backend == "postgres" and backend == "file":
|
||
logger.warning(
|
||
"%s with database.backend='postgres' but agent_storage.backend='file': "
|
||
"custom agents and managed subagents are stored per-node on local disk and are not visible "
|
||
"across workers/nodes. Set agent_storage.backend='db' to share them.",
|
||
signal[0],
|
||
)
|
||
|
||
|
||
def _validate_login_throttle_storage(config: AppConfig) -> None:
|
||
"""Warn when a multi-process deployment counts login failures per process.
|
||
|
||
``auth.local.throttle_storage`` resolves to the shared ``login_throttle``
|
||
table whenever an application database exists, so under the multi-process
|
||
gate (which already requires Postgres) only an explicit ``memory`` lands
|
||
here. That is not fatal — the throttle still works on every replica — but
|
||
with N replicas behind one load balancer an attacker gets N x
|
||
``max_login_attempts`` guesses and a lockout on one replica is invisible
|
||
to the others, exactly the gap the shared table closes. Mirrors the
|
||
``agent_storage.backend='file'`` divergence warning above.
|
||
"""
|
||
signal = _multi_process_signal(config)
|
||
if signal is None:
|
||
return
|
||
local = getattr(getattr(config, "auth", None), "local", None)
|
||
if local is None:
|
||
return
|
||
from deerflow.config.auth_config import LocalAuthConfig, resolve_login_throttle_storage
|
||
|
||
selector = getattr(local, "throttle_storage", LocalAuthConfig.model_fields["throttle_storage"].default)
|
||
db_backend = getattr(getattr(config, "database", None), "backend", None)
|
||
if resolve_login_throttle_storage(selector, db_backend) == "memory":
|
||
logger.warning(
|
||
"%s with auth.local.throttle_storage=%s: failed-login counters and lockouts are kept per Gateway process, "
|
||
"so an attacker behind the load balancer gets N x max_login_attempts guesses and a lockout on one replica "
|
||
"is invisible to the others. Set auth.local.throttle_storage='auto' (or 'db') so the shared login_throttle "
|
||
"table in the application database enforces one limit per IP.",
|
||
signal[0],
|
||
str(getattr(selector, "value", selector)),
|
||
)
|
||
|
||
|
||
def _validate_memory_retrieval_index(config: AppConfig) -> None:
|
||
"""Warn when a declared multi-instance deployment keeps DeerMem's retrieval index on the shared memory volume.
|
||
|
||
DeerMem's derived FTS5 index is one SQLite database in WAL mode. When
|
||
``storage_path`` sits on the home volume several Gateway instances share,
|
||
the default ``{storage_path}/.retrieval`` makes every instance open that
|
||
same file over a network filesystem (where SQLite documents WAL as
|
||
unsupported), empty and refill it under its peers at startup, and delete
|
||
it from under them on corruption recovery. The index is rebuildable, so
|
||
each instance should keep its own copy on local disk through
|
||
``memory.backend_config.retrieval_index_path``. Only the explicit
|
||
declaration counts: uvicorn workers of one process tree share local disk,
|
||
where a shared WAL index is supported. Mirrors ``_validate_agent_storage``:
|
||
a warning, not a refusal, because memory still works, only slower.
|
||
"""
|
||
declaration = multi_instance_declaration(config)
|
||
if declaration is None:
|
||
return
|
||
memory = getattr(config, "memory", None)
|
||
if memory is None or not getattr(memory, "enabled", False) or getattr(memory, "manager_class", "deermem") != "deermem":
|
||
return
|
||
backend_config = dict(getattr(memory, "backend_config", None) or {})
|
||
if backend_config.get("retrieval_adapter", "fts5") != "fts5":
|
||
return # disabled, or a custom RetrievalPort factory that owns its own storage
|
||
from deerflow.agents.memory.backends.deermem.deermem.core.paths import retrieval_index_directory
|
||
from deerflow.agents.memory.manager import resolve_deermem_storage_path
|
||
|
||
storage_path = resolve_deermem_storage_path(backend_config)
|
||
index_dir = retrieval_index_directory(storage_path, backend_config.get("retrieval_index_path"))
|
||
if index_dir is None or not Path(index_dir).resolve().is_relative_to(Path(storage_path).resolve()):
|
||
return
|
||
logger.warning(
|
||
"%s but the DeerMem retrieval index at %s is inside memory storage_path %s: every Gateway instance opens the same "
|
||
"SQLite WAL index over the shared memory volume, rebuilds it under its peers at startup and deletes it from under them "
|
||
"on corruption recovery. Set memory.backend_config.retrieval_index_path to an instance-local directory (the Helm chart "
|
||
"mounts an emptyDir at /var/lib/deerflow/memory-index).",
|
||
declaration.knob,
|
||
index_dir,
|
||
storage_path,
|
||
)
|
||
|
||
|
||
async def _drain_inflight_runs(run_manager: RunManager) -> None:
|
||
"""Drain in-flight runs before the checkpointer is torn down (issue #3373).
|
||
|
||
Shields the (internally-bounded) drain so that even if the lifespan
|
||
coroutine is repeatedly cancelled mid-shutdown — e.g. signal escalation or
|
||
the server's graceful-shutdown timeout — the checkpointer pool is not closed
|
||
while run tasks are still writing checkpoints. Cancellation is remembered
|
||
and propagated only after the already-running drain reaches a safe terminal
|
||
point.
|
||
"""
|
||
drain = asyncio.create_task(run_manager.shutdown(timeout=_RUN_DRAIN_TIMEOUT_SECONDS))
|
||
cancellation: asyncio.CancelledError | None = None
|
||
|
||
while not drain.done():
|
||
try:
|
||
await asyncio.shield(drain)
|
||
except asyncio.CancelledError as exc:
|
||
if cancellation is None:
|
||
cancellation = exc
|
||
except Exception:
|
||
if cancellation is not None:
|
||
logger.exception("In-flight run drain failed after shutdown cancellation")
|
||
raise cancellation
|
||
logger.exception("Failed to drain in-flight runs during shutdown")
|
||
return
|
||
|
||
if cancellation is not None:
|
||
try:
|
||
drain.result()
|
||
except Exception:
|
||
logger.exception("In-flight run drain failed after shutdown cancellation")
|
||
raise cancellation
|
||
|
||
try:
|
||
drain.result()
|
||
except Exception:
|
||
logger.exception("Failed to drain in-flight runs during shutdown")
|
||
|
||
|
||
async def _publish_recovered_run_stream_end(
|
||
bridge: StreamBridge,
|
||
recovered_runs: list[RunRecord],
|
||
*,
|
||
cleanup_delay: float = 60.0,
|
||
on_cleanup_scheduled: Callable[[str, asyncio.Task[None]], None] | None = None,
|
||
) -> list[tuple[str, asyncio.Task[None]]]:
|
||
"""Terminate retained streams for runs recovered as orphaned."""
|
||
cleanup_tasks: list[tuple[str, asyncio.Task[None]]] = []
|
||
for record in recovered_runs:
|
||
stream_exists = getattr(bridge, "stream_exists", None)
|
||
if stream_exists is not None:
|
||
try:
|
||
if not await stream_exists(record.run_id):
|
||
logger.debug("Skipping recovered stream end for %s: stream already expired", record.run_id)
|
||
continue
|
||
except Exception:
|
||
logger.debug("Failed to check recovered stream existence for %s", record.run_id, exc_info=True)
|
||
try:
|
||
await bridge.publish_end(record.run_id)
|
||
except Exception:
|
||
logger.warning(
|
||
"Failed to publish recovered run stream end for %s",
|
||
record.run_id,
|
||
exc_info=True,
|
||
)
|
||
continue
|
||
task = asyncio.create_task(bridge.cleanup(record.run_id, delay=cleanup_delay))
|
||
task.add_done_callback(lambda task, run_id=record.run_id: _log_recovered_stream_cleanup_result(task, run_id))
|
||
cleanup_tasks.append((record.run_id, task))
|
||
if on_cleanup_scheduled is not None:
|
||
on_cleanup_scheduled(record.run_id, task)
|
||
return cleanup_tasks
|
||
|
||
|
||
def _log_recovered_stream_cleanup_result(task: asyncio.Task[None], run_id: str) -> None:
|
||
if task.cancelled():
|
||
return
|
||
try:
|
||
task.result()
|
||
except Exception:
|
||
logger.warning("Failed to clean up recovered run stream for %s", run_id, exc_info=True)
|
||
|
||
|
||
async def _cleanup_recovered_scheduled_goals(recovered_runs: list[RunRecord], *, run_manager: RunManager, checkpointer: Checkpointer) -> None:
|
||
"""Clear only confirmed occurrence-owned goals after durable orphan recovery."""
|
||
from deerflow.runtime.runs.worker import clear_recovered_scheduled_goal
|
||
|
||
for record in recovered_runs:
|
||
metadata = getattr(record, "metadata", None) or {}
|
||
if not isinstance(metadata.get("scheduled_goal_objective"), str):
|
||
continue
|
||
try:
|
||
await clear_recovered_scheduled_goal(record, run_manager=run_manager, checkpointer=checkpointer)
|
||
except Exception:
|
||
logger.warning("Scheduled goal cleanup failed for recovered run %s; retained for guarded next-run cleanup", record.run_id, exc_info=True)
|
||
|
||
|
||
async def _flush_recovered_stream_cleanups(
|
||
bridge: StreamBridge,
|
||
cleanup_tasks: dict[asyncio.Task[None], str],
|
||
*,
|
||
timeout: float = 1.0,
|
||
) -> None:
|
||
"""Cancel delayed cleanups and delete their streams before bridge shutdown."""
|
||
pending = [(task, run_id) for task, run_id in cleanup_tasks.items() if not task.done()]
|
||
if not pending:
|
||
return
|
||
for task, _run_id in pending:
|
||
task.cancel()
|
||
await asyncio.gather(*(task for task, _run_id in pending), return_exceptions=True)
|
||
|
||
run_ids = list(dict.fromkeys(run_id for _task, run_id in pending))
|
||
try:
|
||
results = await asyncio.wait_for(
|
||
asyncio.gather(
|
||
*(bridge.cleanup(run_id, delay=0) for run_id in run_ids),
|
||
return_exceptions=True,
|
||
),
|
||
timeout=max(0.0, timeout),
|
||
)
|
||
except TimeoutError:
|
||
logger.warning(
|
||
"Immediate recovered stream cleanup exceeded %.1fs for run_ids=%s; bridge TTL remains the final safety net",
|
||
timeout,
|
||
run_ids,
|
||
)
|
||
else:
|
||
for run_id, result in zip(run_ids, results):
|
||
if isinstance(result, BaseException):
|
||
logger.warning(
|
||
"Failed to immediately clean up recovered run stream for %s: %r",
|
||
run_id,
|
||
result,
|
||
)
|
||
|
||
|
||
if TYPE_CHECKING:
|
||
from app.gateway.auth.local_provider import LocalAuthProvider
|
||
from app.gateway.auth.repositories.sqlite import SQLiteUserRepository
|
||
from deerflow.persistence.thread_meta.base import ThreadMetaStore
|
||
from deerflow.runtime import RunRecord
|
||
|
||
|
||
T = TypeVar("T")
|
||
|
||
|
||
async def _mark_latest_startup_recovered_threads_error(
|
||
run_manager: RunManager,
|
||
thread_store: ThreadMetaStore,
|
||
recovered_runs: list[RunRecord],
|
||
) -> None:
|
||
"""Project startup recovery before request-serving concurrency begins.
|
||
|
||
This helper must remain on the pre-``yield`` startup path. ``ThreadMetaStore``
|
||
has no ``latest_run_id`` column, so it cannot express an atomic conditional
|
||
update keyed by the recovered run. Periodic recovery deliberately skips this
|
||
projection; moving this helper after request serving starts would reintroduce
|
||
a read/update race with newer runs.
|
||
"""
|
||
recovered_by_thread: dict[str, set[str]] = {}
|
||
for record in recovered_runs:
|
||
recovered_by_thread.setdefault(record.thread_id, set()).add(record.run_id)
|
||
|
||
for thread_id, recovered_run_ids in recovered_by_thread.items():
|
||
try:
|
||
latest_runs = await run_manager.list_by_thread(thread_id, user_id=None, limit=1)
|
||
except Exception:
|
||
logger.warning("Failed to find latest run for thread %s during run reconciliation", thread_id, exc_info=True)
|
||
continue
|
||
if not latest_runs or latest_runs[0].run_id not in recovered_run_ids:
|
||
continue
|
||
try:
|
||
await thread_store.update_status(thread_id, "error", user_id=None)
|
||
except Exception:
|
||
logger.warning("Failed to mark thread %s as error during run reconciliation", thread_id, exc_info=True)
|
||
|
||
|
||
async def _terminalize_recovered_runs(
|
||
bridge: StreamBridge,
|
||
recovered_runs: list[RunRecord],
|
||
*,
|
||
cleanup_delay: float,
|
||
on_cleanup_scheduled: Callable[[str, asyncio.Task[None]], None] | None = None,
|
||
) -> list[tuple[str, asyncio.Task[None]]]:
|
||
"""Publish terminal markers and schedule retained-stream cleanup."""
|
||
return await _publish_recovered_run_stream_end(
|
||
bridge,
|
||
recovered_runs,
|
||
cleanup_delay=cleanup_delay,
|
||
on_cleanup_scheduled=on_cleanup_scheduled,
|
||
)
|
||
|
||
|
||
def get_config() -> AppConfig:
|
||
"""Return the freshest ``AppConfig`` for the current request.
|
||
|
||
Routes through :func:`deerflow.config.app_config.get_app_config`, which
|
||
honours runtime ``ContextVar`` overrides and reloads ``config.yaml`` from
|
||
disk when its mtime changes. ``AppConfig`` is not cached on ``app.state``
|
||
at all — the only startup-time snapshot lives as a local
|
||
``startup_config`` variable inside ``lifespan()`` and is passed
|
||
explicitly into :func:`langgraph_runtime` for the engines that are
|
||
restart-required by design. Routing every request through
|
||
:func:`get_app_config` closes the bytedance/deer-flow issue #3107 BUG-001
|
||
split-brain where the worker / lead-agent thread saw a stale startup
|
||
snapshot.
|
||
|
||
Hot-reload boundary: fields backed by startup-time singletons
|
||
(engines, sandbox provider, IM channels, logging handler) require a
|
||
process restart to change at runtime. The authoritative list lives in
|
||
:mod:`deerflow.config.reload_boundary` and is mirrored by the
|
||
standardised ``"startup-only:"`` prefix on the matching
|
||
``Field(description=...)`` in :class:`AppConfig` — IDE hover on those
|
||
fields will surface the boundary inline. See
|
||
``backend/CLAUDE.md`` "Config Hot-Reload Boundary" for the operator
|
||
summary.
|
||
|
||
Any failure to materialise the config (missing file, permission denied,
|
||
YAML parse error, validation error) is reported as 503 — semantically
|
||
"the gateway cannot serve requests without a usable configuration" — and
|
||
logged with the original exception so operators have something to debug.
|
||
"""
|
||
try:
|
||
return get_app_config()
|
||
except Exception as exc: # noqa: BLE001 - request boundary: log and degrade gracefully
|
||
logger.exception("Failed to load AppConfig at request time")
|
||
raise HTTPException(status_code=503, detail="Configuration not available") from exc
|
||
|
||
|
||
@asynccontextmanager
|
||
async def langgraph_runtime(app: FastAPI, startup_config: AppConfig) -> AsyncGenerator[None, None]:
|
||
"""Bootstrap and tear down all LangGraph runtime singletons.
|
||
|
||
``startup_config`` is the ``AppConfig`` snapshot taken once during
|
||
``lifespan()`` for one-shot infrastructure bootstrap. The engines and
|
||
stores constructed here (stream bridge, persistence engine, checkpointer,
|
||
store, run-event store) are restart-required by design — they hold live
|
||
connections, file handles, or singleton providers — so they bind to this
|
||
snapshot and survive across `config.yaml` edits. Request-time consumers
|
||
must still go through :func:`get_config` for any field that should be
|
||
hot-reloadable. See ``backend/CLAUDE.md`` "Config Hot-Reload Boundary".
|
||
|
||
The matching ``run_events_config`` is frozen onto ``app.state`` so
|
||
:func:`get_run_context` pairs a freshly-loaded ``AppConfig`` with the
|
||
*startup-time* run-events configuration the underlying ``event_store``
|
||
was built from — otherwise the runtime could end up combining a live
|
||
new ``run_events_config`` with an event store still bound to the
|
||
previous backend.
|
||
|
||
Usage in ``app.py``::
|
||
|
||
async with langgraph_runtime(app, startup_config):
|
||
yield
|
||
"""
|
||
from deerflow.persistence.engine import close_engine, get_session_factory, init_engine_from_config
|
||
from deerflow.runtime import make_store, make_stream_bridge
|
||
from deerflow.runtime.checkpoint_mode import freeze_checkpoint_channel_mode, freeze_checkpoint_snapshot_frequency
|
||
from deerflow.runtime.checkpointer.async_provider import make_checkpointer
|
||
from deerflow.runtime.events.store import make_run_event_store
|
||
|
||
# ------------------------------------------------------------------
|
||
# Multi-worker safety gate: reject SQLite when GATEWAY_WORKERS > 1.
|
||
# SQLite write-locks cannot support concurrent multi-process access.
|
||
# ------------------------------------------------------------------
|
||
_enforce_postgres_for_multi_worker(startup_config)
|
||
# Reject agent_storage.backend='db' on a non-durable database, and warn on
|
||
# node-divergent file storage under multi-worker Postgres.
|
||
_validate_agent_storage(startup_config)
|
||
# Warn when login lockouts stay per-process under several Gateway processes.
|
||
_validate_login_throttle_storage(startup_config)
|
||
# Warn when a declared multi-instance deployment shares DeerMem's SQLite
|
||
# retrieval index across instances through the memory volume.
|
||
_validate_memory_retrieval_index(startup_config)
|
||
|
||
async with AsyncExitStack() as stack:
|
||
# Lifecycle and system-model hooks can originate on isolated subagent
|
||
# loops. Bind them to the Gateway's serving loop before any runtime
|
||
# dependency starts, then reset the binding last through the exit
|
||
# stack. Registering the callback synchronously here also covers every
|
||
# startup-failure and cancellation path below.
|
||
try:
|
||
from deerflow.extensions.notify import (
|
||
drain_extension_notify_dispatches,
|
||
reset_extension_notify_loop,
|
||
set_extension_notify_loop,
|
||
)
|
||
|
||
set_extension_notify_loop(asyncio.get_running_loop())
|
||
except Exception:
|
||
logger.exception("Failed to register the extension notify loop; sync observations will be dropped")
|
||
else:
|
||
|
||
async def reset_notify_loop_safely() -> None:
|
||
try:
|
||
await drain_extension_notify_dispatches()
|
||
except Exception:
|
||
logger.debug(
|
||
"Failed to drain pending extension notifications (non-fatal)",
|
||
exc_info=True,
|
||
)
|
||
finally:
|
||
try:
|
||
reset_extension_notify_loop()
|
||
except Exception:
|
||
logger.debug(
|
||
"Failed to reset the extension notify loop (non-fatal)",
|
||
exc_info=True,
|
||
)
|
||
|
||
stack.push_async_callback(reset_notify_loop_safely)
|
||
|
||
config = startup_config
|
||
app.state.checkpoint_channel_mode = freeze_checkpoint_channel_mode(config.database.checkpoint_channel_mode)
|
||
app.state.checkpoint_snapshot_frequency = freeze_checkpoint_snapshot_frequency(config.database.checkpoint_delta.snapshot_frequency)
|
||
|
||
app.state.stream_bridge = await stack.enter_async_context(make_stream_bridge(config))
|
||
|
||
# Initialize persistence engine BEFORE checkpointer so that
|
||
# auto-create-database logic runs first (postgres backend).
|
||
# Own cleanup before initialization so partial startup and host
|
||
# cancellation cannot strand an engine created along the way.
|
||
stack.push_async_callback(close_engine)
|
||
await init_engine_from_config(config.database)
|
||
|
||
app.state.checkpointer = await stack.enter_async_context(make_checkpointer(config))
|
||
app.state.store = await stack.enter_async_context(make_store(config))
|
||
|
||
# Record the checkpointer/Store backend and the provisioner endpoint
|
||
# selected from this startup snapshot so GET /health/ready probes what
|
||
# the running process actually uses. These singletons are
|
||
# restart-required by design and are never rebuilt on config.yaml hot
|
||
# reload, so the probe must not re-resolve process-wide configuration
|
||
# per request. The stream bridge needs no snapshot: the probe pings the
|
||
# singleton stored on app.state.stream_bridge above.
|
||
from app.gateway.health import READINESS_CHECKPOINTER_CONFIG_ATTR, READINESS_PROVISIONER_URL_ATTR, resolve_checkpointer_config, resolve_provisioner_url
|
||
|
||
setattr(app.state, READINESS_CHECKPOINTER_CONFIG_ATTR, resolve_checkpointer_config(config))
|
||
setattr(app.state, READINESS_PROVISIONER_URL_ATTR, resolve_provisioner_url(config))
|
||
|
||
# Initialize repositories — one get_session_factory() call for all.
|
||
sf = get_session_factory()
|
||
|
||
# The login throttle store is resolved once per process from the startup
|
||
# snapshot and the engine above (auth.local.throttle_storage is
|
||
# startup-only); the router reads it through the same hook tests use.
|
||
from app.gateway.auth.login_throttle import install_login_throttle_store, reset_login_throttle_store, resolve_login_throttle_store
|
||
|
||
install_login_throttle_store(resolve_login_throttle_store(config, session_factory=sf))
|
||
stack.callback(reset_login_throttle_store)
|
||
|
||
if sf is not None:
|
||
from deerflow.persistence.feedback import FeedbackRepository
|
||
from deerflow.persistence.personal_access_tokens import PersonalAccessTokenRepository
|
||
from deerflow.persistence.run import RunRepository
|
||
|
||
app.state.run_store = RunRepository(sf)
|
||
app.state.feedback_repo = FeedbackRepository(sf)
|
||
from app.gateway.auth.pat import PAT_LAST_USED_WRITE_INTERVAL_SECONDS
|
||
|
||
app.state.pat_repo = PersonalAccessTokenRepository(sf, last_used_write_interval_seconds=PAT_LAST_USED_WRITE_INTERVAL_SECONDS)
|
||
else:
|
||
from deerflow.runtime.runs.store.memory import MemoryRunStore
|
||
|
||
app.state.run_store = MemoryRunStore()
|
||
app.state.feedback_repo = None
|
||
# Memory backend has no durable PAT store, so Bearer credentials
|
||
# cannot be validated there and are rejected by the middleware.
|
||
app.state.pat_repo = None
|
||
|
||
# Evidence readers are available to Gateway-lifetime extension services,
|
||
# so the configured event store must exist before those services start.
|
||
run_events_config = getattr(config, "run_events", None)
|
||
app.state.run_events_config = run_events_config
|
||
app.state.run_event_store = make_run_event_store(run_events_config)
|
||
|
||
from deerflow.extensions.run_evidence import StoreRunEvidenceReader, StoreRunEvidenceReaderFactory
|
||
|
||
# Gateway-lifetime services are trusted operator extensions without a
|
||
# request principal. None deliberately binds this app-scoped reader to
|
||
# global, cross-user visibility; event content is not secret-redacted.
|
||
app.state.run_evidence_reader = StoreRunEvidenceReader(
|
||
app.state.run_store,
|
||
app.state.run_event_store,
|
||
user_id=None,
|
||
)
|
||
app.state.run_evidence_reader_factory = StoreRunEvidenceReaderFactory(
|
||
app.state.run_store,
|
||
app.state.run_event_store,
|
||
)
|
||
|
||
# Services are app-scoped. Capture this app's immutable extension set
|
||
# once and close over the same object for teardown; the process-wide
|
||
# singleton may be replaced by another app/test before shutdown.
|
||
from deerflow.extensions import EMPTY_EXTENSIONS, record_runtime_diagnostics
|
||
from deerflow.extensions.gateway import start_services, stop_services
|
||
|
||
extensions = getattr(app.state, "extensions", EMPTY_EXTENSIONS)
|
||
attempted_services: list[tuple[str, Any]] = []
|
||
|
||
async def stop_extension_services() -> None:
|
||
record_runtime_diagnostics(
|
||
await await_drained(
|
||
stop_services(
|
||
extensions,
|
||
service_entries=attempted_services,
|
||
)
|
||
)
|
||
)
|
||
|
||
# Register cleanup before starting: start() can partially acquire
|
||
# resources and then fail or be cancelled.
|
||
stack.push_async_callback(stop_extension_services)
|
||
record_runtime_diagnostics(
|
||
await start_services(
|
||
extensions,
|
||
config,
|
||
sf,
|
||
run_evidence_reader=app.state.run_evidence_reader,
|
||
attempted_services=attempted_services,
|
||
)
|
||
)
|
||
|
||
from deerflow.persistence.thread_meta import make_thread_store
|
||
|
||
app.state.thread_store = make_thread_store(sf, app.state.store)
|
||
if sf is not None:
|
||
from deerflow.persistence.mcp_tasks import McpTaskRepository
|
||
from deerflow.persistence.projects import ProjectDocumentRepository, ProjectRepository
|
||
from deerflow.persistence.scheduled_task_events import ScheduledTaskEventRepository
|
||
from deerflow.persistence.scheduled_task_runs import (
|
||
ScheduledTaskRunRepository,
|
||
)
|
||
from deerflow.persistence.scheduled_tasks import ScheduledTaskRepository
|
||
from deerflow.persistence.subagent_batches import SubagentBatchRepository
|
||
from deerflow.persistence.thread_reads import ThreadReadRepository
|
||
|
||
app.state.project_repo = ProjectRepository(sf)
|
||
app.state.project_document_repo = ProjectDocumentRepository(sf)
|
||
app.state.scheduled_task_repo = ScheduledTaskRepository(
|
||
sf,
|
||
run_repository=app.state.run_store,
|
||
)
|
||
app.state.scheduled_task_run_repo = ScheduledTaskRunRepository(
|
||
sf,
|
||
run_repository=app.state.run_store,
|
||
)
|
||
app.state.scheduled_task_event_repo = ScheduledTaskEventRepository(sf)
|
||
# Per-user unread state and the activity feed's read clock.
|
||
app.state.thread_read_repo = ThreadReadRepository(sf)
|
||
app.state.mcp_task_repo = McpTaskRepository(sf)
|
||
app.state.subagent_batch_repo = SubagentBatchRepository(sf)
|
||
else:
|
||
app.state.mcp_task_repo = None
|
||
app.state.project_repo = None
|
||
app.state.project_document_repo = None
|
||
app.state.subagent_batch_repo = None
|
||
app.state.scheduled_task_repo = None
|
||
app.state.scheduled_task_run_repo = None
|
||
app.state.scheduled_task_event_repo = None
|
||
app.state.thread_read_repo = None
|
||
|
||
# RunManager with store backing for persistence
|
||
run_ownership_config = getattr(config, "run_ownership", None)
|
||
sb_config = getattr(config, "stream_bridge", None)
|
||
cleanup_delay = getattr(sb_config, "recovered_stream_cleanup_delay_seconds", 60.0) if sb_config else 60.0
|
||
recovered_stream_cleanup_tasks: dict[asyncio.Task[None], str] = {}
|
||
|
||
def track_recovered_stream_cleanup(
|
||
run_id: str,
|
||
task: asyncio.Task[None],
|
||
) -> None:
|
||
recovered_stream_cleanup_tasks[task] = run_id
|
||
task.add_done_callback(lambda completed: recovered_stream_cleanup_tasks.pop(completed, None))
|
||
|
||
async def terminalize_recovered_runs(recovered_runs: list[RunRecord]) -> None:
|
||
await _cleanup_recovered_scheduled_goals(recovered_runs, run_manager=app.state.run_manager, checkpointer=app.state.checkpointer)
|
||
await _terminalize_recovered_runs(
|
||
app.state.stream_bridge,
|
||
recovered_runs,
|
||
cleanup_delay=cleanup_delay,
|
||
on_cleanup_scheduled=track_recovered_stream_cleanup,
|
||
)
|
||
|
||
app.state.run_manager = RunManager(
|
||
store=app.state.run_store,
|
||
run_ownership_config=run_ownership_config,
|
||
event_store=app.state.run_event_store,
|
||
on_orphans_recovered=terminalize_recovered_runs,
|
||
)
|
||
# Startup recovery: mark inflight runs whose lease has expired as error.
|
||
# In single-worker mode (SQLite / backend=memory), no run has a lease, so
|
||
# all inflight rows are reclaimed (unchanged behaviour). In multi-worker
|
||
# mode (Postgres), only runs with an expired lease are reclaimed; runs
|
||
# owned by another live worker are skipped.
|
||
from deerflow.utils.time import now_iso
|
||
|
||
recovered_runs = await app.state.run_manager.reconcile_orphaned_inflight_runs(
|
||
error=STARTUP_ORPHAN_RECOVERY_ERROR,
|
||
before=now_iso(),
|
||
stop_reason=ORPHAN_RECOVERY_STOP_REASON,
|
||
)
|
||
await terminalize_recovered_runs(recovered_runs)
|
||
await _mark_latest_startup_recovered_threads_error(
|
||
app.state.run_manager,
|
||
app.state.thread_store,
|
||
recovered_runs,
|
||
)
|
||
|
||
from app.gateway.auth.repositories.sqlite import SQLiteUserRepository
|
||
from app.gateway.extension_agent_runs import GatewayAgentRunsHost
|
||
|
||
app.state.agent_runs_host = GatewayAgentRunsHost(app, load_user=SQLiteUserRepository(sf).get_user_by_id if sf is not None else None)
|
||
stack.callback(app.state.agent_runs_host.close)
|
||
|
||
# Start the lease heartbeat if enabled (multi-worker deployments).
|
||
await app.state.run_manager.start_heartbeat()
|
||
|
||
try:
|
||
yield
|
||
finally:
|
||
app.state.agent_runs_host.close()
|
||
# Drain in-flight run tasks BEFORE the AsyncExitStack tears down the
|
||
# checkpointer (and its connection pool). A run still mid-graph would
|
||
# otherwise leak into asyncio.run() shutdown, where langgraph's
|
||
# _checkpointer_put_after_previous aput races the closed pool and
|
||
# raises PoolClosed (issue #3373).
|
||
run_manager = getattr(app.state, "run_manager", None)
|
||
if run_manager is not None:
|
||
shutdown_deadline = asyncio.get_running_loop().time() + _RUN_DRAIN_TIMEOUT_SECONDS
|
||
try:
|
||
await _drain_inflight_runs(run_manager)
|
||
finally:
|
||
await _flush_recovered_stream_cleanups(
|
||
app.state.stream_bridge,
|
||
recovered_stream_cleanup_tasks,
|
||
timeout=min(
|
||
1.0,
|
||
max(
|
||
0.0,
|
||
shutdown_deadline - asyncio.get_running_loop().time(),
|
||
),
|
||
),
|
||
)
|
||
|
||
|
||
# ---------------------------------------------------------------------------
|
||
# Getters – called by routers per-request
|
||
# ---------------------------------------------------------------------------
|
||
|
||
|
||
def _require(attr: str, label: str) -> Callable[[Request], T]:
|
||
"""Create a FastAPI dependency that returns ``app.state.<attr>`` or 503."""
|
||
|
||
def dep(request: Request) -> T:
|
||
val = getattr(request.app.state, attr, None)
|
||
if val is None:
|
||
raise HTTPException(status_code=503, detail=f"{label} not available")
|
||
return cast(T, val)
|
||
|
||
dep.__name__ = dep.__qualname__ = f"get_{attr}"
|
||
return dep
|
||
|
||
|
||
get_stream_bridge: Callable[[Request], StreamBridge] = _require("stream_bridge", "Stream bridge")
|
||
get_run_manager: Callable[[Request], RunManager] = _require("run_manager", "Run manager")
|
||
get_checkpointer: Callable[[Request], Checkpointer] = _require("checkpointer", "Checkpointer")
|
||
get_run_event_store: Callable[[Request], RunEventStore] = _require("run_event_store", "Run event store")
|
||
get_feedback_repo: Callable[[Request], FeedbackRepository] = _require("feedback_repo", "Feedback")
|
||
get_run_store: Callable[[Request], RunStore] = _require("run_store", "Run store")
|
||
get_project_repo = _require("project_repo", "Projects")
|
||
get_project_document_repo = _require("project_document_repo", "Projects")
|
||
|
||
|
||
def get_store(request: Request):
|
||
"""Return the global store (may be ``None`` if not configured)."""
|
||
return getattr(request.app.state, "store", None)
|
||
|
||
|
||
def get_thread_store(request: Request) -> ThreadMetaStore:
|
||
"""Return the thread metadata store (SQL or memory-backed)."""
|
||
val = getattr(request.app.state, "thread_store", None)
|
||
if val is None:
|
||
raise HTTPException(status_code=503, detail="Thread metadata store not available")
|
||
return val
|
||
|
||
|
||
def _scheduler_unavailable(message: str) -> HTTPException:
|
||
"""Coded 503 for the scheduled-task routes (contracts/scheduled_task_errors_contract.json)."""
|
||
from app.gateway.scheduled_task_errors import scheduler_error
|
||
|
||
return scheduler_error(503, "scheduler_unavailable", message)
|
||
|
||
|
||
def get_scheduled_task_repo(request: Request):
|
||
val = getattr(request.app.state, "scheduled_task_repo", None)
|
||
if val is None:
|
||
raise _scheduler_unavailable("Scheduled task repo not available")
|
||
return val
|
||
|
||
|
||
def get_scheduled_task_run_repo(request: Request):
|
||
val = getattr(request.app.state, "scheduled_task_run_repo", None)
|
||
if val is None:
|
||
raise _scheduler_unavailable("Scheduled task run repo not available")
|
||
return val
|
||
|
||
|
||
def get_scheduled_task_event_repo(request: Request):
|
||
val = getattr(request.app.state, "scheduled_task_event_repo", None)
|
||
if val is None:
|
||
raise HTTPException(status_code=503, detail="Scheduled task event repo not available")
|
||
return val
|
||
|
||
|
||
def get_scheduled_task_service(request: Request):
|
||
val = getattr(request.app.state, "scheduled_task_service", None)
|
||
if val is None:
|
||
raise _scheduler_unavailable("Scheduled task service not available")
|
||
return val
|
||
|
||
|
||
def is_scheduler_running(request: Request) -> bool:
|
||
"""Whether this Gateway process's scheduler poller is running.
|
||
|
||
False when the service is absent (no scheduler persistence) or was never
|
||
started (``scheduler.enabled: false``) or its poller stopped. Per process:
|
||
another worker may report differently.
|
||
"""
|
||
state = getattr(getattr(request, "app", None), "state", None)
|
||
service = getattr(state, "scheduled_task_service", None)
|
||
return bool(getattr(service, "is_running", False))
|
||
|
||
|
||
def get_mcp_task_repo(request: Request):
|
||
val = getattr(request.app.state, "mcp_task_repo", None)
|
||
if val is None:
|
||
raise HTTPException(status_code=503, detail="MCP task repo not available")
|
||
return val
|
||
|
||
|
||
def get_mcp_task_service(request: Request):
|
||
val = getattr(request.app.state, "mcp_task_service", None)
|
||
if val is None:
|
||
raise HTTPException(status_code=503, detail="MCP task service not available")
|
||
return val
|
||
|
||
|
||
def get_subagent_batch_repo(request: Request):
|
||
val = getattr(request.app.state, "subagent_batch_repo", None)
|
||
if val is None:
|
||
raise HTTPException(status_code=503, detail="Subagent batch repository not available")
|
||
return val
|
||
|
||
|
||
def get_subagent_batch_service(request: Request):
|
||
val = getattr(request.app.state, "subagent_batch_service", None)
|
||
if val is None:
|
||
raise HTTPException(status_code=503, detail="Subagent batch service not available")
|
||
return val
|
||
|
||
|
||
def get_run_context(request: Request) -> RunContext:
|
||
"""Build a :class:`RunContext` from ``app.state`` singletons.
|
||
|
||
Returns a *base* context with infrastructure dependencies. The
|
||
``app_config`` field is resolved live so per-run fields (e.g.
|
||
``models[*].max_tokens``) follow ``config.yaml`` edits; the
|
||
``event_store`` / ``run_events_config`` pair stays frozen to the snapshot
|
||
captured in :func:`langgraph_runtime` so callers never see a store bound to
|
||
one backend paired with a config pointing at another.
|
||
"""
|
||
host = getattr(request.app.state, "agent_runs_host", None)
|
||
# Internal/channel and PAT runs deliberately receive no retained delegation.
|
||
agent_runs = host.bind(request) if host is not None else None
|
||
return RunContext(
|
||
agent_runs=agent_runs,
|
||
checkpointer=get_checkpointer(request),
|
||
store=get_store(request),
|
||
event_store=get_run_event_store(request),
|
||
run_events_config=getattr(request.app.state, "run_events_config", None),
|
||
checkpoint_channel_mode=getattr(request.app.state, "checkpoint_channel_mode", "full"),
|
||
checkpoint_snapshot_frequency=getattr(request.app.state, "checkpoint_snapshot_frequency", None),
|
||
thread_store=get_thread_store(request),
|
||
mcp_task_repo=getattr(request.app.state, "mcp_task_repo", None),
|
||
app_config=get_config(),
|
||
extensions=getattr(request.app.state, "extensions", None),
|
||
on_run_completed=getattr(request.app.state, "scheduled_task_service", None).handle_run_completion if getattr(request.app.state, "scheduled_task_service", None) is not None else None,
|
||
)
|
||
|
||
|
||
# ---------------------------------------------------------------------------
|
||
# Auth helpers (used by authz.py and auth middleware)
|
||
# ---------------------------------------------------------------------------
|
||
|
||
# Cached singletons to avoid repeated instantiation per request
|
||
_cached_local_provider: LocalAuthProvider | None = None
|
||
_cached_repo: SQLiteUserRepository | None = None
|
||
|
||
|
||
def get_user_repository() -> SQLiteUserRepository:
|
||
"""Return the cached user repository (created on first use).
|
||
|
||
Origin: admin user-management surface (RFC #4063 / #3462 gap 2). Shares
|
||
the ``get_local_provider`` cache so both surfaces see one store.
|
||
"""
|
||
|
||
get_local_provider()
|
||
assert _cached_repo is not None
|
||
return _cached_repo
|
||
|
||
|
||
def get_local_provider() -> LocalAuthProvider:
|
||
"""Get or create the cached LocalAuthProvider singleton.
|
||
|
||
Must be called after ``init_engine_from_config()`` — the shared
|
||
session factory is required to construct the user repository.
|
||
"""
|
||
global _cached_local_provider, _cached_repo
|
||
if _cached_repo is None:
|
||
from app.gateway.auth.repositories.sqlite import SQLiteUserRepository
|
||
from deerflow.persistence.engine import get_session_factory
|
||
|
||
sf = get_session_factory()
|
||
if sf is None:
|
||
raise RuntimeError("get_local_provider() called before init_engine_from_config(); cannot access users table")
|
||
_cached_repo = SQLiteUserRepository(sf)
|
||
if _cached_local_provider is None:
|
||
from app.gateway.auth.local_provider import LocalAuthProvider
|
||
|
||
_cached_local_provider = LocalAuthProvider(repository=_cached_repo)
|
||
return _cached_local_provider
|
||
|
||
|
||
def get_pat_repo(request: Request):
|
||
"""Return the personal-access-token repository from app state.
|
||
|
||
Raises 503 when the process runs on the memory backend (no durable PAT
|
||
storage), so PAT management routes fail explicitly instead of silently
|
||
accepting tokens nobody can validate.
|
||
"""
|
||
pat_repo = getattr(request.app.state, "pat_repo", None)
|
||
if pat_repo is None:
|
||
raise HTTPException(status_code=503, detail="Personal access tokens require a configured database")
|
||
return pat_repo
|
||
|
||
|
||
async def get_current_user_from_request(request: Request):
|
||
"""Get the current authenticated user from the request cookie.
|
||
|
||
Raises HTTPException 401 if not authenticated.
|
||
"""
|
||
state = getattr(request, "state", None)
|
||
state_user = getattr(state, "user", None)
|
||
from app.gateway.auth_disabled import AUTH_SOURCE_AUTH_DISABLED, AUTH_SOURCE_INTERNAL, AUTH_SOURCE_PAT, AUTH_SOURCE_SESSION
|
||
|
||
if state_user is not None and getattr(state, "auth_source", None) in {
|
||
AUTH_SOURCE_SESSION,
|
||
AUTH_SOURCE_AUTH_DISABLED,
|
||
AUTH_SOURCE_INTERNAL,
|
||
AUTH_SOURCE_PAT,
|
||
}:
|
||
return state_user
|
||
|
||
from app.gateway.auth import decode_token
|
||
from app.gateway.auth.errors import AuthErrorCode, AuthErrorResponse, TokenError, token_error_to_code
|
||
|
||
access_token = request.cookies.get("access_token")
|
||
if not access_token:
|
||
raise HTTPException(
|
||
status_code=401,
|
||
detail=AuthErrorResponse(code=AuthErrorCode.NOT_AUTHENTICATED, message="Not authenticated").model_dump(),
|
||
)
|
||
|
||
payload = decode_token(access_token)
|
||
if isinstance(payload, TokenError):
|
||
raise HTTPException(
|
||
status_code=401,
|
||
detail=AuthErrorResponse(code=token_error_to_code(payload), message=f"Token error: {payload.value}").model_dump(),
|
||
)
|
||
|
||
provider = get_local_provider()
|
||
user = await provider.get_user(payload.sub)
|
||
if user is None:
|
||
raise HTTPException(
|
||
status_code=401,
|
||
detail=AuthErrorResponse(code=AuthErrorCode.USER_NOT_FOUND, message="User not found").model_dump(),
|
||
)
|
||
|
||
# Token version mismatch → password was changed, token is stale
|
||
if user.token_version != payload.ver:
|
||
raise HTTPException(
|
||
status_code=401,
|
||
detail=AuthErrorResponse(code=AuthErrorCode.TOKEN_INVALID, message="Token revoked (password changed)").model_dump(),
|
||
)
|
||
|
||
return user
|
||
|
||
|
||
async def is_admin_user(request: Request) -> bool:
|
||
"""Return whether the authenticated caller is an admin user.
|
||
|
||
``AuthMiddleware`` normally stamps ``request.state.user`` before the request
|
||
reaches a router. Falling back to the strict dependency keeps the route safe
|
||
in tests or alternative ASGI compositions that mount a router without the
|
||
global middleware.
|
||
|
||
Centralising this here means a future change to the admin definition (e.g.
|
||
allowing an internal system role, adding audit logging, or switching to a
|
||
permission-based check) lands in one place instead of drifting across the
|
||
per-router copies that previously existed in ``mcp``, ``channel_connections``
|
||
and ``channels``.
|
||
"""
|
||
# PAT credentials never carry admin capability: no scope in the PAT
|
||
# universe grants it, so an admin's automation token must not unlock
|
||
# admin-only routes (skill installs, integration credentials, MCP config).
|
||
from app.gateway.auth_disabled import AUTH_SOURCE_PAT
|
||
|
||
if getattr(request.state, "auth_source", None) == AUTH_SOURCE_PAT:
|
||
return False
|
||
user = getattr(request.state, "user", None)
|
||
if user is None:
|
||
user = await get_current_user_from_request(request)
|
||
|
||
return getattr(user, "system_role", None) == "admin"
|
||
|
||
|
||
async def require_admin_user(request: Request, *, detail: str) -> None:
|
||
"""Require the authenticated caller to be an admin user.
|
||
|
||
``detail`` is the route-specific 403 message. The shared predicate keeps
|
||
read-side redaction and write authorization on the same admin definition.
|
||
"""
|
||
|
||
if not await is_admin_user(request):
|
||
raise HTTPException(status_code=403, detail=detail)
|
||
|
||
|
||
async def get_optional_user_from_request(request: Request):
|
||
"""Get optional authenticated user from request.
|
||
|
||
Returns None if not authenticated.
|
||
"""
|
||
try:
|
||
return await get_current_user_from_request(request)
|
||
except HTTPException:
|
||
return None
|
||
|
||
|
||
async def get_current_user(request: Request) -> str | None:
|
||
"""Extract user_id from request cookie, or None if not authenticated.
|
||
|
||
Thin adapter that returns the string id for callers that only need
|
||
identification (e.g., ``feedback.py``). Full-user callers should use
|
||
``get_current_user_from_request`` or ``get_optional_user_from_request``.
|
||
"""
|
||
user = await get_optional_user_from_request(request)
|
||
return str(user.id) if user else None
|