mirror of
https://github.com/bytedance/deer-flow.git
synced 2026-10-30 23:20:35 +00:00
77 Commits
| Author | SHA1 | Message | Date | |
|---|---|---|---|---|
chaodidiao
|
6e4bcc16aa
|
fix(subagents): drain final cleanup across caller cancellation (#7546)
* fix(subagents): drain final cleanup across caller cancellation * fix(subagents): preserve stop outcomes during cancellation Follow up on PR #7546 discussion_r4374235193 and Issue #7535. Capture the unwinding cancellation before the drained cleanup tail and use it only while the result remains non-terminal. Preserve completed, failed and timed-out decisions. Keep cancellation propagation and cleanup ownership. Share the synchronous result holder and publish its existing FAILED outcome before requesting cancellation. This necessary timeout boundary overlaps #7483; no active-cancel terminal write is added to the cleanup tail. Validation: 12 targeted regressions pass on Windows and Linux; make test 37876 passed / 808 skipped, make test-blocking-io 292 passed / 1 skipped. Format, lint, diff checks and four-file incremental OCR delegated review pass. --------- Co-authored-by: Willem Jiang <willem.jiang@gmail.com> |
||
|
lilei
|
04dc9e5339
|
fix(subagents): report active cancellation to lifecycle observers (#7483)
* fix(subagents): report active cancellation to lifecycle observers * fix(subagents): preserve synchronous timeout lifecycle failures |
||
|
Xuehao Xu
|
8237bc6c9e
|
fix(subagents): preserve test summaries before evidence truncation (#7407)
* fix(subagents): preserve test summaries before evidence truncation * fix(subagents): honor Node counters and cache harvested summaries * fix(subagents): retain warning counts and Go case evidence * fix(subagents): retain long pytest sibling summaries --------- Co-authored-by: Willem Jiang <willem.jiang@gmail.com> |
||
|
Tsai Yuan
|
03c95f5278
|
fix(subagents): report model output length caps (#7355)
* fix(subagents): report model_length_capped when a subagent hits the output limit ModelLengthFinishReasonMiddleware only ran on the lead chain. A subagent whose response was cut at the provider output limit (finish_reason=length, stop_reason=max_tokens, MAX_TOKENS, Responses max_output_tokens) reached the lead as a clean `completed`, and a tool call truncated at that boundary still executed with partial arguments. Register the same middleware on the subagent chain, in the lead's relative position (after configured extensions, before Safety). It now also records each detected cap per context run_id and exposes `consume_stop_reason`, so the executor's existing duck-typed guard collection reports `stop_reason=model_length_capped`. Add the value to the subagent status contract (Python enum, contract fixture, frontend whitelist) with the label "model output length limit", and update the subagent docs. Tests cover the middleware channel, chain placement, the contract, and an end-to-end executor run with a fake model. Co-Authored-By: Claude Code <noreply@anthropic.com> * test(tasks): cover cap reasons on completed and failed runs Extend the existing shared-contract round-trip matrix to preserve both terminal statuses alongside every supported stop reason, including model_length_capped and loop_capped. --------- Co-authored-by: huangzz <110116@nd.com> Co-authored-by: Claude Code <noreply@anthropic.com> Co-authored-by: Willem Jiang <willem.jiang@gmail.com> |
||
|
Totoro
|
2f0cd407fc
|
fix(goal): base no-progress counting on successful operations (#7271)
* fix(goal): track successful operations for no-progress checks * fix(goal): reject failed exits and narration-only progress --------- Co-authored-by: Totoro-qaq <279883115+Totoro-qaq@users.noreply.github.com> |
||
|
trent87
|
0d73798beb
|
feature(skills): keep allowed-tools as compatibility metadata (#6935)
* refactor(skills): keep allowed-tools as compatibility metadata * docs(skills): align current guides with agent tool authority * test(skills): simplify authority regressions and sync Korean docs |
||
|
RongJie G
|
cd28a1cfe9
|
fix(subagents): bound hard timeout terminalization (#7053)
* fix(subagents): bound hard timeout terminalization Signed-off-by: CorgiBoyG <111257566+CorgiBoyG@users.noreply.github.com> * fix(subagents): address timeout lifecycle review Bound durable terminal cleanup waits to 60 seconds while unfinished execution resources stay supervised. Preserve terminal results when cancellation races a child failure, and register deferred registry removal when its fence wait expires. Add five lifecycle regression cases and preload the sandbox prewarm dependency in the executor test fixture. Validate 470 focused tests, 261 strict blocking-I/O tests, and 50 repeated regression executions under asyncio debug. --------- Signed-off-by: CorgiBoyG <111257566+CorgiBoyG@users.noreply.github.com> Co-authored-by: Willem Jiang <willem.jiang@gmail.com> |
||
|
Totoro
|
16d36ccd21
|
feat(sandbox): add bounded AIO prewarming and lifecycle benchmarks (#6998)
* fix(benchmark): validate sandbox lifecycle and concurrent measurements * feat(sandbox): prewarm likely AIO tool turns (cherry picked from commit 3a1017ed032ed9331987bc3dc61f23dcdac4e905) * fix(sandbox): bind bounded prewarming to execution leases * docs: trim frontend guidance below its existing size limit * fix(benchmark): separate failed lifecycle stage timings * test(sandbox): measure prewarm overlap and unused resources through agent graphs * fix(sandbox): finish lease and sampler cleanup after cancellation or errors * docs(benchmark): distinguish initialization wait from model overlap * docs(helm): align the configuration example with version 61 * fix(sandbox): preserve synchronous agent graphs when prewarming is enabled * test(cargo): isolate native fixtures from inherited make jobserver flags * fix(sandbox): require managed cleanup before prewarming * fix(sandbox): log denied prewarming as an expected skip * fix(benchmark): filter provider summary input records --------- Co-authored-by: nothingyuancando <1579035512@qq.com> |
||
|
Guo-Yixin
|
34ccd79351
|
fix(acceptance): recognize native Node test summaries (#6953)
* fix(acceptance): recognize native Node test summaries * test(acceptance): isolate native Node suite labels * fix(acceptance): recognize native Node test summaries * fix(acceptance): filter Node warning points * test(acceptance): cover Node warning reporter points * test(acceptance): cover Node warning labels with native reporters * fix(acceptance): keep failure counts on one diagnostic line |
||
|
Xuehao Xu
|
201b9840cd
|
feat(knowledge): unify read-only retrieval and source contracts (phase 1) (#6716)
* feat(knowledge): unify read-only retrieval and source contracts * fix(knowledge): address citation and contract review feedback * fix(knowledge): resolve citations in delegated task reports * docs(frontend): keep agent guidance within CI size budget * docs(backend): preserve merged guidance chain headroom |
||
|
NanPan
|
78150e9b30
|
fix(subagents): keep isolated-loop shutdown ownership retryable (#6316)
* fix(subagents): retain isolated-loop shutdown ownership * fix(subagents): reap stopped isolated loop before replacement * test(subagents): cover pending-loop recovery after worker exit * docs(subagents): document pending-loop recovery * fix(subagents): revalidate pending loop recovery |
||
|
Zeren Wang
|
d3e8e78f3f
|
feat(authz): middleware-declared tool path Layer 1 now covers tools declared by middlewares (#6105)
* feat(authz): extension tool authorization PR2 — middleware-declared tool path Layer 1 now covers tools declared by middlewares, which LangChain merges into the bound tool set after the host's explicit-list filter: - new agents/middlewares/tool_declarations.py: record the ordinary pass's verdict, collect declarations from the built stack, decide only the delta seeded by the first pass (a name denied for this build stays denied), narrow the stack build-locally on independent state-preserving copies (direct __dict__ write, refused up front for tools data descriptors, never mutating caller-owned instances), and verify the exact bound stack after the final normalization copy. - IsolatedMiddleware ships a state-preserving __copy__ so narrowing survives full/delta normalization and copies of copies. - TodoMiddleware degrades when write_todos is denied: no system-prompt injection, no completion/context-loss reminders, no jump_to. - Layer 2 forwards host-resolved tool provenance into AuthzRequest.context and binds the infrastructure exemption to the concrete host-created tool object instead of a name set (tool=None goes to the provider). - Wired into all four assembly paths (lead x2, native subagent with the decision offloaded via asyncio.to_thread, DeerFlowClient). Tests: new test_middleware_declared_tools.py (35) and test_layer2_tool_provenance.py (7); blocking-IO anchor pinning the subagent offload (teeth verified red->green); placement-guarantee and enforcement suites extended; executor test doubles converted for the async _create_agent. Behavior change called out in CHANGELOG: with authorization enabled and a tools policy that excludes write_todos, plan-mode builds no longer bind it (built-in RBAC default unchanged). * fix(authz): close residual gaps from review — narrow non-collectable declarations, dedupe candidates, warn on skipped pass - apply_declared_tool_view now removes non-BaseTool declarations from the bound view when authorization is enabled (they would otherwise bind via LangChain's create_tool auto-conversion with no provider decision); verify_declared_tool_view fails loudly if one survives normalization - decide_declared_tools dedupes never-submitted names so the provider sees each candidate once - subagent executor logs a warning when the provider is set but the seeded Layer-1 state is missing, instead of skipping the declaration pass silently - docs: middleware-declared tools must be BaseTool instances with usable names * fix(authz): fail loudly on non-list/tuple middleware tools declarations _iter_declared warned and returned () for a 'tools' attribute that is not a list/tuple, but LangChain's factory iterates the attribute as-is — so a set/generator/dict_values declaration bound with no Layer-1 decision. Raise DeclaredToolViewError instead, matching the module's fail-loud standard for every other un-narrowable shape. Disabled path remains a strict no-op. Docs updated to match. |
||
|
Guo-Yixin
|
4d5607229e
|
fix(subagents): preserve direct-return tool results (#6083)
* fix(subagents): preserve direct-return tool results * fix(subagents): include middleware tools and surface direct failures |
||
|
hataa
|
1cd3eb742b
|
feat(authz): enforce skill authorization at assembly and slash-activation (#4063 Phase 3) (#4541)
* feat(authz): enforce skill authorization at assembly and slash-activation (#4063 Phase 3) Phase 3 / Skills — the second of three resource-type PRs (Models, Skills, Sandbox). The RBAC provider already maps "skill" → config key "skills" (rbac.py _RESOURCE_POLICY_KEYS), so no schema change is needed. Layer 1 (assembly-time, mirrors Phase 1B): - filter_available_skills_by_authorization() in skill_filter.py filters the skill-name allowlist (set[str] | None) by the provider's "skill" policy. In lead agent, runs after _available_skill_names() so the catalog (describe_skill) and SkillActivationMiddleware share one filtered set. In subagent executor _load_skills(), filters config.skills before disk load. - filter_resources_by_authorization() in enforcement.py — generic version of filter_tools_by_authorization for any resource with a name attribute. Layer 2 (slash-activation): - No new middleware code needed. SkillActivationMiddleware._resolve_activation already checks "reference.name not in self._available_skills". Since Layer 1 filters the allowlist, denied skills cannot be slash-activated. authorization.enabled: false is a complete no-op. available_skills=None (no agent allowlist) + enabled: resolves all skill names from config and filters. 12 new tests + 243 existing authz/skills tests pass. * test(subagents): wait on the chained future instead of polling done() in caller-loop teardown test * feat(authz): enforce action-scoped skill:activate at runtime and reuse the catalog loader for filter candidates * feat(authz): arm skill:activate on the subagent chain, unify the client skill-surface user, and record the Layer 2 reversal in the design notes * chore(authz): align the subagent filter's fallback user with the chain's bucket convention * feat(authz): gate autonomous skill loading on skill:activate, close the bootstrap assembly leak, and reuse one provider per subagent assembly Review round 4 (willem-bd, CHANGES_REQUESTED): - [P1] Autonomous load path: describe_skill filters its match set on the action-scoped skill:activate decision (all-denied is indistinguishable from "no match"), and ToolErrorHandlingMiddleware stamps skill_context entries only when the action allows it — a denied SKILL.md read gets a skill_context_denied marker instead, so durable context, allowed-tools policy, and autonomous secret bindings never activate a denied skill. File visibility itself stays Layer 1 (sandbox-projection semantics). - [P1] Bootstrap assembly: build_middlewares/apply_prompt_template now receive the authorization-filtered set instead of the raw _BOOTSTRAP_SKILL_NAMES, and an empty skill_names index is preserved when deferred discovery is on rather than round-tripping through None into the legacy full-metadata prompt that re-advertises a denied skill (deferred-off keeps the legacy rendering; the filtered empty allowlist blanks it on denial). - [P2] SubagentExecutor resolves ResolvedSkillAuthorization once per assembly (_resolve_skill_authorization, memoized per executor) and shares it across the Layer 1 filter (authorization=), the activation middleware, and both new gates — uncached provider factories can no longer hand the layers different instances/policy snapshots. All gates share one helper (skill_filter.skill_activation_allowed) so slash, describe, and skill-file-load cannot drift; provider errors keep the configured fail-closed/fail-open policy. skill_authorization is threaded through _build_runtime_middlewares into both runtime builders. Tests: describe gating (deny/fail-closed/fail-open/disabled), read stamping (denied marker vs entry) + extract_skills skip without warning, bootstrap three-way param (allow/deny x deferred on/off), lead + subagent chain wiring for the stamp gate, and the distinct-instance provider factory repro. Each fix is mutation-verified (reverting it fails its test). Full A/B suite run: no new failures vs HEAD. * feat(authz): re-authorize persisted skill_context before reuse and route async paths through aauthorize Review round 5 (willem-bd): - [P1] Persisted entries: skill_context survives across runs, so both consumers now re-check the action-scoped skill:activate decision before reuse. SkillToolPolicyMiddleware._active_skills_for_paths skips a denied skill (same skip as a disabled/unallowlisted one; all-denied keeps the existing fail-closed-to-builtins semantics) and _in_context_secret_sources binds no secrets for a denied entry. skill_authorization is threaded into SkillToolPolicyMiddleware on both chains. - [P2] Async provider API: skill_activation_allowed_async mirrors the sync helper through await provider.aauthorize() (same pattern as authorize_sandbox_execution_async). Async hooks precompute an activation_decisions map on the event loop (slash target + persisted entry names, one batch of awaits) and hand it to the worker-thread handler, which consults the map instead of the sync API: SkillActivationMiddleware.awrap_model_call (activation + secret binding), ToolErrorHandlingMiddleware.awrap_tool_call (_amaybe_stamp awaits the decision before stamping), and SkillToolPolicyMiddleware's awrap_model_call/awrap_tool_call (entry names + slash source name). describe_skill becomes a StructuredTool with both a sync function (invoke -> authorize) and a coroutine (ainvoke -> aauthorize), so async agent execution never routes a loop-affine provider through the sync API while sync callers keep their semantics. Tests: allow-read-then-deny-next-run regressions for secret binding and allowed-tools (partial deny does not poison, all-denied fails closed to builtins), and an _AsyncOnlyProvider (sync authorize always fails, aauthorize works) driving four async-path assertions (activation, stamp allow/deny, describe, tool policy). All six fixes are mutation-verified; full A/B suite run against HEAD shows no new failures. * fix(authz): authorize skill activation by declared Skill.name, not directory basename ShenAC-SAC's P2: bundled skills may declare a name that differs from their directory (skills/public/vercel-deploy-claimable declares name: vercel-deploy), but the runtime gates derived authorization targets from the container path. - New deerflow/skills/container_registry.py: build_container_path_registry() + canonical_skill_name() — the one path-to-Skill.name resolution shared by the activation, tool-policy, and read-stamping middlewares. - ToolErrorHandlingMiddleware resolves the read path through the (user-scoped, via new user_id kwarg threaded from both runtime builders) registry before the skill:activate decision, sync and async (registry lookup in a worker thread, aauthorize() on the loop). RBAC allow by the declared name now activates; allow by the directory name does not. - SkillToolPolicyMiddleware async hooks canonicalize the policy paths first (three-phase: thread registry -> loop aauthorize -> thread filter reusing the registry; the cached per-model-step decision still short-circuits), so the decision map is keyed by the names _active_skills_for_paths checks — no more map miss falling back to the sync provider API from the thread. _entry_names removed: persisted entry["name"] is path-derived by construction; entries are resolved by path instead. - SkillActivationMiddleware collects slash names (already canonical) plus persisted entry *paths*, canonicalized off-loop via _canonical_names_for_paths(). Regressions (test_skills_authorization.py 45->50): sync stamp target, RBAC allow declared-name/deny directory-name both directions, async stamp via _AsyncOnlyProvider, composed async slash -> tool-policy keeps the skill's tools with zero sync fallback, persisted-entry secret binding survives the name mismatch. All three mutations (stamp target, policy map key, entry canonicalization) fail their tests. * fix(authz): close the sync-fallback class on skill activation paths willem-bd's two P2s on the canonical-name fix, plus a fourth instance of the same class found while sweeping the membership: - SkillToolPolicyMiddleware: a transient prepass registry failure is now preserved (_REGISTRY_LOAD_FAILED marker) — the worker-side filter applies its own fail-closed treatment instead of silently retrying storage, which recovered skills absent from the decision map and fell back to the synchronous provider API from the thread (fail-open retaining a denied skill's allowed-tools). - SkillActivationMiddleware: an empty entry-path list no longer pays a full skill-tree scan on ordinary authorization-enabled async model steps (previously a passive path with zero registry I/O). - 4th instance (reproduced with the PR fixtures, then fixed): the entry prepass failure on the secret-binding path let the consumer-side _load_skill_registry_by_path silently recover and resolve entry names outside the decision map — the same sync fallback, fail-open binding the aauthorize-denied skill's declared secret. The fix goes one step further than failure propagation: the prepass hands down the registry *snapshot* the decision map was keyed by, and entry sources resolve against that snapshot, so a map miss is impossible by construction (this also closes the narrower window where both loads succeed but disagree because storage changed mid-step). The slash source is deliberately unaffected: it does not consult the decision map (validated once at activation) and keeps its fresh per-call load. - Class membership swept: skill_activation_allowed() sync call sites are exactly four (stamping sync path and describe's sync implementation are legitimately sync); the only two async-reachable fallback points (the _activation_allowed helpers) are both construction-guarded now. Structural hardening (typed async-batch decision object where a map miss is fail-closed by construction) recorded in the plan notes as a follow-up, anchored at those two helpers. - _RegistryArg type alias replaces the degenerate dict|object|None union in the policy middleware signatures. Tests (test_skills_authorization.py 50->54): transient prepass failure -> builtins-only with exactly one storage attempt and zero provider calls; ordinary async step -> zero registry scans; two-call scenario -> entry binding zeroed while the slash binding survives; entries resolve against the prepass snapshot under a mid-step rename. Mutations verified red: marker reverted to None, over-broad slash kill, snapshot bypass. * fix(authz): correct secret-binding reference points and gate the rendered skill reminder Review round 9 (willem-bd) plus construction-audit follow-ups on the same class — every async divergence between two loads of the same data now has an explicit, pinned reference point: - Slash secret source resolves from a FRESH post-activation registry: the prepass snapshot is taken before the aauthorize awaits, so binding the slash source from it would activate NEW_KEY content while injecting OLD_KEY when a skill's declared secrets change in that window. The 'one scan per step' bonus that motivated sharing the snapshot is withdrawn — consistency wins. The two secret sources resolve independently: a transient fresh-load failure zeroes only the slash binding; entries keep binding from their own snapshot. - Same-skill dominance: when a skill is both the slash source (fresh) and a persisted entry (snapshot), only the activation-era declaration binds — unioning both eras injected a secret the skill no longer declares. Aligns with the tool-policy rule (slash activation dominates for the run); different-skill entries still stack additively. - The rendered 'Active skills' reminder (DurableContextMiddleware) no longer advertises skills the provider denies — surface consistency with describe_skill. Publication/consumption split: the activation middleware publishes per-step path-keyed skill:activate decisions into the run context (async: reused from the prepass snapshot, zero extra I/O; sync: one scan + authorize() only when entries exist — passive steps stay scan-free), the renderer consumes them and filters its render copy (state untouched; unpublished = permissive, matching the other run-context carriers). New carrier key listed in REDACTED_CONTEXT_KEYS with a redaction regression. - Wiring regressions on both chains for everything added above (user_id to the stamp gate, skill_authorization + token pairing + activation- before-durable ordering to the durable renderer): each mutation verified red (M4-M13 across this change set). - Tests 54->65: slash post-activation lookup, source independence, dominance, sync-chain sync-API path, render filter (async/sync/ disabled with load-count assertions), chain wiring x2, redaction listing. Plan notes record rounds 9-2..6 including the reference-point post-mortem and the audit axes exhausted. * fix(authz): anchor same-skill dominance on the slash identity, not binding results Review round 10 (willem-bd) plus a self-audit extension the new checklist items caught in the first fix itself: - R10 (P2): the same-skill exclusion was derived from successfully BOUND slash sources. When the prepass snapshots required-secrets [OLD_KEY] and the operator strips the declaration during the await window, the slash source binds nothing (required_secrets gate) — the exclusion set stayed empty and the stale entry view kept injecting OLD_KEY for a skill that now declares no secrets. The exclusion now derives from the AUTHENTICATED slash activation identity (run-context source path, owner-token checked): the identity's declared name resolves from the fresh registry first, then the entry snapshot, and joins the exclusion set regardless of binding success. - Self-audit (value->gate->downstream on the fix itself): the declared name can also change mid-run (same path, foo->bar). A name-anchored exclusion then holds only the new name while the snapshot entry resolves the same path to the old one — both eras bound again. The PATH is the stable identity shared by the activation and every earlier read: entries at the slash identity's path are now excluded before registry resolution (exclude_paths), independent of what either version calls the skill; the name-based set remains for cross-path same-name shadowing. Tests 65->67: no-secrets-era dominance (ACTIVE secrets None under the OLD_KEY-to-empty interleaving), path-anchored dominance across a mid-run rename (only the activation-era key binds). Mutations verified red: identity-derived exclusion disabled (M14), path exclusion disabled (M16). * test(authz): pass explicit config to lead middleware wiring test --------- Co-authored-by: Willem Jiang <willem.jiang@gmail.com> |
||
|
PeaceMaker-best
|
66597ae89d
|
fix(subagents): retain completed model usage on cancellation (#5910)
* fix(subagents): retain completed model usage on cancellation * docs(subagents): fit guidance within chain budget Signed-off-by: Willem Jiang <willem.jiang@gmail.com> --------- Signed-off-by: Willem Jiang <willem.jiang@gmail.com> Co-authored-by: PeaceMaker-best <221849497+PeaceMaker-best@users.noreply.github.com> Co-authored-by: Willem Jiang <willem.jiang@gmail.com> |
||
|
RongJie G
|
842f561054
|
fix(runtime): drain agent stream close across cancellation (#5806)
* fix(runtime): drain agent stream close across cancellation * docs(runtime): condense stream teardown guidance * fix(runtime): surface stream close failures on cancellation --------- Co-authored-by: CorgiBoyG <CorgiBoyG@users.noreply.github.com> Co-authored-by: Willem Jiang <willem.jiang@gmail.com> |
||
|
yang rui
|
1da284bb3a
|
feat(prompts): add operator overlays for core system prompts (#5794)
* feat(prompts): add operator overlays for core system prompts * test(prompts): isolate SOUL forwarding test from operator config * fix(prompts): wrap complete subagent system message with overlay * fix(subagents): serialize prompt overlays in durable batches --------- Co-authored-by: Willem Jiang <willem.jiang@gmail.com> |
||
|
Nan Gao
|
bcaf6db36a
|
feat(models): add declarative reasoning capability contract (#5780)
* feat(models): add declarative reasoning capability contract (#5073) Introduce an optional per-model `reasoning:` block beside the legacy `supports_thinking` / `supports_reasoning_effort` booleans. It declares thinking availability (unsupported | optional | required), the accepted effort values with aliases, default, and serialization path, the payload dialect, and the reasoning-history requirement. When present the booleans are derived from it and contradictory profiles fail at config load. `deerflow.models.reasoning` normalizes legacy and declared profiles into one immutable contract and resolves every caller's generic request against it. `create_chat_model` is the single enforcement point: required-thinking models never receive a synthesized disable payload, unsupported effort values never reach the provider, and the dialect synthesizes the on/off payload when no template exists. Legacy profiles keep the historical path byte-for-byte. The lead agent and subagent descriptor resolve the same policy first so run metadata reports effective values. `/api/models` and `DeerFlowClient` project the contract as `reasoning`; the frontend derives mode clamping and effort choices from it through `core/models/reasoning.ts`, and user preferences accept provider-specific effort tokens. The setup wizard's Z.AI GLM-5.3-Flash profile migrates to the contract, restoring its low/high/max effort control. * docs(models): move the reasoning contract guidance into models/AGENTS.md backend/AGENTS.md sits within 46 bytes of the 96 KiB effective-chain hard limit for the middlewares, runtime, sandbox, and subagents guides, so the new section pushed scripts/check_agent_guidance.py over it. The models guide is the nearest file for the factory and the contract module, and its chain has ample room. * fix(models): address review on the reasoning contract - vLLM dialect: the enable path now mirrors whichever chat_template_kwargs switch the template declares, like the disable path already did, so a template using the older `thinking` key no longer also grows `enable_thinking`. - Validate operator-supplied effort values at `effort.path` inside `when_thinking_enabled`, `when_thinking_disabled`, and the `thinking` shortcut against the contract, not only the profile-level value; they are forwarded when the caller chooses nothing, so the "unknown values never reach the provider" guarantee now holds for them too. - Constrain `effort.path` to dotted identifiers and reject single-segment paths that would replace a whole mapping (`extra_body`, `thinking`, `model_kwargs`, `default_headers`, `default_query`). - Z.AI GLM-5.3-Flash profile: `default: high` instead of `max`, because the default also governs summarization, title generation, and subagent calls that never choose an effort; `max` stays selectable. Documented in the config example, the configuration manual, and the design note. * fix(models): close reasoning contract review gaps * fix(models): preserve native Ollama reasoning config * fix(models): keep native reasoning strings and project reasoning into the descriptor - `ModelConfig.reasoning` also accepts the level string ChatOllama takes for gpt-oss style models (`low` / `medium` / `high`); the factory forwards both the boolean and the string on the legacy path, so profiles that passed `reasoning: high` through before the contract field existed keep loading and keep sending it. - The assembly descriptor no longer excludes `reasoning` from `model_parameters`: the native ChatOllama value and a declared contract's dialect/history/effort change the request payload, so they now move the assembly fingerprint. - e2e: the sidecar spec pinned the pre-contract body where a model without effort control still received `reasoning_effort: minimal`; the preset is now dropped for such models (the Gateway discarded it anyway), so the assertion expects the key to be absent. |
||
|
RongJie G
|
e01314442c
|
fix(mcp): scope sessions and task access by thread incarnation (#5556)
* fix(mcp): scope sessions and task access by thread incarnation * fix(mcp): preserve thread incarnation in delegated subagents * fix(mcp): preserve incarnation in durable batches * fix(mcp): bind standalone graph lifecycle context * fix(studio): preserve implicit thread creation metadata --------- Co-authored-by: CorgiBoyG <CorgiBoyG@users.noreply.github.com> Co-authored-by: Willem Jiang <willem.jiang@gmail.com> |
||
|
zhangwei-way
|
b6503e9a35
|
feat(knowledge): add per-message RAGFlow retrieval scope (#5238)
* feat(knowledge): integrate RAGFlow retrieval and management * test(knowledge): cover merged listing tool * feat(knowledge): add per-message retrieval scope * chore(docs): remove unrelated document * docs(knowledge): add interaction screenshots * feat(knowledge): simplify scope selector trigger * docs(knowledge): refresh selector screenshot * feat(knowledge): defer standalone management * docs(knowledge): show chat-only scope UI * fix(knowledge): honor scope on clarification replies * fix(knowledge): harden scoped replay validation * docs(knowledge): clarify replay scope precedence * fix(knowledge): keep provider settings on tools * fix(config): preserve tools-only knowledge settings * fix(knowledge): submit custom assistant identity * refactor(knowledge): trim PR scope changes * fix(knowledge): sanitize document scope display * feat(knowledge): enable scope selection in main chat * fix(knowledge): emphasize active scope icon without button frame * fix(knowledge): close context scrubbing and refresh e2e checks * fix(knowledge): preserve idempotent canonical retries * fix(knowledge): accept promptless conversation runs * style(knowledge): format backend regression tests * chore(knowledge): trim PR scope and fix frontend format * fix(knowledge): remove shared-scope notice * fix(knowledge): remove scope persistence notice * docs(knowledge): include main chat in catalog scope * fix(knowledge): preserve scope recovery and upgrades * fix(config): preserve LightRAG knowledge upgrades --------- Co-authored-by: foreleven <for-eleven@hotmail.com> |
||
|
RongJie G
|
94110e5dce
|
fix(subagents): close stream before releasing resources (#5221)
Co-authored-by: CorgiBoyG <CorgiBoyG@users.noreply.github.com> |
||
|
Hyeonsang Cho
|
53798b44cd
|
fix(subagents): scale max_turns into the graph's super-step budget (#5485)
* fix(subagents): scale max_turns into the graph's super-step budget
max_turns was handed to LangGraph as recursion_limit, but the two count
different things. recursion_limit counts super-steps, one per graph node,
and create_agent compiles a node for every middleware lifecycle hook, so
one turn costs before_model + model + after_model + tools nodes — seven to
eight through the subagent chain. The built-in general-purpose agent's
max_turns=150 therefore bought about 18 tool-using turns before failing as
turn_capped, and every middleware added to the chain shrank the effective
budget again.
Resolve the limit from the chain each subagent was actually assembled with
(subagents/turn_budget.py) instead of passing the turn count through, so
raising max_turns buys the turns it names.
No config keys or defaults changed; existing max_turns values now grant
their full budget, bounded as before by subagents.timeout_seconds and
subagents.token_budget.
* fix(subagents): warn when a counted hook can jump the agent loop
Review follow-up. The per-turn cost is a flat multiplier over the straight
before_model -> model -> tools loop. A hook that declares can_jump_to and
returns {"jump_to": ...} re-enters the loop without traversing tools,
spending another before_model + model + after_model pass that buys no tool
result, so the resolved limit becomes a lower bound rather than an exact
budget — silently re-creating the short budget this translation fixes.
Measured against a compiled graph: with one jumping after_model hook,
three tool turns need the resolved limit plus one jump pass, and the run
raises GraphRecursionError at the resolved limit.
How often a jump fires is data-dependent and unbounded, so it cannot be
folded into the arithmetic. find_jumping_hooks reports the condition off
the same __can_jump_to__ attribute the factory reads, and the executor
warns when a counted hook declares one. Nothing in today's subagent chain
does, so this changes no budget.
* fix(subagents): detect jumps declared on agent-level hooks too
Review follow-up. find_jumping_hooks exempted before_agent/after_agent on
the grounds that a jump out of them lands in the loop the budget already
pays for. That does not hold on langchain 1.3.14:
- after_agent jumps re-enter the loop after it finished, and the hook runs
again on the next exit, so the extra passes are unbounded. Even
jump_to "end" is routed to exit_node, the head of the after_agent chain,
so it reruns the chain; destinations are no safe filter.
- a before_agent hook that stages a tool call and jumps to tools runs a
tools step no model turn paid for. It is O(1), but the resolved limit
has zero headroom, so one step caps the last turn.
The detector now scans every hook pair the factory wires jump edges for.
The compiled-graph pin is parametrized over after_model->model,
after_agent->model, after_agent->end and before_agent->tools, each raising
GraphRecursionError at the resolved limit and completing once the jump's
cost is added. No middleware in the subagent chain declares a jump on any
hook, so this changes no budget.
|
||
|
PeaceMaker-best
|
14c9d44440
|
feat(runtime): persist tool-progress phase transitions (#5214)
* feat(runtime): persist tool-progress phase transitions Record bounded warn, block, and recover decisions for lead and task subagent runs while preserving event-loop isolation, fail-open behavior, and concurrent transition order. * fix(runtime): trust server-owned tool progress attribution * fix(runtime): centralize trusted audit attribution * fix(runtime): preserve complete tool progress audit state * docs: trim tool progress guidance to pass size check * fix(runtime): fence subagent audit recorder loop * docs(readme): sync tool-progress event coverage Signed-off-by: PeaceMaker-best <221849497+PeaceMaker-best@users.noreply.github.com> --------- Signed-off-by: PeaceMaker-best <221849497+PeaceMaker-best@users.noreply.github.com> Co-authored-by: PeaceMaker-best <221849497+PeaceMaker-best@users.noreply.github.com> Co-authored-by: 嗜鵼 <hy2010hy2010@qq.com> Co-authored-by: Willem Jiang <willem.jiang@gmail.com> |
||
|
Wenchao An
|
6d5d7bb1d5
|
feat(subagents): add opt-in parent context snapshots (#5367)
* feat(subagents): add opt-in parent context snapshots * test(subagents): package synthetic snapshot evaluation * fix(subagents): preserve output text and defer snapshot capture * docs(subagents): keep snapshot guidance within chain budget * fix(subagents): omit unpaired tool calls from snapshots * fix(subagents): safely omit unserializable snapshot media |
||
|
PeaceMaker-best
|
3c36217a51
|
feat(observability): persist deferred tool promotions (#5183)
* feat(observability): persist deferred tool promotions Signed-off-by: PeaceMaker-best <221849497+PeaceMaker-best@users.noreply.github.com> * fix(ci): trim agent guidance chain Signed-off-by: PeaceMaker-best <221849497+PeaceMaker-best@users.noreply.github.com> --------- Signed-off-by: PeaceMaker-best <221849497+PeaceMaker-best@users.noreply.github.com> Co-authored-by: PeaceMaker-best <221849497+PeaceMaker-best@users.noreply.github.com> |
||
|
PeaceMaker-best
|
fb28ed0122
|
feat(subagents): enable historical upload discovery (#5170)
Signed-off-by: PeaceMaker-best <221849497+PeaceMaker-best@users.noreply.github.com> Co-authored-by: PeaceMaker-best <221849497+PeaceMaker-best@users.noreply.github.com> |
||
|
Michael
|
85ffb66d6e
|
fix(subagents): stamp UTC-aware datetimes on SubagentResult lifecycle (#5153)
## Why DeerFlow declares one timestamp convention in deerflow/utils/time.py: every lifecycle timestamp is UTC (now_iso / datetime.now(UTC)). SubagentResult writers in subagents/executor.py still used naive datetime.now(), so on any non-UTC host the in-memory lifecycle metadata (started_at / completed_at) was local wall-clock time. The sibling durable-batch path (subagents/batch_service.py) already stamps datetime.now(UTC), so the same run model carried two different conventions depending on which path wrote it. ## What changed - Added an executor-local _utcnow() helper that stamps datetime.now(UTC). - SubagentResult.completed_at default in try_set_terminal(), result.started_at in _aexecute(), and the started_at default in _aexecute_admitted() now route through _utcnow(). - Explicit caller-supplied timestamps (completed_at=...) still pass through unchanged. - Added regression tests asserting the default writers produce UTC-aware datetimes. ## Surface area - [x] Backend runtime (deerflow.subagents.executor) - internal dataclass lifecycle metadata, no wire format change - [ ] Frontend UI / Backend API / Sandbox / Skills / Dependencies / Default behavior change ## Bug fix verification - New tests: tests/test_subagent_executor.py::test_timestamp_writers_stamp_utc_aware_datetimes and test_utcnow_helper_returns_utc_aware_datetime encode the convention. - Updated BlockingDateTime.now() in the terminal-publication-order test to mirror datetime.now's optional tz argument. ## Validation - cd backend && python -m pytest tests/test_subagent_executor.py: 136 passed; 2 pre-existing TestBashExecutionHarvest failures reproduce identically on clean main (Windows sandbox env), unrelated to this change. - ruff format + ruff check clean on both changed files. ## AI assistance **Tool(s) used:** Codex (coding agent) **How you used it:** analysis of the timestamp conventions, implementation, and regression tests authored with AI assistance; change reviewed before commit. - [ ] I've read and understand every line of this change and take responsibility for it — it's not unreviewed AI output. |
||
|
Aari
|
9e0fbd60fa
|
fix(sandbox): isolate concurrent subagent shell sessions (#5134)
* fix(sandbox): isolate concurrent subagent shell sessions * fix(sandbox): make execution acquire idempotent * fix(sandbox): close execution lifecycle gaps * fix(sandbox): serialize retained client lifecycle * fix(sandbox): close remaining client lifecycle gaps * fix(sandbox): unwind failed client lookup * fix(sandbox): protect internal lease identities * fix(sandbox): make cancellation reconciliation durable * fix(sandbox): fence cancelled workers and IM uploads |
||
|
AoHanBei
|
9b32b5d841
|
feat(observability): persist loop detection events (#5127)
* feat(observability): persist loop detection events * fix(observability): persist subagent loop events * fix(observability): narrow subagent loop event bridge * fix(observability): attribute subagent loop events * fix(tests): isolate subagent executor imports |
||
|
Hyeonsang Cho
|
9146bfa03d
|
feature(gateway): issue request trace ids unconditionally (#5119)
* refactor(gateway): issue request trace ids unconditionally The request trace id was gated behind logging.enhance.enabled at every entry point, so downstream code had to keep asking whether one existed: a header-provenance flag in its own ContextVar, a precedence resolver, and three-level carrier fallbacks at each consumer. Bind one unconditionally instead. TraceMiddleware covers Gateway HTTP; ensure_trace_context covers the entry points that never touch ASGI -- scheduled occurrences, MCP task notification runs, IM channel messages, and the embedded client -- each scoped to one unit of work so a long-lived worker task cannot leak one occurrence's id into the next. The ContextVar becomes the only source; the response header, runtime context, run metadata and log records are derived outputs. Consumers now use ensure_trace_id() or resolve_trace_id(*carriers) and drop their presence guards. Removed: resolve_deerflow_trace_id, the header-provenance flag and its three helpers, set/reset_current_trace_id, is_trace_correlation_enabled and its gateway alias. BREAKING CHANGE: every Gateway HTTP response now carries X-Trace-Id and it cannot be turned off; logging.enhance.enabled controls log output only. Installations on the default enabled: false will start seeing the header. No config keys were added or removed. * fix(gateway): stop persisting a caller-supplied trace id on the run record body.metadata forks two ways: through build_run_config into the live run config, which the run worker restamps, and through create_or_reject into the run record that the runs API echoes verbatim. Only the first was covered, so a client sending metadata.deerflow_trace_id made the most durable and most visible surface of a run disagree with the X-Trace-Id and the log lines the same request produced -- a correlation id that does not match the logs is worse than none. Stamp the server-issued id once at the trust boundary so both forks receive it, preserving the caller's own metadata keys. Close the same gap on config.context, which reaches the runtime context by a separate path: _build_runtime_context no longer merges server-owned keys from the caller, and _install_runtime_context assigns rather than setdefaults. A thread's metadata is no longer seeded with the run-scoped id of whichever run created it -- one thread spans many runs and as many trace ids. Found by driving a real run through the Gateway and reading the run back from the runs API; every unit test built its metadata by hand and so could not see it. * fix(gateway): expose X-Trace-Id to split-origin browser clients X-Trace-Id is not on the CORS safelist, so a browser client served from a separate origin could not read it -- and those are exactly the clients that cannot read the Gateway's logs either, leaving them with nothing to quote in a bug report. Same-origin nginx deployments were unaffected, which is why this stayed hidden. Add it to CORS_EXPOSED_HEADERS beside Content-Location, referencing TRACE_ID_HEADER rather than repeating the literal. * fix(gateway): keep X-Trace-Id on unhandled-exception 500s Starlette's ServerErrorMiddleware sits outside every user middleware and emits unhandled-exception 500s through the raw send, so those responses never pass TraceMiddleware's header-writing wrapper. The 500 for a server bug is exactly the response a user most needs to correlate with a log line, and it was the one response that shipped without the id. TraceMiddleware now tracks whether http.response.start has been sent. On an exception with no response started it emits its own plain 500 carrying the header, then re-raises: the outer ServerErrorMiddleware sees the response already started and only re-raises too, so the server's exception logging is untouched. An exception mid-stream keeps propagating unchanged — a second response start cannot be sent, and the already-written header stands. The trace id is printable ASCII by construction (normalize_trace_id / generate_trace_id), which is what makes the raw latin-1 header encoding safe. * fix(gateway): strip the forged trace id from the persisted request echo The run-record fix stopped a forged metadata.deerflow_trace_id on the authoritative metadata surface, but the raw request echo still carried one: create_or_reject persists body.config verbatim as runs.kwargs_json, which the runs API serves back. A client posting config.context.deerflow_trace_id therefore still got its forged value stored and echoed on one API surface while the header, logs, run metadata, and checkpoint all carried the real id — the id is ignored as input there, so echoing it back only manufactures disagreement. Two changes close it. redact_config_secrets — already the shared scrub for that echo, applied at admission and again at serve time, so historical records are covered too — now also drops deerflow_trace_id from config.metadata and config.context. And build_run_config now merges run metadata onto a copy of the caller's config["metadata"] instead of updating it in place: the nested values of the request config are reference copies, so the in-place merge was writing the server-stamped key through into body.config, contaminating the "what the client sent" record before it was persisted (and incidentally masking the forged-value echo on the metadata container). The regression test posts a forged id through body.metadata, config.metadata, and config.context at once and reads the kwargs echo back off the run record, failing if either leak returns. * docs(harness): record the trace-echo scrub, 500 fallback, and accepted retry divergence The trace section of the harness AGENTS.md now covers the two fixes that close the derived-output rule (the kwargs-echo scrub in redact_config_secrets plus build_run_config's copy merge, and TraceMiddleware's own 500 for unhandled exceptions), and CHANGELOG gains their Fixed entries. It also writes down the one accepted divergence: a crash-recovered scheduled launch reuses the durable run through its idempotency key, and start_run returns early on idempotency_reused without restamping — so the run record keeps the first attempt's deerflow_trace_id while the retry's own log lines carry the freshly minted id of its ensure_trace_context binding. The divergence is confined to the crash-recovery window and is accepted rather than fixed: restamping on reuse would rewrite a persisted record for a run that already exists, which is worse than two ids that each correlate their own attempt's logs. Written down so the next reader of the scheduler recovery path does not diagnose it as a bug. * docs(config): align the logging.enhance schema note with the unconditional trace id The config-module AGENTS.md still described logging.enhance as the gate for the Gateway X-Trace-Id header and Langfuse deerflow_trace_id. That model is gone: ids are issued unconditionally and this block decides log output only. Left as-is, the stale wording invites an agent to "restore" a header gate it believes was lost. Reworded to match the sibling AGENTS.md files and config.example.yaml, with a pointer to the Request Trace Context section that owns the full model. * docs(changelog): link the trace entries to #5119 The five new entries pointed at the ([#XXXX]) placeholder with no reference definition, rendering as literal text instead of a link — and RELEASING.md step 2 relies on those references when the section becomes release notes. All five now point at #5119, with the definition appended to the reference block. * refactor(harness): rename _stream_without_trace_context to _stream_turn The name asserted the opposite of what the method now does. It was accurate while logging.enhance.enabled could route stream() around the trace scope; with the gate gone it is the only stream implementation left, and it binds the id itself via ensure_trace_id(). Private, so the rename touches only the definition and the one stream() call site. * docs(harness): fit the trace-context guidance inside the AGENTS.md chain budget The expanded Request Trace Context section pushed the effective AGENTS.md chain for agents/middlewares to 99,815 bytes, past the 98,304 hard limit scripts/check_agent_guidance.py enforces in CI (AG002). Compressed the section from 7,359 to 4592 bytes with no facts removed: the entry-point table, the derived-output rule and its enforcement points, the accepted scheduled-retry divergence, the two resolution helpers, the stream() binding rationale, the log-output-only gate, the CORS listing, the 500 fallback, and the test map all remain. Sized against the merge, not just the branch: current main grew the same chain by ~724 bytes, so the check was verified on the merged tree as well (97,772 bytes; branch tree 97,048). * fix(gateway): declare content-length on the fallback 500 The pre-response 500 declared content-type but no content-length, leaving the framing to the ASGI server: chunked on HTTP/1.1, close-delimited on HTTP/1.0 — the one wire difference from the ServerErrorMiddleware response it replaces, which sends content-length: 21. The explicit header keeps the fallback byte-identical to what clients saw before. * docs(readme): drop the trace-correlation condition from the translations The zh/ja/fr/ru Langfuse sections still said metadata.deerflow_trace_id matches X-Trace-Id "when request trace correlation is enabled". The id now always matches and that condition no longer exists, so each bullet states the unconditional match and that logging.enhance.enabled only controls whether the id is printed into logs — the one piece of the feature a user can still configure. * test(gateway): pin TraceMiddleware wiring through create_app() Every X-Trace-Id test exercised a hand-built four-route app, so the real stack's add_middleware(TraceMiddleware) line was pinned by nothing: deleting it — or short-circuiting above it — passed CI while silently dropping both the response header and the ambient id the run-record stamp and enhanced log records derive from. One case now drives /health through create_app() and asserts the inbound id round-trips; mutation-checked by removing the wiring line, which fails exactly this test. * docs(gateway): note the fallback 500 is CORS-opaque The pre-response 500 is emitted outside CORSMiddleware — the exception has already unwound past it — so it carries no Access-Control-Allow-Origin and a split-origin browser client cannot read the id on this one response, unchanged from the ServerErrorMiddleware 500 it replaces. Documented on the class and in the CHANGELOG entry rather than fixed: replicating the origin allowlist outside CORSMiddleware would let the two policies drift. * fix(harness): keep abandoned-stream cleanup inside the trace binding stream() binds the turn's id around each next(inner) and resets it before yielding, but the finally's inner.close() ran after that binding was gone. Abandoning the stream therefore drove the inner LangGraph generator's GeneratorExit/finally path with no trace id — or an unrelated ambient one from whichever context ran the close — so cancellation and finalization logs and callbacks did not correlate with the turn they belong to. inner.close() is now wrapped in a local bind/reset of the same turn id. The token is set and reset in the same frame, never across a yield, so the per-step cross-context safety is preserved even when GC closes the generator from another Context — pinned by the existing copy_context close test, which now exercises this path. The regression test records the id from the inner generator's finally and fails without the binding. * test(harness): teach the worker-trace fake about RunManager.cleanup Upstream #5112 (bound gateway memory after terminal runs) added a run_manager.cleanup(run_id) call to run_agent's finalization, so the merge-commit CI run failed all five worker-trace-binding tests with AttributeError on this PR's _FakeRunManager. The fake gains the same no-op shape as its other methods. * docs(gateway): bring the gateway AGENTS.md back under its soft budget Upstream #5092 grew backend/app/gateway/AGENTS.md to 40,966 bytes, 6 over the 40,960 soft budget that test_agent_guidance_check.py::test_repository_guidance_stays_below_soft_budgets_and_avoids_doc_indexes enforces — its Unit Tests run on main was cancelled by push concurrency, so main is currently red on that test and every PR merge-run inherits the failure. Two whitespace/wording trims in the row #5092 touched (a doubled space, and "its configured `context_window`" → "its `context_window`") bring the file to 40,953 with no content change. --------- Co-authored-by: Willem Jiang <willem.jiang@gmail.com> |
||
|
Zeren Wang
|
a06a6fed7e
|
feat(harness): deterministic acceptance checklist for subagent delegations (RFC #4651, layer 2) (#5109)
* feat(harness): deterministic acceptance checklist for subagent delegations (RFC #4651, layer 2) PR4 of RFC #4651: check lead-supplied acceptance_criteria in code when a subagent completes, so objectively checkable requirements can never be silently passed by a self-report. - subagents/acceptance_checks.py: deterministic leaf families — file:<path> exists|non-empty and file_written:<path> read through read_current_file_content scoped to the shared thread workspace; the read uses the sandbox-native virtual path form (the local read validator and provider mount tables resolve /mnt/user-data/... paths, not host paths); the scope decision canonicalizes with realpath on the local sandbox so workspace symlinks cannot escape into uploads; a remote provider's "Error: ..." return string is normalized to a failed check (provider-typed via is_local_sandbox); a UnicodeDecodeError marks a binary deliverable as existing and non-empty; out-of-scope paths degrade to UNVERIFIED. tests_passed:<command> anchors to a matching recorded bash execution with status=success and a test-summary shape; matching is shell-structure aware with control-flow attribution (span must end at the last segment with provable execution), negating-option values are ineligible evidence and a target negated anywhere in the command degrades the match, extra flags must be selection-preserving, extra positionals widen only after a path-scoped criterion, truncated commands degrade via command_truncated, the summary shape is read only from output attributable to the matched segment (preceding segments provably silent by invocation form), and pass shapes require a nonzero passed count. Criterion text is neutralized with neutralize_untrusted_tags before storage/rendering. Anything else renders UNVERIFIED, never silently passed. - executor: accumulate bounded bash command/output evidence per streamed chunk (merged by tool_call_id, newest-capped) so subagent summarization compacting earlier messages cannot erase a recorded execution; the recorded status is the actual shell exit status parsed from the output's exit marker (signed codes included; the remote Command exited with code N form is accepted only as the whole trimmed output), falling back to deerflow_tool_meta only when no marker exists. - sandbox providers: e2b/opensandbox/tenki/boxlite append the LocalSandbox-style "Exit Code: N" marker on nonzero exit even with non-empty output; aio propagates the SDK's structured exit_code on both exec paths the same way; local timeouts append Exit Code: 124; and _truncate_bash_output always preserves a trailing exit marker (signed included) inside its budget, with a 32-char floor raising any smaller configured limit, so the actual shell outcome always survives in the output text. - task_tool: run the checklist offloaded (asyncio.to_thread) on the completed branch, failure-isolated; stamp the verdict into result metadata and render the per-criterion section into the model-visible result text. - status contract: additive subagent_acceptance_verdict transport with read-side structural validation. - delegation ledger: entry carries the verdict and renders a compact acceptance segment; gateway strips caller-forged verdicts from both ledger entries and message metadata, like the citation verdict. - blocking-IO anchor pins the offload (teeth proven red->green); leaf read errors catch only OSError/SandboxError so unexpected errors reach the task-tool-level isolation instead of being mislabeled. * fix(harness): close acceptance evidence gaps from review (RFC #4651 PR4) - negating options: overlap with a matched criterion target is now checked by path/nodeid prefix, not exact token equality — excluding a sub-path of the criterion's selection (pytest tests --deselect tests/unit/test_auth.py) degrades to UNVERIFIED instead of holds - output attribution: any redirection token in the matched final segment makes the recorded tail non-attributable (> / >> / 2> are word characters to the parser, so redirection was invisible to the matcher) - silent-source allowlist narrowed from any *activate suffix to the */bin/activate shape - status_contract docstring: restore the shared-fixture sentence and note subagent_acceptance_verdict is deliberately outside the fixture - executor: update_bash_executions publishes [] (stream carried no bash-family calls) instead of collapsing it into None, mirroring update_tool_receipts * fix(harness): close acceptance residual gaps from re-review (RFC #4651 PR4) - tests_passed: add error outcomes to the fail shapes — "4 passed, 1 error" and pytest's "ERROR <nodeid>" short summary no longer satisfy the pass shape when the exit status is swallowed (|| true) or absent; zero-error counts stay clean. - file leaves: bound the deliverable read — a "wc -c" shell size probe answers files above 50k bytes without loading ~2x their size, honoring the host-bash kill switch and falling back to the full read on any non-integer rendering, so verdicts never get less sound. - executor: record the exit marker text as status_marker on harvested bash evidence; the leaf detail now reports the marker actually seen instead of asserting a failure indistinguishable from the command's own trailing text. - extend the blocking-IO anchor to drive the probe branch inside the offload; teeth re-verified red->green. * fix(harness): close acceptance forgery and bound gaps from P2 re-review (RFC #4651 PR4) - file leaves: never read unbounded — size is established first (os.stat on the validated local host path, so the host-bash-disabled configuration needs no shell; a guarded wc -c on remote providers that renders missing/unreadable in its own words). Above the 50k cap the leaf answers from the size alone, at/below it the full read runs, and an unestablishable size degrades to UNVERIFIED instead of an unlimited fallback read. - output attribution: source/. prefixes are never provably silent — a crafted */bin/activate path shape says nothing about what the script prints, so sourced segments can no longer lend a passing summary. - executable identity: an explicitly path-spelled criterion now requires the same normalized executable path; the basename rule stays only for deliberately bare criterion commands. * fix(harness): run acceptance size probe outside subagent-controlled state (RFC #4651 PR4) - remote probe no longer runs in the sandbox's persistent shell: a fresh env -i /bin/sh with absolute-path stat/realpath (poisoned functions, aliases, PATH, exported functions, IFS, locale cannot steer it), plus a marker env routing AIO onto a fresh per-call bash.exec session. - metadata-only: stat never opens content, so a FIFO deliverable cannot block the parent for the provider's idle timeout; non-regular files (fifo/dir/symlink) degrade to UNVERIFIED. - containment canonicalized against the literal mount root: a final-component symlink or a swapped parent directory (root included) cannot redirect the check outside shared storage; unprovable layouts degrade to UNVERIFIED. * fix(harness): canonicalize probe containment against the canonical mount root (RFC #4651 PR4) Literal-root equality made every remote file leaf permanently UNVERIFIED on e2b and Tenki, which realize /mnt/user-data as a symlink to the home dir by default (e2b bootstrap 'sudo ln -sfn', Tenki best-effort symlink). Containment now compares the file's realpath against the mount root's realpath — exactly what the provider's own read path resolves, so probe and read-back stay consistent; final-component symlinks stay rejected by the non-dereferencing stat, and an intermediate dir-link escape under a sane root still lands ESCAPED. The inner script is a module constant and the suite now executes the composed probe for real against on-disk layouts (real dir, symlinked prefix, final symlink, fifo, missing, dir-link escape), which the canned-output stub could not see. * fix(harness): close bare-criterion negation and CDPATH summary channels (RFC #4651 PR4) - matching: a criterion with no positional selection target (bare pytest, make test) stands for the runner's default selection, so ANY negating option (--ignore/--deselect/...) makes the recorded run a different selection — unprovable. The overlap guard only sees consumed criterion tokens, which a bare criterion does not have; scoped criteria keep the unrelated-exclusion behavior. - attribution: cd is no longer blanket-silent — CDPATH makes cd print the resolved (subagent-chosen) destination and the pass shapes match as substrings, so one mkdir 'all tests passed' plus an export minted a pass for any quiet command. A cd argument or CDPATH= value (export or leading assignment) carrying any summary shape makes the segment non-silent; shape-free cd dir wrappers keep matching. - docs: _truncate_bash_output states the effective 32-char floor (the guarantee previously read as an unconditional max_chars bound). * fix(harness): close env-assignment and expansion channels in acceptance matching (RFC #4651 PR4) Self-audit in the shape of the last review rounds — channels the matcher classified as accounted-for that can change what runs, narrow the selection, or lend the summary text: - env assignments are no longer blanket-stripped: only an allowlist of inert display/CI knobs (CI, NO_COLOR, PY_COLORS, ...) may prefix a matched span, and a non-allowlisted assignment in any preceding segment (pure-assignment or export NAME=) is state pollution — PATH redirects the executable, LD_PRELOAD/PYTHONPATH/NODE_OPTIONS inject code, PYTEST_ADDOPTS/GOFLAGS/MAKEFILES inject selection-changing inputs, BASH_ENV runs arbitrary shell startup. All degrade to unprovable. - runtime expansions: any span token carrying /$( )/backticks, any negating-option value carrying an expansion or glob (unknown excluded set), and any extra executed token carrying glob metacharacters (crafted option-looking filenames narrow invisibly) are unprovable. Criterion-side globs stay self-consistent (literal match). - cd: an argument carrying a runtime expansion or glob is non-silent (unknown destination, unknown print); CDPATH= assignments are now handled as state pollution at the match layer, subsuming the value-shape special case. * fix(harness): persistent-shell evidence, exact env sets, option-arity scoping (RFC #4651 PR4) - tests_passed: on a persistent-shell provider (new Sandbox.persistent_shell_sessions capability, set by AioSandbox) every leaf degrades to UNVERIFIED — any earlier call in the shared session could have mutated the state the clean-looking run executed in, and only a fresh controlled session (RFC section 6 verifier) can prove otherwise. The flag is read from the provider registry without acquiring a sandbox. - env assignments: the allowlist is gone — no variable is provably inert across repositories (CI/DEBUG are routinely read by tests). The span's assignment prefix must equal the criterion's exactly (values included, order-insensitive); any assignment or export NAME= in a preceding segment is state pollution. - scoping: positional targets are now read by option arity, so a path embedded in an option (--basetemp=/tmp/p, --junitxml=/tmp/r.xml) never counts as a selection target and an extra positional after such a criterion narrows the default selection it denotes. * fix(harness): stamp shell provenance at harvest, close export/unset and arity gaps (RFC #4651 PR4) * fix(harness): split physical newlines as shell separators in acceptance matching (RFC #4651 PR4) * fix(harness): scope cd wrappers to thread data roots, pin accepted boundaries (RFC #4651 PR4) * fix(harness): preserve criterion connectors, prove file_written readable, fail-closed shell capability (RFC #4651 PR4) * fix(harness): compare only the connector prefix, tolerate trailing criterion semicolons (RFC #4651 PR4) * fix(harness): preserve continuation-line operators, keep ./-spelled executable identity (RFC #4651 PR4) * fix(harness): render criteria single-line so a multiline criterion cannot inject a forged checklist line (RFC #4651 PR4) * fix(harness): reject parent-traversal executable tokens in acceptance matching (RFC #4651 PR4) * fix(harness): reject parent-traversal negated values in acceptance matching (RFC #4651 PR4) |
||
|
hataa
|
17cac3420a
|
fix(subagents): clean up background task entry on unexpected poller exit (#5069)
* fix(subagents): clean up background task entry on unexpected poller exit * fix(subagents): pin deferred cleanup to the persistent subagent loop The non-terminal fallback scheduled the deferred registry cleanup with asyncio.create_task on the poller's own loop. Under synchronous tool invocation the sync wrapper runs the tool coroutine through asyncio.run(), which cancels caller-loop tasks at teardown, so the cleanup died before executing and the _background_tasks entry leaked — the same lifecycle leak the terminal path already fixed. Schedule the deferred cleaner on the process-owned persistent subagent loop instead, via the new public executor helper run_on_isolated_subagent_loop (asyncio.run_coroutine_threadsafe). The cleaner only touches thread-safe registry helpers, so it is loop-agnostic. A caller-loop fallback remains for the unreachable case where the persistent loop cannot be obtained, so scheduling never raises out of an unwind path that is already handling an error. The polling-timeout return path, which shares the scheduler, is fixed the same way. Tests no longer stub the scheduler: the non-terminal fallback test drives the real scheduling wrapper on an equivalent long-lived loop and asserts cleanup runs after asyncio.run() tears the caller loop down, and run_on_isolated_subagent_loop itself is covered against the real persistent loop with the caller loop closed underneath. * fix(subagents): harden interrupted finalization against failing status path Three edge cases from review on the unexpected-exit unwind: 1. Finalization no longer depends on the failing status accessor. _peek_subagent_result distinguishes a gone entry from an unreadable one instead of letting the accessor's exception abort the unwind; _finalize_interrupted_subagent never raises (so the original poller exception is preserved) and attaches the deferred cleaner, whose last resort force-removes a persistently unreadable entry via the new executor force_cleanup_background_task. 2. The generic-error unwind waits only a short grace period (_UNEXPECTED_EXIT_GRACE_SECONDS) instead of the full execution timeout before re-raising; the remaining lifecycle stays with the deferred cleaner on the persistent subagent loop. 3. The deferred cleaner reports the subagent's final usage (deltas since the unwind snapshot included, via final=True bypassing usage_reported; the journal dedupes by source_run_id) before removing the terminal entry. The report is transferred in a plain worker thread so the RunJournal's loop-bound progress flush is skipped rather than scheduled on a foreign loop. * fix(subagents): pin deferred final usage delivery to the parent run loop _report_deferred_final_usage ran record_external_llm_usage_records in a to_thread worker, making it the first cross-thread RunJournal writer: the unlocked accumulators can lose token updates and _tokens_by_model mutations race get_completion_data() iteration on the parent loop. Capture the parent loop at unwind time (it is alive in every path that continues the run) and deliver the final report onto it with call_soon_threadsafe, serialized with all other journal access; when that loop is already closed (asyncio.run teardown) the report is dropped on purpose — the run has persisted and nothing reads the counters back. The live-loop test exercises the real recorder path (journal captures the running loop of every call) instead of stubbing _report_subagent_usage, so a cross-thread report would surface as a wrong-loop entry. Also gates the two teardown tests on the caller loop actually closing (the deferred cleaner could otherwise legitimately deliver while that loop is still winding down), and adds a task_tool-level regression test for execute_async submit failure leaving no registry residue (rolled back inside execute_async since #5086). * docs(subagents): document the reverse loop boundary; observability + test fixes Extend subagents/AGENTS.md's Isolated-loop callback boundary with the reverse-direction contract from #5069: deferred registry cleanup is pinned to the persistent subagent loop via run_on_isolated_subagent_loop, and the final usage report is handed back onto the parent run's loop captured at unwind time — never invoked from the persistent loop or a worker thread, which would silently reintroduce the journal accumulator/iteration race. Dropping the final report (closed parent loop) is the one path where a subagent's tail usage goes permanently unaccounted, so both drop branches now log at info with the execution id and the unaccounted record count. Restore the retention assert in test_deferred_cleanup_task_retained_and_ survives_gc: the bounded wait observes the production done-callback discard, the assert (not a manual discard) is what fails if that callback is deleted. * fix(subagents): honour grace-wait cancellation and shrink deferred-cleaner captures Two review follow-ups on the unwind: - The shared unwind absorbs CancelledError (never-raise contract), so a graph-node cancellation landing inside the generic-error grace wait was swallowed and the node ended as a failed tool call instead of an interrupted run. The generic branch now re-checks task.cancelling() after the unwind and re-raises CancelledError; the absorb site documents why the cancellation path needs it and where the discrimination lives. - The deferred cleaner captured the whole run runtime, pinning the parent run's journal and event store for up to a full poll budget (~31 min) via the strongly-held task handle — worst on the polling-timeout path, where a stuck subagent pinned its run's journal for a second full timeout after the tool returned. The recorder is now resolved on the unwind path and is the only capture (plus ids and the report loop); a None recorder skips reporting entirely. Both behaviours are regression-tested (red on the previous head, green after): a cancellation parked inside the grace wait surfaces as CancelledError, and the runtime is collectable while the cleaner still polls. The closed-loop drop test now uses a real recorder via runtime.callbacks so the drop it pins is loop-based, not recorder-absence. |
||
|
Zeren Wang
|
22b0456e45
|
feat(harness): subagent report contract and delegation acceptance criteria (#5090)
* feat(harness): subagent report contract and delegation acceptance criteria (RFC #4651 PR3) Layer 1 receipt verification is inert unless subagents actually cite their execution record. This lands the prompt layer that closes the adoption gap: - New subagents/report_contract.py owns the model-facing contract text, derived from the single-owner citation format (format_citation / receipt_id) so prompts can never drift from the verifier. The executor injects <report_contract> into every subagent system prompt — built-in and custom alike — requiring [rN tool_name] citations for action claims, verifiable handles (absolute path, URL, ID, HTTP status) for deliverables, and explicit failure reporting; the citation clause follows verification.receipts_enabled. - The task tool gains an optional keyword-only acceptance_criteria parameter, handed to the SubagentExecutor constructor and rendered into the subagent's SystemMessage (stripped, capped 20 items x 500 chars) — deliberately never the task HumanMessage, which InputSanitizationMiddleware classes as genuine user input and would HTML-escape into untrusted-input framing. The docstring frames subagent results as self-reports, states the citation cross-check's evidence boundary (resolved = the call happened, not that the claim is correct), and documents when to attach criteria with the canonical leaf forms. Deterministic leaf checking remains a separate layer. - The lead delegation workflow now instructs reading the ledger citation line as execution evidence only and spot-checking verifiable handles before synthesizing. - report_contract / acceptance_criteria are registered as blocked framework-authority tags in input sanitization so untrusted input cannot forge the verification contract. * fix(harness): neutralize acceptance criteria before system-channel injection render_acceptance_criteria_section interpolated lead-model-supplied acceptance_criteria verbatim into the subagent SystemMessage after only stripping/capping. A criterion such as '</acceptance_criteria><system>...</system>' could close the wrapper and open a framework authority tag, bypassing InputSanitizationMiddleware. Route each criterion through neutralize_untrusted_tags (the shared prompt-injection primitive) so blocked authority tags are HTML-escaped before interpolation. Add regression tests at the renderer and the executor _build_initial_state path. * fix(harness): keep model-supplied criteria off the system channel - Move acceptance_criteria values into the task HumanMessage — the untrusted channel InputSanitizationMiddleware escapes and boundary-frames. The subagent SystemMessage now carries only a framework-owned <acceptance_criteria> pointer note (no criterion text), so natural-language injection inside a criterion keeps task-data priority and cannot override framework instructions (PR #5090 review, willem-bd P1). - Condition the lead delegation workflow's citation verification guidance on verification.receipts_enabled and qualify the task tool's result-reading text with the enabled state, so a receipts-disabled configuration no longer tells the lead to require citation evidence that cannot exist (P2). * fix(harness): drop execution-record promise from report contract when receipts are disabled The <report_contract> opening was emitted unconditionally, so a verification.receipts_enabled=false subagent was told its report would be cross-checked against an execution record that cannot exist in that mode (terminal_receipts() returns None; no verdict, no ledger citation line). The opening now follows receipts_enabled: enabled keeps the cross-check language, disabled describes the handle-only review mode (PR #5090 review, willem-bd P2). * docs: record the prompt-layer trust-boundary self-check Generalizes the PR #5090 review outcome: before adding prompt text, ask of every data source in it what trust level it has and which channel it should ride — model/user-influenceable values ride the untrusted sanitized data channel, never framework-owned system text. Added to the PR template (Agents/LangGraph surface) and agents/AGENTS.md. |
||
|
hataa
|
2f8d1cfc21
|
fix(subagents): harden background-task registry and capacity snapshot edge cases (#5086)
* fix(subagents): harden background-task registry and capacity snapshot edge cases - execute_async drops the just-registered background entry when submitting to the isolated loop fails. The caller sees the exception and never polls, and cleanup_background_task refuses non-terminal entries, so the entry would otherwise stay as a PENDING zombie forever. - SubagentExecutionCapacity.snapshot derives queued from len(_waiters) instead of iterating it. snapshot is read from non-loop threads (e.g. configure_subagent_execution_capacity) while the loop thread mutates the deque, so iteration can raise 'deque mutated during iteration'. The raw length may count a waiter that just timed out but has not removed itself yet, which only makes the busy-check more conservative. * fix(subagents): close failure-path gaps around background submit Address both review findings on the background submission lifecycle: - execute_async() copies the isolated-loop context before registering the _background_tasks entry, so a context-copy failure (callback- manager copy or loop-bound handler filtering) can no longer strand a permanent PENDING entry the caller will never poll. - _submit_to_isolated_loop_in_context() resolves the loop before calling the coroutine factory. As direct run_coroutine_threadsafe arguments the coroutine was created first, so a loop-startup failure stranded a never-awaited coroutine (RuntimeWarning + retained captures until collection). Both call sites share the fix. New tests verified red on the previous implementation, green after: - context-copy failure leaves no registry residue - the real submit helper (only the loop getter patched) never invokes the coroutine factory when loop startup fails * fix(subagents): close the coroutine when scheduling rejects it run_coroutine_threadsafe can itself raise once the coroutine exists (e.g. the loop closes between the lookup and the internal call_soon_threadsafe). Wrap the call, close the rejected coroutine, and re-raise; a focused test patches only run_coroutine_threadsafe and asserts the created coroutine reaches CORO_CLOSED. |
||
|
Zeren Wang
|
3b592c2053
|
feat(harness): subagent receipt citation verification (#5076)
* feat(harness): subagent receipt citation verification - add receipt citation verification core - harvest subagent tool receipts at terminal status - transport subagent receipts and citation verdict via status contract - verify subagent report citations at task write-back - render citation verdicts in the delegation ledger * fix(gateway): strip forged receipt verdicts from the delegations channel normalize_input() and the checkpoint-state mutation sanitizer only stripped server-owned metadata from message-shaped values, so an external caller could submit a delegation entry carrying a forged receipt_verdict that render_delegation_ledger would present as runtime-owned execution evidence. Strip receipt_verdict from caller-supplied delegation entries on both the run-creation and thread-state mutation paths, with regression coverage for each. * fix(harness): close silent-pass gaps in the zero-citation heuristic The action-claim detector missed the most common completion verbs (fixed/added/tested/changed/...) and had no CJK coverage at all, so reports like 'I fixed the bug and added tests.' or '我已经创建了文件并运行了测试。' were treated as claim-free: citation_resolved=True with no ledger warning. Broaden the verb lists and add a language-independent safety net: when the run harvested receipts and a nontrivial (>=240 char) report cites none of them, flag it UNVERIFIED. Short claim-free confirmations remain a vacuous pass. |
||
|
Aari
|
ff0a6768c2
|
feat(subagents): add unified capacity and durable batch execution (#4998)
* feat(subagents): add capacity controls and durable batches * fix(helm): sync subagent config schema version * fix(subagents): preserve batch history without worker * fix(subagents): support explicit factory runtimes * fix: address durable batch review findings |
||
|
Nan Gao
|
13f0a7f263
|
feat(extensions): let an out-of-tree extension observe what the agent did (#4863)
* feat(extensions): let an out-of-tree extension observe what the agent did
DeerFlow's extension system can contribute middleware, services and routes,
but an extension cannot answer basic questions about a run without reaching
into host internals. Several of the facts it would need are destroyed by the
operations that produce them:
* The middleware chain injects and rewrites a lot of context — date
reminders, recalled memory, compaction summaries, durable-context data,
image payloads, activated skill bodies. Downstream, none of it is
attributable: at the model-call boundary an injected HumanMessage is
indistinguishable from the user's own, and anything wanting to tell them
apart has to pattern-match prompt wording, which breaks on the next copy
edit.
* Two runs of "the same agent" are only comparable if the chain enforced the
same limits, prompts and thresholds. Recovering that from outside means
reading private attributes and guessing which of them change behaviour — a
guess that rots silently as middlewares gain fields.
* The lead-agent factory resolves a model after runtime overrides, renders a
prompt, filters tools through authorization and composes a stack, all
inside one synchronous call, and none of it survives: a middleware sees its
neighbours but not the prompt, the run worker sees a graph but not what
went into it.
* Summarization is destructive by design. N messages leave the context and
one summary enters it; afterwards only the summary exists, so "which
messages became this?" is not reconstructible.
This adds seven neutral facilities so those facts are recorded where they are
still true, and releases the contract package as 0.2.0.
Message provenance
Producers stamp `deerflow_content_kind` / `deerflow_producer_kind` onto the
messages they inject or rewrite. Stamping is unconditional — a fact whose
presence depends on whether an observer is installed is not a fact — and the
keys are server-owned, so provenance cannot be forged from a request.
Middleware self-description
Twelve middlewares declare their own behaviour-affecting parameters through
a duck-typed `release_policy_parameters()`. Long text is hashed rather than
embedded: a declaration is an identity, not a copy of the prompt.
Agent assembly descriptor
`assemble_lead_agent()` returns the graph plus a descriptor whose fingerprint
answers "did anything about this agent change between these two runs?".
`make_lead_agent()` keeps its graph-only signature — it is the LangGraph
Server ABI declared in langgraph.json. Tools and skills are sorted before
hashing because their assembly order is incidental; middlewares are not,
because stack order decides what wraps what. Host build identity is reported
but excluded from the fingerprint, so a redeploy does not invalidate every
agent's identity.
Context compaction observation
Summarization emits the content hashes of the messages it is about to remove
joined to the summary that replaced them. Content is the only identity
available at that seam: the summary does not become a message, and what later
projects it into a request renders it bounded and escaped rather than
verbatim.
Neutral policy, transform and MCP-source facts
Guardrail decisions are published to runtime context under a `__`-prefixed
key; result-rewriting middlewares append a declared, ordered transform trail;
MCP tools carry their credential-free logical origin.
Extension route identity
Contributed routes are session-authenticated and cannot opt out, but
"logged in" and "administrator" are different questions. Extensions get a
neutral projection of the caller rather than the host's auth context, and
`require_admin` fails closed when identity cannot be determined.
Extension-owned tables
An extension that persists data owns its own MetaData and migration chain, so
its tables are absent from Base.metadata and `alembic revision --autogenerate`
proposes dropping them. Extensions declare a table prefix, which is rejected
at registration if it would shadow a host table.
The contract package stays dependency-free and imports no host code; every new
Protocol method has a default so later additions remain additive. The loader's
pre-1.0 rule requires an exact major.minor match, so extensions written against
0.1 are now refused at startup with an actionable install hint rather than
loading into a host that implements a different surface.
uv.lock records the contract package's new version, so `uv sync --locked` still
resolves on a fresh checkout.
* fix(backend): sort gateway service imports
|
||
|
ChiHaYa
|
88252e9b31
|
fix(subagents): isolate background tasks from reused tool call IDs (#4758)
* fix(subagents): isolate background execution IDs * fix(subagents): preserve correlation scope and isolate usage * fix(subagents): make usage attribution idempotent |
||
|
Nan Gao
|
7389331e65
|
feat(extensions): observe task lifecycle and system model calls (#4684)
* feat(extensions): observe task lifecycle and system model calls PR 1 (#4636) gave extensions a middleware chain, and a middleware only sees what passes through the agent graph. Two runtime surfaces stay invisible to it: when a lead run or a subagent begins and ends, and the DeerFlow-owned model calls made outside the graph. This slice adds both, with no new Gateway surface -- routers, services, and the reference extension stay in PR 3. Contract (deerflow-extension-api 0.1.1) --------------------------------------- Two contribution kinds join `middlewares` on the registry: `task_lifecycle` (`on_task_start` / `on_task_stop`, receiving a `TaskInfo` and a conservative `TaskOutcome` of completed / aborted / failed) and `system_model_observer` (`on_system_model_call`, receiving a `SystemOperationKind`, a `SystemModelRequest` snapshot, and a `SystemModelResult` carrying either the response or the provider exception plus a duration). `SystemModelRequest.messages` normalizes to a tuple at construction. Goal evaluation and memory extraction pass a message list while title generation and summarization pass one prompt string, and a bare `str` already satisfies `Sequence` -- without normalization an observer iterating `request.messages` would silently walk characters. Copying also makes the frozen snapshot immutable in fact rather than only by declaration, since observations may run after the call site returns and keeps mutating its own list. Registry marks and rollbacks become per-bucket and positional, so an `install()` that fails after registering two different kinds cannot leave one of them behind. `needs_task_store` now covers all three kinds: a deployment that registers only lifecycle hooks still gets a task store. Task lifecycle -------------- The lead worker notifies start after the run has started and stop after completion persistence and the completion hook, but before clearing the finalizing barrier and publishing the stream end -- holding the barrier across stop is what keeps a same-thread replacement run from overlapping this task's lifecycle. Cancellation raised out of the stop notification is deferred, not propagated in place, so a cancelled run still clears the barrier and emits its end frame. A subagent with a parent `run_id` wraps its execution in the same pair inside `finally`, reporting `parent_task_id` so a delegation tree is reconstructable; a subagent without a `run_id` (embedded client, standalone LangGraph Server) logs and skips rather than inventing a parent. Contributors run in registration order inside one shared 3s budget and every failure is logged and failed open. System model calls ------------------ Four kinds cover the model calls the middleware chain cannot see: goal evaluation, memory extraction, title generation, and summarization. Each site reports both terminal paths without changing the provider exception the host observes, short-circuits on `has_system_model_observers`, and passes the live task store when the runtime has one (detached work gets an isolated store). The sync summarization half stays unobserved on purpose -- it and its only host caller are the sync side of an async-only runtime, so notifying there would block a thread on a call site the host never reaches; the reason is recorded at the call site. The DeerMem backend must stay vendorable and cannot import the extension API, so it reports through a new `MemoryCallbacks.on_memory_llm_result` host hook that the DeerFlow-side callbacks translate into an observation. Notification loop ----------------- Extension resources must be touched on the loop that created them, but subagents can execute on isolated loops and DeerMem runs on a worker thread. The Gateway registers its serving loop before any runtime dependency starts and resets it last through the exit stack, so every startup-failure and cancellation path is covered. Awaited hooks raised on another loop are dispatched across with `run_coroutine_threadsafe` and awaited under the same budget; synchronous sites submit fire-and-forget work. Shutdown stops accepting detached observations before the memory flush -- that flush runs on a worker thread and can emit memory observations -- while keeping the loop alive for awaited task hooks until run and subagent drain completes. Tests ----- `test_extension_task_lifecycle.py`, `test_extension_subagent_lifecycle.py`, and `test_extension_system_model_calls.py` cover ordering, fail-open, budget exhaustion, snapshot binding under a concurrent singleton replacement, the loop-dispatch and shutdown-suspension paths, and both terminal paths at every call site. `test_gateway_run_drain_shutdown.py` pins the stop-before-barrier and drain ordering. * fix(extensions): decide notification fail-open by origin, observe cancellation `_notify_each` only guarded `Exception`, so a contributor letting a `CancelledError` escape — an extension implementing an internal timeout with cancellation, say — skipped its successors and reached the worker's deferred-interrupt path, ending an otherwise successful run as cancelled. Fail-open is about where a failure came from, not its base class: only a genuine cancellation of the host task increments `Task.cancelling()`, so propagate on that and contain everything else. `KeyboardInterrupt` / `SystemExit` still propagate. `observe_system_model_call` skipped observers on cancellation for the same base-class reason, leaving goal / title / summarization silent on a terminal path that is routine — interrupt/rollback admission and shutdown both cancel the run task, with the provider tokens already spent. Awaiting observers there is unreliable (a repeated cancel interrupts that await before any of them runs), so report through the same non-blocking submission the synchronous memory bridge uses, then propagate the cancellation untouched. DeerMem keeps `BaseException` around its provider call, now with the reason recorded: that path runs on a worker thread, where cancelling the awaiting side never interrupts the running thread, so `CancelledError` cannot arrive at all. Its host-hook wrapper narrows to `Exception` — only the hook's own failures are non-fatal, and an observability path must not swallow a process teardown signal. * fix(extensions): warn on budget exhaustion, scope observer logs by task, propagate teardown Review response on #4684: - The memory observation bridge caught BaseException, which would swallow a teardown signal raised while dispatching; it now catches Exception, matching the boundary the DeerMem-side call site documents and tests. - A notification-budget timeout raised mid-hook fell into the generic hook-failure path and logged an asyncio-internal traceback; it now logs a warning like the pre-hook budget skip, while a TimeoutError a contributor raises on its own stays classified as a hook failure. - System model observer logs passed the operation kind as the task id, so log lines said "task goal/title/..."; they now carry the task scope id alongside the kind. |
||
|
Nan Gao
|
1f792d0f4b
|
feat(extensions): add middleware plugin foundation (#4636)
* feat(extensions): add middleware plugin foundation * fix(extensions): stop config resolution from masking extension loading `create_app()` resolved the configured plugin list inside the fail-open guard around `load_extensions()`. CI has no `config.yaml` (gitignored and never generated by the workflow), so `get_app_config()` raised `FileNotFoundError` there and was swallowed as an extension failure -- `load_extensions()` never ran at all, and the four `create_app()` tests in `test_extension_app_loading.py` passed locally but failed on every runner. Resolve the plugin list before the guard. Only an absent `config.yaml` is tolerated, mirroring `_resolve_trace_enabled_for_app_construction()`: `create_app()` runs at import time, and lifespan still performs strict config loading before serving. A `config.yaml` that exists but fails to parse or validate now propagates instead of being reported as an extension failure -- reporting it as the latter silently dropped a `required: true` extension rather than failing the boot. Make the tests config-independent with an autouse `stub_app_config` fixture, following the existing pattern in `test_gateway_lifespan_shutdown.py`, and cover both new branches of the config-resolution boundary. * fix(extensions): bind the run's extension snapshot through subagent delegation The lead-agent path resolves one immutable loaded-extension snapshot per run and binds it through task-store allocation and graph construction, but the subagent path re-read the process-wide singleton at execution time. In production both are the same object, yet a `set_loaded_extensions()` between the lead run's start and a subagent's execution (test teardown, a future hot-reload path) would let one run mix two extension generations — exactly what the documented invariant exists to prevent. The graph-build binding is a ContextVar scoped to synchronous construction, so it has already exited by the time a tool delegates; the snapshot has to travel through runtime context instead. The run worker publishes it under the host-internal `EXTENSION_SNAPSHOT_CONTEXT_KEY` (written after the caller merge, popped when the run has none, so a caller-supplied value is never authoritative), `task_tool` reads it back through the type-checking `resolve_run_extensions()`, and `SubagentExecutor` binds it at construction. Callers outside the Gateway run path — embedded `DeerFlowClient`, standalone LangGraph Server — install no snapshot and keep the existing `get_loaded_extensions()` fallback. * refactor(extensions): defer the ordering table by call, not by a lying tuple `CORE_ORDERING_CONSTRAINTS` was a `tuple` subclass that overrode only `__iter__` and resolved into a class-level `_resolved` side channel. A tuple cannot populate its own storage after construction, so the instance stayed the empty tuple it was built as: `len()` was 0, `bool()` was False, `in` was always False, indexing raised, slicing and `reversed()` came back empty, and it compared unequal to the plain tuples tests substitute for it — all while iteration yielded the real constraints. Only `assert_ordering` consumed it, and only by iterating, so the split went unnoticed. The sibling `_AnchorTable(dict)` uses the same idea soundly because dict is mutable: `self.update()` fills the real storage, making every inherited operation correct. That trick does not survive the port to an immutable type. Replace it with `core_ordering_constraints()`, matching how `stack.py` defers the same kind of table via `_anchors()`. The deferral is kept — it is about dependency direction, not just cycles: `extensions/` is the layer the middleware layer calls into, so a module-scope `agents.middlewares` import here points the dependency backwards and closes a cycle as soon as any middleware imports something under `extensions/` at module level. Resolution stays at `assert_ordering` time, which already runs inside the middleware builder. Tests pin both halves: the returned value is a plain tuple whose len/bool/ membership/indexing/reversal/equality agree with iteration, and a subprocess probe asserts importing `extensions.ordering` does not load the middleware layer while calling the function does. |
||
|
Aari
|
a5059b8284
|
fix(subagents): isolate callbacks and activate skills lazily (#4497) | ||
|
hataa
|
7857fa0cce
|
feat(authz): enforce tool authorization at assembly and runtime (#4370)
* feat(authz): enforce tool authorization at assembly and runtime * fix(middleware): guard deferred tool setup lookup (#4370) --------- Co-authored-by: Willem Jiang <willem.jiang@gmail.com> |
||
|
March7
|
8c78d1f41f
|
fix(subagents): load user-scoped skills (#4356) | ||
|
hataa
|
10890e10a8
|
feat(authz): propagate trusted authorization principal context (#4203) | ||
|
Nan Gao
|
de55982c5a
|
fix(subagents): preserve parent checkpoint namespace (#4215)
* fix(subagents): preserve parent checkpoint namespace * test(subagents): align stream isolation coverage |
||
|
Aari
|
42544755ac
|
fix(skills): escape untrusted skill metadata before it enters the model prompt (#4128)
* fix(skills): escape untrusted skill metadata before it enters the model prompt Skill name/description/allowed-tools come from the frontmatter of a user-installable .skill archive (POST /api/skills/install or a drop into skills/custom/); the parser only strips them. The slash-activation and durable-context siblings already html.escape these exact fields before rendering them into a model-visible block -- but five other render sites emit them raw. The sharpest is the default path, <available_skills> in the system prompt (skills.deferred_discovery: false): a community skill whose description closes the block can forge a framework-trusted <system-reminder> into the lead-agent system prompt. Driven through the real apply_prompt_template(), the forged tag reaches the system prompt raw on main and is neutralized here. Escape at every render site that emits untrusted skill metadata/content: - <available_skills> (name/description/location) and <disabled_skills> (name) in lead_agent/prompt.py; - describe_skill output (name/description/allowed-tools/location) and <skill_index> (name) in skills/describe.py; - the subagent <skill name=...> attribute plus the raw SKILL.md body in subagents/executor.py::_load_skill_messages -- its direct sibling skill_activation escapes both, this escaped neither. quote=False in element-text positions (matching skill_context and the #4097 correction), quote=True in the one attribute position (matching skill_activation). category is a controlled enum and is left as-is; escaping is render-time only, so stored skills are unchanged and re-rendering never double-escapes. * fix(skills): escape skill name in the slash-activation prose line The slash-activation reminder emitted `activation.skill_name` raw in its prose line while escaping the same value in the adjacent <skill name="..."> attribute. skill_name is grammar-gated to [a-z0-9-] by resolve_slash_skill before it reaches the renderer, so this is a defense-in-depth / consistency fix rather than a reachable injection: the two positions can never drift if a future caller builds an activation from an unconstrained name. Reuse the already-computed escaped_skill_name. |
||
|
Daoyuan Li
|
2bd0f56a0f
|
fix(subagents): classify recursion-capped LLM error fallbacks as failed (#4056)
The GraphRecursionError except-block in SubagentExecutor._aexecute derives usable_partial from the last AIMessage's raw non-empty text, without checking _extract_llm_error_fallback (#4042) first. A handled provider failure (LLMErrorHandlingMiddleware's deerflow_error_fallback marker) always carries non-empty user-facing text, so when it lands on the same turn that trips max_turns, it is indistinguishable from genuine partial output and gets misclassified as a completed task instead of the failed provider error it is. Consult _extract_llm_error_fallback in this except-block too, same as the normal-completion path above it, and classify FAILED with stop_reason=turn_capped when it detects the marker. |
||
|
Nan Gao
|
aafd5077b2
|
feat(subagents): show effective model and token usage on task cards (#4049)
* feat(subagents): show runtime metadata on task cards * fix(subagents): stop task-card render loop and dedupe model fetches Address code review on the runtime-metadata cards: - P1 render loop: the terminal ToolMessage is re-parsed on every MessageList render and always carries modelName/usage, so the presence-based setTasks condition fired a fresh state object each render -> "Maximum update depth exceeded". computeNextSubtask now returns a value-compared `changed` flag and a pure subtaskNotification() routes terminal transitions through the deferred after-render path while skipping no-op re-parses. - Per-card useModels refetch: add staleTime: Infinity to the ["models"] query so every subtask card shares one /api/models fetch instead of refetching on each mount. * make format * refactor(subagents): dedupe token-usage validators + tidy event narrowing Address PR review follow-ups: - DRY: extract one shared token-usage validator per side. Backend status_contract.normalize_token_usage() now backs both the terminal ToolMessage metadata and the subagent.step/.end run events (step_events.py), and frontend messages/usage.normalizeTokenUsage() backs both the live task_running event (lifecycle.ts) and the terminal ToolMessage metadata (subtask-result.ts). Prevents the input/output/ total_tokens validation from drifting across the four former copies. - Nit: onCustomEvent narrows event.type once instead of re-checking the object shape per branch; the redundant task_started early-return (already validated by taskEventToSubtaskUpdate) is dropped. |
||
|
Nan Gao
|
bbb3deb231
|
fix(subagents): classify LLM error fallbacks as failed (#4042)
* fix(subagents): classify LLM error fallbacks as failed (#4041) * fix(subagents): address #4042 review on LLM error fallback - clarify _extract_llm_error_fallback docstring: tail-only scan is safe because subagents append their own terminal message, so the last AIMessage is never a stale parent-history marker (cross-ref worker.py) - compute final_result and pop stop_reason only on the COMPLETED branch so the guard stop-reason pop no longer fires on the discarded FAILED path - note the error_detail/"LLM request failed" fallbacks are defensive; the middleware always populates a non-empty content - add regression test locking the stale parent-history marker invariant --------- Co-authored-by: Willem Jiang <willem.jiang@gmail.com> |
||
|
Ryker_Feng
|
ebc09ce130
|
feat(mcp): auto-promote deferred MCP tools from routing hints (#4019)
* feat(mcp): auto-promote deferred MCP tools from routing hints
When tool_search.enabled=true defers MCP tool schemas, PR1 routing hints
still require the model to spend a tool_search discovery round trip before
it can call the tool the routing metadata already points at. This adds a
McpRoutingMiddleware that matches the latest user message against PR1
routing keywords and promotes the matching deferred schemas before the
model call, removing that round trip.
Design (soft routing, opt-in, additive):
- Matches only the latest real HumanMessage (shared is_real_user_message
helper, reused by SkillActivationMiddleware so the two cannot drift);
case-insensitive substring match, no tokenizer dependency.
- Ordering: priority desc, then tool name asc; capped by the new global
tool_search.auto_promote_top_k (default 3, clamped 1..5). Does not add or
consume a per-tool auto_promote_top_k (PR1 schema unchanged); a per-tool
value is ignored with a DEBUG note.
- Returns a plain {"promoted": ...} state update (not a Command) and relies
on ThreadState.merge_promoted for union/dedupe, so auto-promote and a
model-triggered tool_search converge on the same catalog hash.
- Installed before DeferredToolFilterMiddleware on every deferred-tool path
(lead agent, subagent, embedded client, webhook via shared builders);
a construction-time assert rejects the reversed order. catalog_hash is
None / no routing index is a complete no-op, so bootstrap and ACP skip it.
- Privacy: never executes tools, never promotes policy-filtered tools, adds
no routing keywords or matched tool names to trace metadata or INFO/WARN
logs.
No behavior change when tool_search.enabled=false.
Tests: index construction, matching semantics, middleware state updates,
same-cycle deferred-filter interaction, lead/subagent/embedded-client
builder wiring + order invariant, config clamping, config.example.yaml
parseability, and privacy assertions.
* refactor(mcp): address auto-promote review nits
- executor: access app_config.tool_search.auto_promote_top_k directly to match
the lead-agent and embedded-client paths (drop the over-defensive getattr that
masked missing config); update the subagent test mock to carry tool_search.
- tool_search / mcp_routing_middleware: cross-reference the duplicated routing
priority/keyword normalization between the builder and the middleware's
defensive _normalize_index so they cannot silently drift.
- MCP_SERVER.md: document that auto-promote keyword matching is a case-insensitive
substring test (not word-boundary), advising distinctive keywords.
|