* fix: bound MCP server bring-up timeouts and exclude externalized tool outputs from delivery verification
Two related robustness fixes:
1. MCP server bring-up was unbounded. tool_call_timeout only covered
session.call_tool(); tool discovery (subprocess spawn + initialize +
tools/list) and persistent stdio session initialization could hang
forever, blocking agent construction (and on the Gateway event loop,
the whole process). Add a per-server session_init_timeout
(default DEFAULT_MCP_SESSION_INIT_TIMEOUT = 60s, null disables) that
bounds both discovery and pooled-session initialization. The session
pool's existing cancellation handling tears down a session stuck
mid-creation in its own task.
2. ToolOutputBudgetMiddleware externalizes oversized tool outputs into
outputs/.tool-results/ (configurable tool_output.storage_subdir). The
workspace-change scanner and run delivery verification counted those
files as produced artifacts, so any run that externalized a tool output
without also presenting a real artifact failed with
"Artifact delivery incomplete". Exclude TOOL_RESULTS_DIRNAME via a
shared constant (mirroring BROWSER_FRAMES_DIRNAME) and thread the
configured storage_subdir through snapshot capture so both
workspace-changes events and delivery verification stay clean.
* review: enforce single-segment tool_output.storage_subdir; document discovery-timeout cleanup
Address review feedback:
1. A custom tool_output.storage_subdir with a path separator (e.g.
cache/tool-results) silently no-oped the workspace-scanner exclusion:
os.walk yields one-segment dirnames, so a nested value never matched and
its files were counted as produced artifacts again. ToolOutputConfig now
validates storage_subdir as a single directory name (rejects separators,
.., absolute, empty) with tests, so the exclusion is always sound.
2. The discovery-timeout path now documents why cancellation is safe, mirroring
the session-init note: discovery runs inside the adapter's nested async
context managers, and stdio_client's finally terminates the process tree
(SIGTERM->SIGKILL on POSIX, process-tree on Windows), so a timed-out npx
subprocess and its children are reaped rather than accumulating.
* review: log session-init timeouts and align API response model default with runtime config
Address second-round review feedback:
1. A session-init timeout raised TimeoutError without any log, unlike the
discovery timeout which logs a WARNING. Wrap the bounded get_session in a
try/except that logs the timeout (server name + seconds) and re-raises, so
operators can diagnose tool-call failures caused by hung MCP sessions.
2. McpServerConfigResponse.session_init_timeout defaulted to None while
McpServerConfig defaults to 60s: a server created via PUT /api/mcp/config
without the field was persisted with null (no timeout) while the same
server created in the config file got 60s. Align the response-model default
to DEFAULT_MCP_SESSION_INIT_TIMEOUT so API-created and file-created servers
behave the same; an explicit null still opts out.
* review: narrow the discovery-timeout handler to the bounded wait_for path
The except TimeoutError clause covered both the bounded wait_for branch and
the bare discovery branch. With session_init_timeout opted out (None), a
TimeoutError raised by discovery itself would hit the %.1f format with None:
logging raises TypeError internally, the WARNING is silently dropped, and a
--- Logging error --- traceback goes to stderr.
Narrow the handler to wrap only the wait_for call, where the branch condition
guarantees the timeout value is not None. A discovery-internal TimeoutError on
the opted-out path now falls through to the generic failure handler and is
reported as 'tool discovery failed' with exc_info. Covered by a regression
test that asserts the skip is reported without any broken format.
* feat(channels): add Buzz (Nostr) channel connector
Adds a Buzz (https://github.com/block/buzz) channel so DeerFlow can join a
Nostr-relay workspace as a member: it answers @mentions in channels, replies
to DMs, and streams answers by editing one message in place.
* app/channels/buzz_nostr.py — pure NIP-01 helpers: canonical event ids,
BIP-340 signing/verification, chat/edit/auth builders, relay frames.
* app/channels/buzz.py — BuzzChannel: one NIP-42-authenticated websocket,
channel discovery (kind 39000) with one subscription per channel, live
membership tracking (44100/44101), per-channel replay watermarks, and
replies posted once then edited in place (kind 40003).
* app/channels/buzz_run_policy.py — same-thread serialization, mirroring
the Feishu precedent.
Inbound is gated in order: signature verification, self-drop, /connect
bind-and-return, pubkey allowlist, then mention / DM / mention-free /
thread-follow. Off by default; needs the new optional `buzz` extra
(coincurve, lazily imported), which detect_uv_extras resolves from
channels.buzz.enabled the same way it already handles channels.discord.
Two relay behaviours drove the design and are worth knowing when reviewing:
a global {"kinds":[9]} subscription receives nothing from buzz-relay and a
multi-value "#h" filter receives nothing either, so one REQ per channel is
required; and a single global `since` cursor skips quiet channels, so
watermarks are per channel.
Signed-off-by: Ajay R <ajayr@formbuddy.com>
* fix(channels): only publish assistant messages from the IM stream
`_accumulate_stream_text` decided what streamed `messages-tuple` payloads
become displayable text by rejecting ONLY payloads whose `type` contained
"tool", so it published everything else. DeerFlow writes hidden model
context into the messages channel as ordinary messages -- memory recall and
the rewritten user turn as hidden HumanMessages (DynamicContextMiddleware),
the `<durable_context_data>` block as another (DurableContextMiddleware) --
and LangGraph fans those state writes out on the messages stream, so they
reached every streaming IM channel as the assistant's reply.
Proved live on a Buzz relay: the connector published a `<memory>` fact block
and, in another run, a verbatim echo of the user's own inbound message.
Affects Feishu, Telegram, WeCom and Buzz; worst on Buzz, where each update
is an immutable public Nostr event that a corrective edit cannot unpublish.
Invert the filter to an allowlist of assistant message types. Two new pure
helpers keep it testable:
- `_stream_payload_type` resolves the type from both shapes the function
already handles: the `model_dump()` shape the gateway emits, and
LangChain's `to_json()` constructor shape whose own `type` is the literal
"constructor" and whose class name is the tail of the `id` path.
- `_is_assistant_stream_type` matches "ai"/"assistant" by PREFIX, not
substring -- ordinary words contain "ai" ("chain", "domain"), and a
substring test would admit a foreign type name by accident.
The bare-`str` branch is removed: an untyped payload cannot be attributed to
the assistant, nothing in DeerFlow produces one (serialize_messages_tuple
always emits `[message_dict, metadata]`), and a runtime that emitted raw text
deltas would emit hidden context the same way. Per-message-id buffering and
merging are unchanged.
Tests pin both directions, including multi-chunk merging across one message
id, so the allowlist cannot silently kill streaming, plus an end-to-end
`_handle_streaming_chat` test asserting the live payload never reaches an
outbound message.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Signed-off-by: Ajay R <ajayr@formbuddy.com>
* chore(helm): bump config_version to 33 in chart values and README
config.example.yaml moved to 33 for the buzz channel block; the chart's
embedded config example and its README copy track it (config_version only
drives the outdated-config warning, per scripts/check_config_version.sh).
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Signed-off-by: Ajay R <ajayr@formbuddy.com>
---------
Signed-off-by: Ajay R <ajayr@formbuddy.com>
Co-authored-by: Claude Opus 5 <noreply@anthropic.com>
* 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.
* feat(authz): enforce model authorization at Gateway routes and runtime (#4063 Phase 3)
Phase 3 / Models — the first of three resource-type PRs (Models, Skills,
Sandbox). The RBAC provider already maps "model" → config key "models"
(rbac.py _RESOURCE_POLICY_KEYS), so no schema change is needed.
Gateway route layer (mirrors Phase 2A):
- resolve_model_authorization() in authz.py returns (provider, principal),
reusing _get_cached_route_provider and build_principal_from_context,
including the INTERNAL_SYSTEM_ROLE → None pop for internal callers.
- list_models filters via provider.filter_resources(principal, "model", names).
- get_model checks provider.authorize("model", "use"). Deny → 403 (not 404,
since the model exists but the role lacks permission).
Runtime resolution layer (mirrors Phase 1B):
- _authorize_model_name() in agent.py runs after _resolve_model_name. On deny,
falls back to the first allowed model (RFC §9: graceful, not crash). All
models denied + fail_closed → ValueError (matches existing contract).
authorization.enabled: false is a complete no-op on both layers. Anonymous
requests (user=None) bypass filtering. 18 new tests + 314 existing tests pass.
* fix(authz): enforce model:use on the embedded DeerFlowClient path (Phase 3 follow-up)
Round 4 review (willem-bd): _authorize_model_name only covered the Gateway
runtime path (_make_lead_agent). The parallel lead-agent construction path
DeerFlowClient._ensure_agent (client.py) filtered tools but not the model,
so a library/embedded consumer with role-scoped model policies could run a
model the role is denied model:use for.
- Insert _authorize_model_name in _ensure_agent, mirroring _make_lead_agent.
- Resolve None default to the first configured model before the gate so the
implicit default (create_chat_model(name=None)) is also authorized.
- Update test_authorization_filters_framework_tools_and_reuses_provider: the
stub provider now returns an allow decision for model:use (checked during
assembly) and patches resolve_authorization_provider in the agent namespace.
- Add 3 DeerFlowClient._ensure_agent path tests (real-path fallback,
None-default resolution, disabled no-op); 24 tests total.
* docs(authz): document get_model provider-unavailable fail-open path + test
zhfeng review (round 5): get_model's docstring only mentioned the deny→403
path, not the provider-resolution-error + fail-open path (which allows the
request, mirroring list_models's documented fail-open semantics). The
behavior itself is correct and symmetric with list_models, but it was
undocumented and the _AuthorizationUnavailable path had no test coverage.
- Extend get_model docstring to state the provider-error fail-closed/fail-open
outcome, matching list_models's wording.
- Add test_get_model_provider_unavailable_fail_closed_vs_open exercising the
_AuthorizationUnavailable path (provider cannot be resolved at all), pinning
fail-closed→403 / fail-open→200.
---------
Co-authored-by: Willem Jiang <willem.jiang@gmail.com>
* fix(mcp): constrain stdio launcher args and env at the config API
The API-managed stdio allowlist validated only `command`. Since `npx`/`uvx`
exist to fetch and run code, allowlisting the command name alone named a
binary without constraining what that binary ran, so the restriction did not
match the intent stated in `_validate_mcp_update_request`'s own docstring:
reducing the blast radius of a compromised authenticated browser session.
Screen two more fields on the stdio path, shared by `PUT /api/mcp/config` and
the enable branch of `PATCH`:
- `_ARBITRARY_EXEC_ARGS` rejects the flags that make a launcher evaluate a
string (`-c`, `--call`, `-e`, `--eval`, `--print`, `--shell`, `--node-arg`,
`--node-options`), matching `--flag=value` as well as `--flag value`, for
every allowlisted command rather than just `npx`/`uvx`. `-p` is deliberately
absent: it means `--package` for npx and `--python` for uv.
- `_CODE_INJECTING_ENV_VARS` rejects env names that inject code at process
startup (`NODE_OPTIONS`, `LD_PRELOAD`, `PYTHONSTARTUP`, ...), which would
otherwise sidestep the argument check.
Remote transports skip both -- they spawn nothing. Rejection messages echo
only the normalized flag, never the caller's value.
This is defense in depth, not a trust boundary: `npx`/`uvx` fetch and execute
remote packages by design, so an admin can still point one at a package they
published. Gateway admin remains equivalent to code execution on the host, and
the code, backend/AGENTS.md, and README all say so explicitly to keep the
check from being mistaken for a boundary.
`npx`/`uvx` stay in the default allowlist: nearly every official MCP server is
`npx -y @modelcontextprotocol/server-*`, so dropping them would break the
primary UI flow for adding one.
Tests cover each rejected flag spelling, case-insensitivity, `python -c` under
an extended allowlist, env injection, and the PATCH enable path, plus nine
positive cases pinning that real-world `npx`/`uvx`/`python -m` invocations
still validate.
* fix(mcp): screen PYTHONPATH and interpreter short-flag clusters
Review follow-up on the stdio launcher screen.
`PYTHONPATH` bypassed the env denylist on the *default* allowlist: `site`
imports `sitecustomize.py` from any `sys.path` entry before the tool's
entry point runs, so a caller-controlled directory is code execution
under plain `uvx` (verified against the real launcher). `PYTHONHOME` is
the same class via a repointed stdlib. Both are now rejected.
`PYTHONSTARTUP` is inert for non-interactive launchers and stays only as
belt-and-braces, now documented as such.
`-p` was exempted for every command, but it is node's `--print` --- the
long spelling was blocked while the short one passed once an operator
extended the allowlist. Whole-token matching also missed combined short
options (`node -pe`, `perl -we`, `python -Ic`). Both rules now apply to
commands outside `{npx, uvx}`, which are interpreters rather than package
runners; scoping them that way leaves the default allowlist unchanged,
since npx/uvx do not cluster short options and their trailing arguments
belong to a third-party server's own CLI where `-name` is ordinary.
`LD_LIBRARY_PATH`/`DYLD_LIBRARY_PATH` are recorded as an accepted
residual: conditional on the process loading a shadowable library, and
legitimately set by native-dependency servers.
Still defense in depth, not a trust boundary.
* docs(mcp): record NODE_PATH as an accepted search-path residual
Reviewer asked whether NODE_PATH belongs in _CODE_INJECTING_ENV_VARS for
parity with NODE_OPTIONS/PYTHONPATH. It does not, and the mechanism in the
report does not hold: verified on node v22, NODE_PATH is searched *after* the
local node_modules chain (the resolver unshifts the requiring module's paths
ahead of it), so it cannot shadow an installed dependency, and ESM import
ignores it entirely. It can only supply a CJS module that would otherwise
fail to resolve.
That puts it with LD_LIBRARY_PATH as a conditional search path, not with the
unconditional startup execution every entry in the set provides. Widen the
residual note from native-library shadowing to the general class, name
NODE_PATH in it, and pin the boundary with a test so it reads as a decision.
* fix(mcp): scope the stdio arg screen to the launcher's option region
`_arbitrary_exec_arg` inspected every element of `args`, justified by the claim
that the launchers in scope treat the token after `--` as a package name. That
is inaccurate: only the *first* token after `--` is the package name, and with
or without the separator every token from the package name onward is handed to
the third-party server's own argv, where `-c` is routinely "config" and `-e`
"env". So `npx -y @scope/server -c config.json` was rejected even though npm
never parses that `-c`, and the positive cases stayed green only because they
were picked to dodge the denylist. Over-blocking, not a bypass -- a regression
against the PR's own "the primary UI flow must keep working" goal.
Screen the launcher's own option region instead. Finding that region needs each
launcher's option *arity*, because a value is not a positional:
`npx -p <pkg> -c '<command>'` runs the command -- `-p` is `npm exec`'s
`--package` (it overrides the global `--parseable` shorthand), so `<pkg>` is
its value and npm keeps parsing its own flags. Ending the region at the first
non-flag token, the obvious rule, walks straight past that and turns an
over-block into a real eval bypass. The tables are therefore generated, not
guessed: `_NPX_BOOLEAN_ARGS` from `@npmcli/config` (npm 10.9.4) with the `-p`
override applied, `_UVX_VALUE_ARGS` from `uvx --help` (uv 0.11.1).
The unknown-option default is deliberately opposite per launcher, following the
exec set rather than symmetry. npx owns real exec flags, so an unknown option
consumes a value and keeps the region open; npm errors on options it does not
define, so that direction cannot reject an invocation that would otherwise
work. uvx owns no string-eval flag at all, so its screen is a tripwire rather
than a control, and an unknown option consumes nothing -- which keeps uv's
large and growing boolean surface from over-blocking. uvx also drops the short
spellings from its exec set, since `-c` is uv's `--constraints <file>`: an
ordinary documented option the old screen rejected outright.
Short options are matched case-sensitively now, because their case selects a
different option -- npm reads `-C` as `--prefix`, and folding it onto `-c`
rejected it. Long spellings stay case-insensitive: npm accepts `--CALL` and
runs it. Commands outside the package-launcher table are interpreters rather
than package runners and keep the unchanged whole-args screen, including the
`-p` rule and cluster decomposition.
Verified against the real launchers rather than by reading their docs. For
every argument vector in the new tests the validator's verdict matches whether
`npx` actually executes the string or passes it through to the server, and the
bypass-guard cases were confirmed to fail under the naive first-non-flag-token
boundary before being pinned.
Reported by @willem-bd on #4617. Two further false positives found while
confirming it are covered here: `uvx -c constraints.txt <tool>` above, and
`docker run -i --rm -e KEY <image>` -- the canonical GitHub MCP server
invocation -- which an operator who extends the allowlist with `docker` still
hits, since the conservative non-package-launcher path is unchanged.
---------
Co-authored-by: Willem Jiang <willem.jiang@gmail.com>
* fix(gateway): unify thread ID validation at the API boundary
Thread ID entry points accepted arbitrary strings while downstream
consumers (filesystem paths, Kubernetes Provisioner, JSONL event store)
each enforced different character restrictions, so invalid IDs were
persisted first and only failed later during sandbox/workspace init.
Centralize validation in deerflow.utils.thread_id (pattern
^[A-Za-z0-9_-]{1,64}$): validate at routers, RunCreateRequest,
scheduler dispatch, paths.py, JSONL store, embedded client, and align
the Provisioner pattern (pinned by a parity test). UUIDs are still
generated only when no ID is supplied; caller-supplied opaque IDs stay
supported.
Deliberate exceptions: DELETE /threads/{id} keeps str as the legacy
cleanup escape hatch (filesystem cleanup guarded), read-only
client.get_thread stays unvalidated, and scheduler rows with legacy
invalid IDs record a failed dispatch instead of raising out of the
poll loop.
* docs: document canonical thread ID contract
README: caller-supplied thread IDs need not be UUIDs; the canonical
pattern and per-endpoint behavior. AGENTS.md: the shared
deerflow.utils.thread_id contract, its enforcement boundaries, and the
legacy-ID escape hatches.
* fix(gateway): close thread ID validation gaps at remaining entry points
Follow-up to the canonical thread ID contract: a full audit found the
uniform-422 coverage only reached about half of the thread_id surfaces.
- routers: 18 routes still took a bare thread_id: str — 13 in
thread_runs.py (including the five messages/events/workspace-changes
reads that returned 500 on the JSONL event store vs 404/empty on the
DB store), 4 read routes in threads.py, and the suggestions route
flagged in review. DELETE /api/threads/{id} keeps str as the declared
legacy-cleanup escape hatch.
- client: upload_files/delete_upload/list_uploads/get_artifact now
validate up front, fulfilling the RFC's 'all mutating entry points'
clause (get_thread stays unvalidated as the declared legacy read path).
- tui: the /resume literal-ref fallback validates against the canonical
contract and reports a descriptive error instead of failing deep in
the client.
- scripts/support_bundle.py: replace the drifted dot-allowing pattern
with a byte-identical copy of THREAD_ID_PATTERN (kept local so the
script still runs with a broken venv).
* test(gateway): guard the canonical thread ID contract against regressions
- test_thread_id_route_contract.py: static AST sweep asserting every
route handler with a thread_id parameter annotates ThreadId
(whitelist: the DELETE escape hatch), plus a runtime sweep hitting
all 44 thread_id routes with a non-canonical ID and asserting a 422
that names thread_id, plus a websocket upgrade-rejection case.
- test_thread_id_validation.py: client entry-point validation,
support_bundle pattern parity, and TUI literal-ref fallback tests.
- Align two tests that encoded the old contract (dotted IDs).
* feat(artifacts): inline editing for text artifacts in the panel
Add a PUT /api/threads/{id}/artifacts/{path} endpoint that atomically
replaces an existing UTF-8 text file under /mnt/user-data/outputs after
verifying its SHA-256 revision. Active runs conflict (409); binary,
symlink, oversized, and non-output paths are rejected.
Frontend: edit/save/discard buttons, draft state with conflict detection,
CodeEditor onChange/onSave, loader SHA-256 from ETag, i18n, beforeunload guard.
Backend: PUT endpoint with thread reservation, atomic temp-file replacement,
sandbox sync for non-mounted providers, rollback on failure, ETag on GET.
Tests: 8 backend + 1 blocking-IO + 3 frontend test files.
* fix(artifacts): scope replacement permissions and release sandboxes
---------
Co-authored-by: Willem Jiang <willem.jiang@gmail.com>
* feat: show real-time context window usage in chat UI (#3125)
Adds a `context_usage` block to `GET /api/threads/{id}/token-usage`
(token count from the live checkpoint, the thread model's
`context_window`, and a percentage), introduces a new
`ModelConfig.context_window` distinct from the per-call `max_tokens`
output cap, and surfaces the percentage in the chat header — inside
`TokenUsageIndicator` when token-usage tracking is on, or as a
standalone badge when it's off so context capacity stays visible
independent of cost tracking.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
* feat: per-category breakdown for context window usage
Replace the single-number context_usage payload with a Claude-Code-style
breakdown — messages, system prompt, skills, system/MCP tools (active +
deferred), custom agents, memory injection, autocompact buffer, and free
space — and surface it in the chat UI with a segmented progress bar and
per-row table.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
* docs(config): document context_window across model examples
Add `context_window` to every example model in config.example.yaml so the
new chat-UI "% context used" indicator works out of the box for whichever
example a user adopts. Each value is the published default at the time of
writing; users are pointed at the official model spec to verify. Bumps
config_version to 11 so `make config-upgrade` flags outdated user configs.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
* style: ruff format (line-length 240)
No behavior change — collapses two multi-line expressions that fit on
one line under the project's 240-char limit. Picked up by `make format`.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
* review: address Copilot bot comments on #3183
- token-usage-indicator: switch `{contextPercentage && (...)}` to an
explicit `!= null` check. (The string `"0"` is actually truthy in JS so
the original code wasn't buggy, but the explicit check is clearer.)
- context-usage-breakdown: drop the `useMemo` around segments/totals — the
computation is O(n) over a handful of rows and the previous memo deps
omitted `t.contextUsage.categories`, so the bar's tooltips/aria-labels
could stay in the old language after a locale switch.
- context_usage._split_tools: snapshot MCP names from
`get_cached_mcp_tools()` directly instead of re-reading
`extensions_config.json` after `get_available_tools()` already loaded
it. Removes redundant file I/O on every `/token-usage` poll.
(`get_available_tools()` still emits its own INFO logs — silencing
those is out of scope here.)
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
* style(frontend): prettier --write context-usage-breakdown
CI's `pnpm format` (prettier --check) caught two lines previously
formatted by hand. Collapses one comma to fit on one line; no behavior
change.
Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>
* fix(gateway): correct context-usage breakdown + add exact token counting
The context-usage indicator shipped two bugs that silently zeroed whole
breakdown rows (both caught by try/except, so the feature looked alive but
produced wrong numbers):
1. _count_system_prompt passed app_config= to get_deferred_tools_prompt_section,
which only accepts deferred_names -> TypeError swallowed -> system_prompt
row always 0, and used_tokens/percentage undercounted by the full prompt.
Also subtracted the deferred section twice (the rendered prompt already
excluded it). Fix: derive deferred names deterministically and pass them to
apply_prompt_template; drop the redundant subtraction.
2. _split_tools imported a non-existent get_deferred_registry -> ImportError
swallowed -> all four tool-category rows always 0. Fix: classify via the
public is_mcp_tool predicate + tool_search.enabled (mirrors
build_deferred_tool_setup); the MCP tag is set by get_available_tools.
Added token_usage.counting (approximate|exact). 'exact' routes text/schema/
message counting through the model tokenizer (tiktoken cl100k_base) via the
existing memory-module machinery (lazy load + cache + cooldown + CJK-aware
fallback), so CJK-heavy threads stop being undercounted by chars//4.
Regression + e2e tests added; 6621 backend tests pass.
* fix(gateway): harden context usage accounting
* fix(gateway): count promoted MCP tools as active in context usage
Promoted tools (deferred MCP tools the thread has fetched via tool_search)
have their full schema bound on every subsequent turn by
DeferredToolFilterMiddleware, so they consume context like any active tool.
The breakdown previously left them in the reserved *_deferred rows, under-
counting the thread's used_tokens.
Classification now treats a tool as deferred only when tool_search is enabled,
it is MCP-sourced, AND it has not been promoted. The promoted set is read from
the checkpoint's channel_values and scoped by catalog hash — matching the
runtime middleware, so a stale promotion from MCP-config drift cannot inflate
the active count.
The static system prompt still lists all deferred tool names (promotions only
affect schema binding, not the prompt), so _count_system_prompt's deferred
rendering is intentionally left unchanged.
8 new tests cover classification, catalog-hash scoping (match / drift /
compute-failure / malformed), and checkpoint extraction.
* fix(context): address review feedback
* fix(context): count structured message payloads
* fix(context): harden usage accounting
* fix(config): bump schema for context usage fields
* refactor: narrow context usage to core indicator
---------
Co-authored-by: Claude Opus 4.7 <noreply@anthropic.com>
Co-authored-by: Willem Jiang <willem.jiang@gmail.com>
* fix(memory): reject duplicate facts inside the create critical section
memory_add's duplicate check ran outside the storage critical section
(the tools.py comment called this out): two concurrent tool calls for
the same user could both pass the check and both store the same fact.
Move authoritative duplicate rejection into
MemoryUpdater.create_memory_fact: the candidate's normalized content
key is checked against the fresh memory snapshot on every
revision-conflict retry, so the loser of a concurrent create reloads,
sees the winner's fact, and is rejected with ValueError("Duplicate
fact"). The tool-layer pre-check stays as a fast path.
The REST router now maps the duplicate ValueError to 409 with a clear
detail instead of the misleading "content cannot be empty" 400.
Tests: backend-level duplicate rejection, a simulated
concurrent-commit-during-conflict-retry race, and the router 409
mapping.
* fix(memory): retry legacy-path fact creation on save conflict
Wrap the legacy single-file save in the same bounded conflict-retry
loop as the apply_changes path: reload the fresh snapshot and re-run
_raise_if_duplicate_fact_content on every retry, so a concurrent
duplicate commit is rejected with ValueError("Duplicate fact") (tool
error / REST 409) instead of the generic OSError save failure
(REST 500). No-duplicate conflicts now retry and store instead of
failing the create.
Addresses review feedback on bytedance/deer-flow#4599.
* fix(sandbox): project enabled skills into sandbox views
* fix(skills): keep projection mutations consistent
* fix(skills): fail closed on projection errors
* fix(skills): isolate per-scope failures during boot projection rebuild
rebuild_all_skill_projections() propagated any exception from the public
rebuild or from a single user's rebuild straight out of the gateway
lifespan startup, uncaught. A single broken user directory (bad
permissions, corrupted _skill_states.json, unreadable content) would
therefore abort gateway boot for every user, not just that one -
_rebuild_*_locked already fails closed internally (clears the view and
re-raises), so the boot loop only needed to stop treating that re-raise
as fatal.
Each scope's rebuild now fails closed independently and boot continues;
a scope left empty by a boot failure self-heals on the next sandbox
acquire via ensure_skill_projections().
Also patches deerflow.skills.projection.rebuild_all_skill_projections in
the memory-flush lifespan test fixture, matching the two sibling
fixtures in the same file — this call is now on the lifespan startup
path and the fixture's minimal SimpleNamespace config predates it.
* test(skills): update authz test for the projection-aware public toggle
_persist_shared_skill_state (introduced earlier in this branch) reads
the shared extensions_config.json fresh from disk under the projection
lock instead of through the cached get_extensions_config() singleton -
that's the whole point of the fix (stale worker caches must not clobber
another worker's concurrent update). The name no longer exists on the
skills router module, so the test's monkeypatch of it started raising
AttributeError instead of exercising the endpoint.
The mock storage in this test isn't a real LocalSkillStorage instance,
so _persist_shared_skill_state's projection-mutation branch is already
skipped (nullcontext) and it falls back to a fresh ExtensionsConfig()
for the nonexistent tmp config_path - no replacement monkeypatch needed.
* fix(sandbox): make skill projection ensure best-effort in acquire
acquire() called _ensure_skills_projection() directly, outside any
try/except, in both LocalSandboxProvider and AioSandboxProvider. Every
other skill-mount setup path in these providers has always caught
exceptions and logged a warning rather than failing sandbox acquire
outright (e.g. when config.yaml can't be resolved) - these two new call
sites broke that contract, so any projection failure (including simply
not having a config.yaml, as in CI's test environment) now failed
acquire() itself instead of just leaving skill mounts off.
_ensure_skills_projection now catches its own exceptions and returns
None; both providers' callers already tolerate that (a None projection
skips the skill-specific mounts, matching the existing degrade path)
after making _append_public_skill_mapping and the custom/legacy mount
block in LocalSandboxProvider explicitly None-safe.
Caught by running the full suite with config.yaml removed, matching
CI's environment - not caught locally because a real config.yaml was
present, masking the failure.
* fix(sandbox): make E2B skill projection mounts best-effort
_skill_projection_mounts called ensure_skill_projections with no guard,
unlike Local/AIO's _ensure_skills_projection. A raise propagated out of
_apply_mounts before the configured-mounts loop ran, so a skills
projection failure dropped the operator's own configured mounts too -
only caught by create()'s outer warning, with nothing applied at all.
Swallow here and return an empty mount list on failure, matching the
Local/AIO pattern: still fail-closed for skills, but no longer widens
the blast radius to unrelated configured mounts.
Review feedback from PR #4178.
* docs(skills): document projection trade-offs flagged in review
- _update_tree_digest: note the metadata-only (not content) hashing
trade-off and why runtime writes through this codebase are still
covered regardless (rebuild-under-lock + rename always changes inode).
- LocalSandboxProvider.acquire: note the acquire-time self-heal cost
(cheap on a fresh manifest, ~400ms rebuild under lock on stale/drift).
- skill_projection_mutation: drop the no-op except-Exception-then-raise;
a raise from the mutation already propagates past the yield with the
view left cleared, no explicit re-raise needed.
- provisioner README: spell out that hostPath skills volumes require
the gateway and K8s node to share DEER_FLOW_HOST_BASE_DIR (single-node
or shared storage), and that the custom/legacy volumes' hostPath type
Directory (not DirectoryOrCreate) makes a violation of that assumption
a visible Pod-creation failure instead of a silent empty mount.
Review feedback from PR #4178.
* fix(skills): lazily repair user projections
* fix(skills): close projection review gaps
* fix(skills): refresh user projection enable state
* fix(skills): close projection review follow-ups
* fix(skills): preserve state across projection writes
---------
Co-authored-by: Willem Jiang <willem.jiang@gmail.com>
* feat(runtime): seed empty run-event feed from checkpoint history
Threads created before the journaled run-event model hold their history
only in the LangGraph checkpoint. Before the first journaled run, backfill
an empty run-event message feed from the existing checkpoint head so
legacy history receives earlier thread-global seq numbers and remains
visible in the unified feed. Threads with no checkpoint or an already
populated feed skip the path.
The seed guard resolves the user explicitly instead of relying on the
store's AUTO default, which raises without a user contextvar (scheduler
launch path on the DB event store).
* docs: document checkpoint history seeding in thread runs
Before the first journaled run, an empty run-event message feed is
seeded from an existing checkpoint head so legacy checkpoint-only
history stays visible with earlier thread-global sequence numbers.
* fix(gateway): make checkpoint-history seed guard thread-scoped
The emptiness guard filtered by the current user whenever one was in
context, answering "does this user have any messages?" rather than
"has this thread's feed ever been journaled?". Seed rows stamped with
a different principal (NULL for ownerless seeds, or another user on a
shared NULL-owner thread) were invisible to the guard, so each new
principal re-seeded a duplicate history. Pass user_id=None
unconditionally; None also opts out of AUTO resolution, so the
ownerless scheduler path still cannot raise.
Adds a DbRunEventStore-backed regression test (the MemoryRunEventStore
tests cannot catch this — the memory store ignores user_id) proving the
ownerless-seed -> authenticated-run sequence seeds exactly once.
* fix(scheduler): retain launched run when post-launch bookkeeping fails
`dispatch_task()` created a `queued` task-run row, then `_launch_run()`
returned a live `run_id`, and only afterward did the queued->running
bookkeeping (`update_status` + `update_after_launch`) run. When that
bookkeeping raised on a transient DB error, the `except` handler marked the
task-run `failed` with `last_run_id=None`. Because `failed` is outside the
partial unique index `uq_scheduled_task_run_active`, this released the
task's single active slot: the next dispatch cycle could no longer see the
still-live run and launched a duplicate. The launched `run_id` was also
dropped, breaking later recovery / reconciliation / cancellation.
Track `launched_run_id` / `launched_thread_id`, set only after `_launch_run`
returns. In the `except` handler:
- If launch already succeeded, keep the task-run row `running` (so it keeps
holding the active slot and no duplicate launch can occur) and persist the
launched `run_id` on the parent task for retention. The bookkeeping
retries are best-effort with logging; if they fail too the row stays
`queued`, which is still active and still holds the slot, so we still
report the run as launched.
- If launch itself failed (no live run was created), behave as before:
mark the task-run `failed` and release the active slot.
The overlap-skip branch is now guarded by `launched_run_id is None` so a
run that already launched can never be reclassified as a skip / failed.
Adds a stateful regression test (`test_post_launch_bookkeeping_failure_does_not_release_active_slot`)
that injects a failure on the queued->running write and asserts a second
dispatch does not launch another run (`launch_count` stays 1) while the
first `run_id` is retained on the task-run row. The test is verified to
fail on `main` and pass with this change. A complement test pins the
pre-launch-failure path (launch itself raises) to ensure the slot is still
released when no live run exists.
Fixes#4452
* style(scheduler): apply ruff format to fix lint-backend CI
Reformat the two files touched by the previous commit with
`ruff format` (line-length=240 config joins the hand-wrapped
condition/log lines). No semantic change.
Fixes the `lint-backend` CI failure on PR #4504.
Co-Authored-By: Claude <noreply@anthropic.com>
* fix(scheduler): key retention on launch_succeeded flag
The previous invariant keyed the retention branch off
`launched_run_id is not None`, but the assignment
`launched_run_id = result["run_id"]` is itself post-launch code that
can raise (KeyError/TypeError on a malformed _launch_run result).
In that case launched_run_id stays None and the dispatch falls through
to the pre-launch generic-failure path, marking the task-run row failed
and releasing the active slot -- even though a live run was just
created (same class of bug as #4452, narrower trigger).
Flip a `launch_succeeded` flag immediately after `await _launch_run(...)`
returns, before any further code that can raise, and key both the
overlap-conflict guard and the retention branch off that flag.
Add a regression test with a malformed launch result (missing run_id):
the dispatch reports outcome="launched", the row stays running, and a
second dispatch does not launch a duplicate.
Addresses willem-bd review point 1 on #4504.
Co-Authored-By: Claude <noreply@anthropic.com>
* fix(scheduler): don't surface bookkeeping transient as task last_error
In the post-launch retention path the parent task's last_error was set
to the bookkeeping exception -- an infrastructure-level transient, not
a run-level failure. Between the failed bookkeeping write and the run
completing, the task list showed an error on a task whose run was
actively running.
Clear last_error (like the success path's clear-on-launch model): the
run's real terminal outcome is written by handle_run_completion, and
the transient itself is already recorded via logger.exception.
Assert in the retention regression test that the parent task update
carries last_error=None.
Addresses willem-bd review point 2 on #4504 (taking the drop option).
Co-Authored-By: Claude <noreply@anthropic.com>
---------
Co-authored-by: now-ing <24534365+now-ing@users.noreply.github.com>
Co-authored-by: now-ing <now-ing@users.noreply.github.com>
Co-authored-by: Claude <noreply@anthropic.com>
The console cost estimator summed per-run spend across every priced model
without checking that they shared a currency, so a deployment that priced
one model in CNY and another in USD produced a meaningless aggregate under
a single currency label. Track the first priced currency and, on a
mismatch, log a warning and drop all pricing so cost/currency fields report
null instead of an invalid sum. Currency codes are now trimmed and
case-normalized so equivalent spellings don't false-trip the guard.
* fix(skills): offload update_skill's blocking IO and share the config write lock
Re-applied on top of main rather than merged: update_skill has since gained
admin gating, user-scoped storage and a PUBLIC vs CUSTOM/LEGACY split, so the
offload is applied per path.
- PUBLIC: the extensions_config.json read-modify-write (path resolve, snapshot,
merge, write, reload) moves to a worker thread via asyncio.to_thread. The
payload is built from a snapshot so the cached singleton is never mutated in
place while the write is still in flight.
- CUSTOM/LEGACY: set_skill_enabled_state is offloaded; the non-user-scoped
fallback takes the same shared-file RMW path as PUBLIC.
- Both load_skills calls (and storage construction) are offloaded.
The RMW lock now lives next to reload_extensions_config as
get_extensions_config_write_lock() and is acquired by both the skills router and
the MCP router, which performs the same read-modify-write on the same file.
Previously each router held its own module-local lock, so once both sides
offloaded, a PUT /api/mcp/config could run inside a skill toggle's read->write
window and the later write would silently drop the other's change.
The lock is keyed by the running event loop rather than being a module-level
singleton: asyncio primitives bind to the first loop that awaits them, which
makes a plain module-level lock unusable in a process that runs more than one
loop.
Adds a cross-router regression anchor asserting a skill toggle and an MCP update
never overlap inside the RMW (max in-flight 1); it observes 2 when the routers
use separate locks.
* fix(config): own the extensions_config RMW lock from the worker thread
An asyncio.Lock held around `await asyncio.to_thread(...)` protects only the
awaiting task. If that task is cancelled the context manager releases the lock
immediately while Python keeps running the worker thread, so a second skills or
MCP writer could acquire it and operate on extensions_config.json concurrently
with the first worker — reopening the lost-update window this was meant to close.
The per-event-loop keying had a second hole: writers on different loops got
different locks and so did not exclude each other at all.
Replace it with a process-wide threading.Lock acquired *inside* the worker that
performs the RMW (`_write_extensions_skill_state` and `_apply_mcp_config_update`),
so ownership belongs to the thread doing the writing and is held until the write
and reload actually finish, regardless of what happens to the caller. A
threading.Lock also has no event-loop affinity.
Adds a cancellation regression: the skills worker is paused inside the lock, its
route task is cancelled, and the MCP writer is started — the MCP RMW must not
enter until the skills worker is released. Against the previous asyncio-lock
design this test fails with ['skills-enter', 'mcp-enter'].
The two existing serialization tests instrumented by replacing the worker
functions, which now bypasses the lock under test; they instead patch inside the
real workers (each module's reload_extensions_config, the last step under the
lock) so the production lock is exercised.
---------
Co-authored-by: ly-wang19 <ly-wang19@users.noreply.github.com>
Co-authored-by: Willem Jiang <willem.jiang@gmail.com>
* fix(runs): show run duration once per run and label it as elapsed work
turn_duration is the run's wall-clock lifetime (updated_at - created_at),
but both message endpoints stamped it onto every AI message of the run
and the frontend rendered each stamp as 'Thought for X seconds' — the
same number repeated per message, and tool-wait time presented as model
thinking latency (#4152).
- share one stamping helper between list_messages and list_run_messages:
only the run's final non-middleware AI message carries turn_duration
(list_run_messages already tried to do this but iterated reversed()
without stopping at the first match)
- completed-state trigger copy becomes 'Worked for X seconds' since the
number includes tool execution, not just thinking
Fixes#4152
* fix(threads): stamp turn_duration once per run in /history replay too
get_thread_history stamped turn_duration on every AI message of a run
instead of only the last one, so a run with several AI messages (e.g.
a tool call followed by a final answer) showed the same duration badge
more than once on reload.
Rebasing onto main picked up #4118, which replaced the /history
duration path with a checkpoint-metadata fast path plus an
event-store/run-manager fallback for legacy checkpoints - both of
which stamped every AI message per run_id, reintroducing this same
bug on both paths. This commit folds the once-per-run stamping
(stamp_turn_duration_on_last_ai) into both paths instead of only the
legacy one, and narrows the fast-path/fallback boundary to a per-run
completeness check (turn_run_ids - checkpoint_run_durations) instead
of a per-message one, so a run whose duration is already in checkpoint
metadata never re-triggers the event-store correlation fallback just
because its non-final AI messages correctly have no turn_duration.
* fix(frontend): stop caching a client-measured turn_duration per message
A run can produce several standalone content-only AI messages (e.g. a
subagent handoff), each briefly becoming the newest message and each
mounting its own Reasoning timer via the shared turnStartTime. Wiring
onTurnDurationChange let that per-message timer cache and display a
"Worked for X seconds" badge the moment a later message superseded it
- with a different, premature number, since the timer only measured
that message's own streaming window, not the run's total elapsed
time. That reintroduced the #4152 duplicate-badge bug through a path
the backend fix (once-per-run stamping) doesn't cover.
cachedDuration now only ever mirrors a backend-confirmed
turn_duration (already handled by the other effect here), so a
superseded message can no longer display a stale or wrong duration.
* fix(frontend): use 'Worked for' framing even without a persisted duration
The completed-state copy diverged depending on whether turn_duration
had landed yet: "Thought for a few seconds" (no duration) vs. "Worked
for X seconds" (duration present). Both describe the same completed
run, just with or without a persisted number, so the framing
shouldn't disagree on what happened.
* docs(runs): note why the middleware skip is inert on /history, add test
stamp_turn_duration_on_last_ai's middleware-caller skip only has an
effect on the event-store message shape (metadata.caller); checkpoint
messages replayed on /history never carry that field. Document why
that's not a gap - middleware writes go to thread metadata, not the
messages channel, so no middleware message ever reaches a
checkpoint's messages list.
Also lock the middleware-skip contract on list_thread_messages with
its own regression test, mirroring the existing one for
list_run_messages - the thread feed only gained this skip through the
same shared helper, and previously stamped every AI message,
middleware included.
* chore(frontend): retain current run duration UI
The frontend behavior was superseded by #4348; keep the latest main implementation while retaining this PR's backend fix.
* test(frontend): retain current reasoning coverage
Align the old PR test with the frontend implementation already merged in #4348.
---------
Co-authored-by: fancyboi999 <fancyboi999@users.noreply.github.com>
A browser client served from a different origin than the Gateway never
learns the id of the run it just created, so a brand-new thread keeps its
placeholder route for the whole session and every action gated on an
established thread — edit and rerun, regenerate, branch — stays hidden
until the page is reloaded.
Run-creating routes return the run's id in `Content-Location`, and the
LangGraph SDK resolves run metadata from that header alone. It is not
CORS-safelisted, so a cross-origin response hides it from JS unless the
server lists it in `Access-Control-Expose-Headers`. `useStream`'s
`onCreated` therefore never fires and the app cannot rewrite its route.
Expose it. `GATEWAY_CORS_ORIGINS` is a supported deployment mode, so the
CORS middleware has to carry everything that mode needs to read. Same-origin
nginx deployments are unaffected because CORS never applies to them.
* feat(checkpoint): make delta snapshot_frequency configurable
* fix(config): carry legacy checkpoint_delta_snapshot_frequency with warning
Addresses review on #4516: the rename from the flat
database.checkpoint_delta_snapshot_frequency key to nested
database.checkpoint_delta.snapshot_frequency silently dropped the old
value (pydantic extra="ignore"). Add a before-validator that maps the
legacy key onto the nested one with a deprecation warning (nested key
wins when both are set), plus a CHANGELOG breaking-change note covering
the rename and the 1000 -> 10 default change.
* fix(checkpoint): validate frozen snapshot frequency
* feat(lark): sidecar credential broker for sandbox lark-cli (Pattern B)
Removes the plaintext Lark credential mounts (appSecret + OAuth tokens)
from the sandbox container. A long-running broker sidecar owns lark-cli
and the per-user config/data dirs and serves the command surface over
Pod loopback; the sandbox gets only a forwarding shim on PATH, so the
raw credential files never exist in the sandbox filesystem.
- lark_broker.py: stdlib-only loopback broker (argv passthrough with
shell=False, server-injected credential env, bounded I/O) + shim
script constant + install-shim mode.
- docker/lark-cli-broker: init(install-shim) + serve image.
- provisioner: LARK_CLI_BROKER_IMAGE + provision_lark_cli_broker →
shim init container + lark-cli-broker sidecar (config/data mounted
sidecar-only); credentials dropped from the sandbox container;
/api/capabilities reports lark_cli_broker_image. Broker supersedes
the Pattern A init-container binary when both are configured.
- gateway: lark_cli_env_overlay(broker=True) omits config/data env;
sandbox_lark_broker_active() TTL-cached mode resolver; broker added
to sandbox_runtime_mode / readiness and the settings UI.
Opt-in and off by default (empty LARK_CLI_BROKER_IMAGE ⇒ no change).
Closes#4338
* fix(lark): address Pattern B broker review findings (#4501)
Follow-up to the sidecar credential broker addressing the PR #4501 review:
- shim: split the on-PATH lark-cli into a /bin/sh launcher + Python shim body
so broker mode fails loudly (exit 127, actionable message) instead of ENOEXEC
when the sandbox image ships no python3; interpreter pinnable via
DEERFLOW_LARK_BROKER_PYTHON. Launcher bakes in the shim's absolute path since
$0 is the bare command name when run off PATH.
- broker: drop the dead cwd payload field (broker can't see the sandbox FS) and
document the command-surface-only / no-file-IO limitation.
- broker: return a structured 500 JSON on unexpected exec errors so the shim
gets a meaningful message, not an opaque transport failure; set a handler
socket timeout to bound slow/stuck connections.
- broker: add an opt-in DEERFLOW_LARK_BROKER_DENY_SUBCOMMANDS denylist that
refuses secret-dumping subcommands before spawning the binary, forwarded from
the provisioner sidecar.
- gateway: tighten the per-bash-call broker probe timeout (1.5s) and cache
negatives longer (300s) so non-broker remote-provisioner users don't pay a
latency hit; guard the mode cache with a lock; drop the dead
_probe_provisioner_lark_cli_init_image wrapper.
- docs: remove the broken design-doc link from the broker README.
Adds tests for launcher python resolution, cwd omission, denylist enforcement,
500-on-error, hot-path probe timeout + negative caching, and provisioner
denylist-env wiring.
Editing the only turn of a thread reran the original prompt: the model
answered the question the edit was replacing while the UI showed the
edited text, and the edit vanished on reload.
The replay-base lookup decided whether a checkpoint predates the target
user message by message id alone. DynamicContextMiddleware re-keys the
first user turn to `{id}__user` mid-run, so every checkpoint written
before it holds the same prompt under an id the lookup cannot match. The
scan walked past those and anchored inside the run that produced the
turn — a checkpoint that still contains the original prompt and owns the
injection node's pending writes, which the replay then re-added after the
edited message.
Require the replay base to be a settled checkpoint (no pending tasks) in
both the lineage walk and the chronological fallback. That rule is
middleware agnostic: the first turn now anchors on the thread's empty
initial checkpoint and later turns on the previous run's tail, which also
drops the existing reliance on LangGraph discarding a stale `__start__`
write.
Edit replay additionally passes `head_checkpoint` so it resolves its base
lineage-first like regenerate does, and a replayed user message is
restored to its pre-swap id: replaying `{id}__user` into a state that has
no reminder yet makes the middleware treat the turn as already injected
and silently drops its date and memory block.
Frontend: a prepared replay masks the turn it supersedes, so the
optimistic-message baseline is taken from the post-mask human count. The
pre-mask count can never be exceeded when the replay puts exactly one
human message back, and on the first turn the runtime re-keys the
replacement message so identity comparison cannot stand in for the count.
Fixes#4531
