🧵 feat: Background Tool Calls for Agents & Model Specs - #14197
Merged
Merged
Conversation
Opt-in, poll-based background tool execution. The model marks an eligible tool call with `run_in_background: true`; the host executor registers a task, returns a handle immediately (so the graph turn resolves), runs the tool as a detached promise, and the model retrieves the result via a new `check_background_task` poll tool. Host-side only — no `@librechat/agents` change. - Opt-in mirrors `deferred_tools`: admin capability `run_in_background` (off by default) + per-tool `tool_options.run_in_background`. - Model specs / ephemeral agents: `TModelSpec.runInBackground` / `TEphemeralAgent.run_in_background` synthesize per-tool options; both paths converge at `initializeAgent`. - In-process task registry: scoped per user+conversation, idempotent by toolCallId (safe across resume/replay), capped, TTL-swept. - Excludes direct-path / host-special / code-session tools. Subagents and push notifications are deferred follow-ups.
Collaborator
Author
|
@codex review |
Contributor
There was a problem hiding this comment.
Pull request overview
This PR introduces an opt-in “background tool call” execution mode for agents/model specs, allowing eligible tool calls to return an immediate handle and be polled later via a new check_background_task tool, without requiring SDK changes.
Changes:
- Adds
run_in_backgroundopt-in flags to per-tool options, ephemeral agents, and model specs, plus validation support. - Implements host-side background dispatch, in-process task registry, schema injection, and a polling tool (
check_background_task). - Adds unit/integration tests covering eligibility, schema injection, registry behavior, and real
ON_TOOL_EXECUTEbehavior.
Reviewed changes
Copilot reviewed 14 out of 14 changed files in this pull request and generated 3 comments.
Show a summary per file
| File | Description |
|---|---|
| packages/data-provider/src/types/assistants.ts | Adds run_in_background per-tool option. |
| packages/data-provider/src/types.ts | Adds run_in_background to TEphemeralAgent. |
| packages/data-provider/src/models.ts | Adds runInBackground to TModelSpec and schema. |
| packages/data-provider/src/config.ts | Adds AgentCapabilities.run_in_background and Constants.CHECK_BACKGROUND_TASK. |
| packages/api/src/agents/validation.ts | Accepts run_in_background in tool options Zod schemas. |
| packages/api/src/agents/load.ts | Synthesizes ephemeral tool_options from spec/ephemeral background flag. |
| packages/api/src/agents/initialize.ts | Applies background-tool schema injection + poll-tool registration after tool registration. |
| packages/api/src/agents/handlers.ts | Adds ON_TOOL_EXECUTE interception for background dispatch + polling. |
| packages/api/src/agents/background.ts | New implementation: eligibility rules, schema injection, registry, poll tool, synthesis helpers. |
| packages/api/src/agents/background.spec.ts | New unit tests for background tooling and registry. |
| packages/api/src/agents/handlers.background.spec.ts | New tests driving the real execute handler behavior. |
| packages/api/src/agents/added.ts | Synthesizes background tool_options for “added” agents (primary + non-primary). |
| packages/api/src/agents/tests/load.spec.ts | Tests for background tool_options synthesis via ephemeral flag / model spec. |
| api/server/services/Endpoints/agents/initialize.js | Wires the new capability flag into initializeAgent params. |
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
Comment on lines
+277
to
+280
| function toStoredContent(content: unknown): string { | ||
| const asString = typeof content === 'string' ? content : JSON.stringify(content ?? ''); | ||
| return asString.length > MAX_RESULT_CHARS ? asString.slice(0, MAX_RESULT_CHARS) : asString; | ||
| } |
Comment on lines
+192
to
+215
| let changed = false; | ||
| const nextDefs = defs.map((def) => { | ||
| const optedIn = toolOptions[def.name]?.run_in_background === true; | ||
| if (!optedIn || !isBackgroundEligibleToolName(def.name)) { | ||
| return def; | ||
| } | ||
| const injected = injectRunInBackgroundParam(def); | ||
| if (injected === def) { | ||
| return def; | ||
| } | ||
| changed = true; | ||
| const registryEntry = toolRegistry?.get(def.name); | ||
| if (registryEntry) { | ||
| toolRegistry?.set(def.name, { ...registryEntry, parameters: injected.parameters }); | ||
| } | ||
| return injected; | ||
| }); | ||
|
|
||
| if (!changed) { | ||
| return { toolDefinitions: defs, enabled: false }; | ||
| } | ||
|
|
||
| const withPoll = registerBackgroundTaskTool({ toolRegistry, toolDefinitions: nextDefs }); | ||
| return { toolDefinitions: withPoll.toolDefinitions, enabled: true }; |
Comment on lines
+3237
to
+3247
| const backgroundReq = mergedConfigurable?.req as ServerRequest | undefined; | ||
| const backgroundUserId = backgroundReq?.user?.id ?? ''; | ||
| const backgroundConversationId = | ||
| ((metadata as Record<string, unknown>)?.thread_id as string | undefined) ?? | ||
| (mergedConfigurable?.thread_id as string | undefined) ?? | ||
| (backgroundReq?.body as { conversationId?: string } | undefined)?.conversationId ?? | ||
| ''; | ||
| const backgroundActive = | ||
| (mergedConfigurable?.toolRegistry as LCToolRegistry | undefined)?.has( | ||
| CHECK_BACKGROUND_TASK_NAME, | ||
| ) === true; |
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 8b1f4111b2
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
Comment on lines
+3283
to
+3287
| const result = (await tool.invoke( | ||
| normalizeToolInvokeArgs(strippedArgs, tool), | ||
| { | ||
| toolCall: { id: tc.id, stepId: tc.stepId, turn: tc.turn }, | ||
| configurable: mergedConfigurable, |
There was a problem hiding this comment.
The detached background invoke reuses the current request's tool instance and mergedConfigurable; those are built with the job abort signal and request-scoped MCP connections, and the resumable controller aborts/cleans them when the main turn completes. For a genuinely long-running background call that is still running after the response finishes, tools that honor the signal or use the request MCP context can be canceled/closed before the next-turn poll can retrieve a result.
Useful? React with 👍 / 👎.
Collaborator
Author
There was a problem hiding this comment.
Intentional scope for this PR, not a silent gap. Raw background tool calls are deliberately same-run here — matching LangChain deepagents, which backgrounds subagents (their own durable run/thread), not raw tool calls, for exactly this request-lifecycle reason. Since the detached invoke shares the run's abort signal, a tool still running at turn end is aborted and our catch marks the task error (it never hangs as running), so a later poll reports cleanly. Durable, request-detached execution that outlives the run is a tracked follow-up, modeled as background subagents over the existing checkpointer + GenerationJobManager substrate (a run/thread is the right durable unit; a raw tool call is not). Leaving this thread open as the follow-up marker.
- Reliable per-agent execution gate: thread the injected `run_in_background` tool names from `initializeAgent` through `configurable.backgroundToolNames` (`toolRegistry` only reaches the executor for PTC/tool_search), fixing the silent no-op + unstripped-arg leak for ordinary event-driven tools. - Enforce the per-tool opt-in at execution (`backgroundToolSet.has(name)`) so a non-opted-in tool can't be backgrounded via an extra arg. - Gate the `check_background_task` interception on the run actually enabling background, so a user tool sharing that name still executes. - Forward `backgroundToolsAvailable` to added-convo (multi-convo) agents. - Exclude `web_search`/`file_search` from eligibility — their results are turned into user-visible attachments/citations only by the foreground toolEndCallback.
Collaborator
Author
|
Codex round 1 addressed in
Request-lifecycle detachment (tasks outliving the run) is intentionally deferred to a durable follow-up (background subagents over the checkpointer substrate) — see the remaining open thread. |
Collaborator
Author
|
@codex review |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 7a45465705
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
| const strippedArgs = stripRunInBackgroundArg(tc.args); | ||
| void (async () => { | ||
| try { | ||
| const result = (await tool.invoke( |
There was a problem hiding this comment.
The detached invocation still uses the same tool instance that was loaded for this request, and those tools were constructed with the request/job abort signal. In the agent controller, normal final cleanup calls GenerationJobManager.completeJob(), which aborts that runtime controller, so a long-running background MCP/action call can be cancelled as soon as the turn returns the synthetic handle instead of continuing until the user polls it. Start background work with a signal/lifecycle that survives normal job completion, or delay cleanup until the task settles.
Useful? React with 👍 / 👎.
Collaborator
Author
There was a problem hiding this comment.
Same intentional scope as the earlier request-lifecycle thread. This PR is same-run by design (deepagents backgrounds subagents, not raw tool calls, for exactly this reason): the value is parallelism across a multi-step turn, and if a tool is still running when the turn's cleanup aborts, our catch marks the task error rather than hanging. Surviving completeJob()/the request abort to outlive the turn is the durable follow-up — modeled as background subagents over the existing checkpointer + GenerationJobManager substrate (a run/thread is the right durable unit; detaching a raw tool also means owning its MCP-connection lifetime). Keeping this open as the follow-up marker.
- Idempotency scoped to run+turn: provider tool-call ids repeat across turns (e.g. `call_0`), so key the dedupe map by `runId::toolCallId` and sweep orphaned mappings — a later turn no longer collides with a retained task. - Artifacts preserved: a backgrounded tool's artifact is processed through the same `toolEndCallback` as the foreground path (images/files/citations no longer silently dropped), best-effort/guarded. - Forward the `run_in_background` capability to connected-agent discovery and subagent `processAgent` init, so a child agent's own event-driven tools work the same as when it runs as primary. - Strip the injected flag on foreground calls of background-capable tools (the model may emit it as `false`) so strict MCP/action schemas don't reject. - `check_background_task` list path returns metadata only (result_available / result_chars), never full results — prevents context overflow; the full result is returned only when a specific id is requested.
Collaborator
Author
|
Codex round 2 addressed in
Request abort/lifecycle detachment (tasks outliving the turn) remains the intentional same-run boundary; durable execution is the tracked subagent follow-up — see the open thread. |
Collaborator
Author
|
@codex review |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: 445bb71f8c
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
Comment on lines
+3293
to
+3297
| { | ||
| toolCall: { id: tc.id, stepId: tc.stepId, turn: tc.turn }, | ||
| configurable: mergedConfigurable, | ||
| metadata, | ||
| } as Record<string, unknown>, |
There was a problem hiding this comment.
For backgrounded MCP tools that outlive the foreground turn, passing the same mergedConfigurable into the detached invocation also passes the request-scoped MCP connection store. The agents request controller cleans that store with cleanupMCPRequestContextForReq(req) as soon as the turn completes, so a genuinely long-running background MCP call can have its connection/auth torn down while it is still executing, causing the task to fail even though the feature is meant for calls that continue after the response proceeds.
Useful? React with 👍 / 👎.
Collaborator
Author
There was a problem hiding this comment.
Same same-run boundary as the other lifecycle threads, now specific to the MCP request-scoped connection store. Backgrounded MCP calls are meant to run within the turn (parallel to the model's other steps) and be polled before cleanupMCPRequestContextForReq runs at turn end; a call that outlives the turn is the durable follow-up, which is exactly where owning the MCP connection/auth lifetime off-request belongs (modeled as background subagents over the checkpointer substrate). Keeping this open as the follow-up marker.
Comment on lines
+3305
to
+3308
| if (toolEndCallback && result.artifact != null) { | ||
| try { | ||
| await toolEndCallback( | ||
| { |
There was a problem hiding this comment.
When a backgrounded tool finishes after the main turn has already saved the response, this callback is too late to persist its attachment: BaseClient awaits the current artifactPromises only during response save (api/app/clients/BaseClient.js:808-810), so promises appended here later are never attached to the message, and the standard SSE stream may already be closed. Because artifact-producing tools like image generation are still eligible, polling then only reports hasArtifact while the actual generated file/resource is lost to the conversation.
Useful? React with 👍 / 👎.
Collaborator
Author
There was a problem hiding this comment.
Partly addressed in 59559c035, partly scoped. The named case — image-generation built-ins — is now excluded from eligibility (dalle/flux/gemini_image_gen/image_gen_oai/image_edit_oai), so their files aren't silently dropped. For other tools that finish within the turn, the artifact is still processed best-effort through the same toolEndCallback while artifactPromises are being collected. Attaching an artifact to an already-saved turn is genuinely impossible (BaseClient awaits promises only at save), so that case is the durable follow-up, not something this same-run PR can fix — the poll result flags hasArtifact so the model isn't misled. Keeping this open as a follow-up marker.
- Exclude background-capable tools from eager execution (run.ts): a speculative eager dispatch of a `run_in_background` call could launch the detached task with partial/stale args, and that side effect can't be canceled. - Reserve the `check_background_task` name: overwrite a colliding user/MCP tool with the host poll schema (with a warning) so the advertised schema matches the executor's interception instead of hijacking a mismatched tool. - Don't inject background schemas into pure subagents (spawn-tool child graphs) whose tools don't reach the host interceptor; keep it for primary/added/ connected agents. Subagent background is the durable follow-up. - Thread `backgroundToolsAvailable` + `backgroundToolNames` through the OpenAI-compatible and Responses agent routes (was chat-only), so the same agent/model spec behaves consistently across surfaces. - Exclude image-generation built-ins (dalle/flux/gemini_image_gen/image_gen_oai/ image_edit_oai) — artifact-first tools whose files can't reliably attach to an already-saved turn when backgrounded.
Collaborator
Author
|
Codex round 3 addressed in
Two facets of the same-run boundary remain open threads (intentional, → durable subagent follow-up): request/MCP-connection lifetime for tasks that outlive the turn, and attaching artifacts to an already-saved turn. Image-generation built-ins are excluded as a concrete safeguard. |
|
Codex Review: Didn't find any major issues. Hooray! Reviewed commit: ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
If Codex has suggestions, it will comment; otherwise it will react with 👍. Codex can also answer questions or update the PR. Try commenting "@codex address that feedback". |
- Error immediately (matching foreground) when a background-requested tool failed to load, instead of returning a success handle for a dead task - Exclude ephemeral request-scoped MCP tools at injection time so the model never sees a run_in_background param the executor would silently downgrade; flip the execute-time tag to fail closed on a missing server config - Source image-tool background exclusions from the shared imageGenTools set (adds missing stable-diffusion, an artifact-first live tool) instead of a hand-copied list - Add check_background_task to the eager-execution exclusion list: artifact collection is a one-shot claim that must not fire from a speculative snapshot the SDK may discard - Strip an imitated run_in_background arg on tools the executing agent never opted in (multi-agent history bleed), unless the tool's own schema declares the parameter - Truncate oversized stored results with an explicit marker via the shared truncateMiddle (moved to utils/text) instead of a silent slice - Document the at-most-once artifact delivery semantics honestly (the callback's downstream persistence is fire-and-forget, as in foreground)
- Use the SDK's JsonSchemaType instead of a local duplicate; drop all as-unknown casts and type the poll-tool serializer explicitly - Drop derivable BackgroundTask state (progress, hasArtifact) and the dead `enabled` param/return on applyBackgroundToolCalls (guarded at the call site), which also skips the defs pass when nothing opted in - Fold the enable expression into synthesizeBackgroundToolOptions so the three load/added call sites can't drift - Throttle the registry's all-buckets sweep and always sweep the accessed bucket, so a hot poll loop is no longer O(total tasks server-wide); bound retained artifact memory with a size cap - Single-pass stripBackgroundFromToolDefinitions; pass metadata through to the poll-turn callback instead of a no-op reconstruction - Collapse the client's copy-pasted boolean option families into a keyed factory (also removes the shared-object mutation in the bulk toggles) and the six toggle-button copies into one OptionToggle component
|
Codex Review: Didn't find any major issues. Breezy! Reviewed commit: ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
If Codex has suggestions, it will comment; otherwise it will react with 👍. Codex can also answer questions or update the PR. Try commenting "@codex address that feedback". |
Proves the full contract through the real pipeline (mock harness): an agent opts an MCP tool in via tool_options.run_in_background, the model dispatches it detached and receives the synthetic handle while the tool is still running (status=running in the rendered ack — the non-blocking guarantee without timing assertions), the tool completes after its turn finalized, and a later user turn recovers the task id from replayed history, polls check_background_task, and renders the collected result. - fake-mcp-server: slow_echo fixture tool (delayed echo) - fake-model: E2E_BACKGROUND_DISPATCH / E2E_BACKGROUND_COLLECT markers - e2e yaml: agents capabilities = defaults + run_in_background
Collaborator
Author
|
@codex review |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: cd6d15a2f5
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
- Thread backgroundToolsAvailable through the OpenAI-compatible service (derived from app capabilities like codeEnvAvailable/statefulSessions), so agents with tool_options.run_in_background keep the feature on that route; fold the three capability derivations into one helper - Index ephemeral MCP servers by normalizeServerName when excluding tools from background injection: tool names embed the normalized server name while mcpConfig keys the original, so exotic server names previously escaped the injection-time exclusion
|
Codex Review: Didn't find any major issues. Keep it up! Reviewed commit: ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
If Codex has suggestions, it will comment; otherwise it will react with 👍. Codex can also answer questions or update the PR. Try commenting "@codex address that feedback". |
Collaborator
Author
|
@codex review |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: d8d52c5981
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
…oping The in-repo routes merge req into the tool-execute configurable, but external hosts of the exported OpenAI-compatible service inject their own loadTools and may not — tasks would then register under an empty user id, collapsing registry isolation to conversationId alone. Resolve the scoping id from req.user.id, then configurable.user_id / user, and cover the isolation with a foreign-user not_found test.
|
Codex Review: Didn't find any major issues. Already looking forward to the next diff. Reviewed commit: ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
If Codex has suggestions, it will comment; otherwise it will react with 👍. Codex can also answer questions or update the PR. Try commenting "@codex address that feedback". |
|
Codex Review: Didn't find any major issues. 👍 Reviewed commit: ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
If Codex has suggestions, it will comment; otherwise it will react with 👍. Codex can also answer questions or update the PR. Try commenting "@codex address that feedback". |
|
Codex Review: Didn't find any major issues. Hooray! Reviewed commit: ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
If Codex has suggestions, it will comment; otherwise it will react with 👍. Codex can also answer questions or update the PR. Try commenting "@codex address that feedback". |
This was referenced Jul 22, 2026
fuuuzzy
pushed a commit
to fuuuzzy/LibreChat
that referenced
this pull request
Aug 7, 2026
…14197) * 🧵 feat: Background Tool Calls for Agents & Model Specs Opt-in, poll-based background tool execution. The model marks an eligible tool call with `run_in_background: true`; the host executor registers a task, returns a handle immediately (so the graph turn resolves), runs the tool as a detached promise, and the model retrieves the result via a new `check_background_task` poll tool. Host-side only — no `@librechat/agents` change. - Opt-in mirrors `deferred_tools`: admin capability `run_in_background` (off by default) + per-tool `tool_options.run_in_background`. - Model specs / ephemeral agents: `TModelSpec.runInBackground` / `TEphemeralAgent.run_in_background` synthesize per-tool options; both paths converge at `initializeAgent`. - In-process task registry: scoped per user+conversation, idempotent by toolCallId (safe across resume/replay), capped, TTL-swept. - Excludes direct-path / host-special / code-session tools. Subagents and push notifications are deferred follow-ups. * 🩹 fix: Harden background tool calls (Codex review) - Reliable per-agent execution gate: thread the injected `run_in_background` tool names from `initializeAgent` through `configurable.backgroundToolNames` (`toolRegistry` only reaches the executor for PTC/tool_search), fixing the silent no-op + unstripped-arg leak for ordinary event-driven tools. - Enforce the per-tool opt-in at execution (`backgroundToolSet.has(name)`) so a non-opted-in tool can't be backgrounded via an extra arg. - Gate the `check_background_task` interception on the run actually enabling background, so a user tool sharing that name still executes. - Forward `backgroundToolsAvailable` to added-convo (multi-convo) agents. - Exclude `web_search`/`file_search` from eligibility — their results are turned into user-visible attachments/citations only by the foreground toolEndCallback. * 🩹 fix: Address Codex round 2 on background tool calls - Idempotency scoped to run+turn: provider tool-call ids repeat across turns (e.g. `call_0`), so key the dedupe map by `runId::toolCallId` and sweep orphaned mappings — a later turn no longer collides with a retained task. - Artifacts preserved: a backgrounded tool's artifact is processed through the same `toolEndCallback` as the foreground path (images/files/citations no longer silently dropped), best-effort/guarded. - Forward the `run_in_background` capability to connected-agent discovery and subagent `processAgent` init, so a child agent's own event-driven tools work the same as when it runs as primary. - Strip the injected flag on foreground calls of background-capable tools (the model may emit it as `false`) so strict MCP/action schemas don't reject. - `check_background_task` list path returns metadata only (result_available / result_chars), never full results — prevents context overflow; the full result is returned only when a specific id is requested. * 🩹 fix: Address Codex round 3 on background tool calls - Exclude background-capable tools from eager execution (run.ts): a speculative eager dispatch of a `run_in_background` call could launch the detached task with partial/stale args, and that side effect can't be canceled. - Reserve the `check_background_task` name: overwrite a colliding user/MCP tool with the host poll schema (with a warning) so the advertised schema matches the executor's interception instead of hijacking a mismatched tool. - Don't inject background schemas into pure subagents (spawn-tool child graphs) whose tools don't reach the host interceptor; keep it for primary/added/ connected agents. Subagent background is the durable follow-up. - Thread `backgroundToolsAvailable` + `backgroundToolNames` through the OpenAI-compatible and Responses agent routes (was chat-only), so the same agent/model spec behaves consistently across surfaces. - Exclude image-generation built-ins (dalle/flux/gemini_image_gen/image_gen_oai/ image_edit_oai) — artifact-first tools whose files can't reliably attach to an already-saved turn when backgrounded. * 🩹 fix: Address Codex round 4 on background tool calls - Sanitize self-spawn subagent inputs: strip `run_in_background` + the `check_background_task` def from the parent AgentInputs reused for self-spawn, so the isolated child (direct/child-graph path) doesn't advertise a background schema it can't honor. The SDK resolver keeps a provided `agentInputs` even with `self: true`. - Exclude `check_background_task` from PTC (`run_tools_with_code`) tool definitions — it's host-only and not callable from generated code. - Parse stringified JSON args before deciding background dispatch and before stripping the flag, so string-delivered `run_in_background` is honored and never leaks to strict object-schema tools. - Skip injection for tools that already declare their own `run_in_background` param (would otherwise hijack/strip it), and for non-object (string-input) schemas (would otherwise rewrite the input contract). * 🩹 fix: Address Codex round 5 on background tool calls - check_background_task now parses stringified JSON args, so providers that deliver args as a string can retrieve a specific task by id (not just list). - Include agentId in the background dedupe key (`agentId::runId::toolCallId`): two agents in the same run emitting the same provider id (e.g. `call_0`) now launch independent tasks instead of colliding. - Self-spawn sanitization also strips the background entries from the reused toolRegistry (not just toolDefinitions), so a child using tool_search/deferred loading can't rediscover the host-only run_in_background / check_background_task. * 🩹 fix: Strip run_in_background from PTC target tool schemas (Codex round 6) The PTC path already filtered out the host-only check_background_task poll tool but still exposed target tool schemas with the injected `run_in_background` param (the shared toolRegistry entries were mutated by applyBackgroundToolCalls). PTC codegen doesn't go through the host background interceptor, so it could pass the flag to an MCP/action tool (strict-schema rejection or silent foreground with no poll). Sanitize the PTC toolDefs like the self-spawn path does. * 🩹 fix: Sanitize background from explicit subagent inputs (Codex round 7) A child agent reachable as a top-level/handoff agent is initialized WITH the background capability, then reused as an explicit subagent via buildSubagentConfigs. Round 4 only sanitized the self-spawn case; this now applies the same stripBackgroundFromToolDefinitions/Registry to explicit child agentInputs when `child.backgroundToolNames` is non-empty, so an isolated child graph doesn't advertise a run_in_background / check_background_task contract it can't honor. * 🩹 fix: Reap stuck/expired background tasks (Codex round 8) - get() now sweeps before returning, so repeatedly polling a known background_task_id can't keep an expired completed task (and its retained result, up to 100k chars) alive past the one-hour completed TTL. - sweep() now reaps `running` tasks older than a 30-min running TTL, marking them errored. Previously a detached call that never settled (hung network / lost MCP connection) held a running slot forever, exhausting the per-conversation cap and rejecting every later dispatch. * 🩹 fix: Evict oldest settled tasks instead of blocking at the cap (Codex round 9) Only the running-task cap gates dispatch now. The total-tasks cap (MAX_TASKS_PER_BUCKET) bounds memory but no longer rejects new background calls: when full, it evicts the oldest settled (completed/error) tasks to make room. Previously 200 quick background calls in one conversation would block all new dispatches for up to the completed-task TTL, since polling doesn't remove settled tasks. Running is already capped, so room always frees. * 📝 docs: Frame background tool calls as within-turn (Codex P1 contract) Codex escalated the request-lifecycle findings to P1 on the grounds that the advertised "poll later" contract can't be honored for genuinely long-running calls (request-scoped MCP connections + the run abort signal are torn down at turn end). Align the model-facing contract with what the same-run implementation actually delivers: the run_in_background param, check_background_task, and the dispatch handle now instruct the model to collect the result WITHIN THE SAME TURN (backgrounded work isn't guaranteed to survive past the turn). This is within-turn parallelism; cross-turn survival of long-running calls remains the deliberate durable subagent follow-up. Copy/comment-only; no behavior change. * ♻️ refactor: Cross-turn background tool calls, leak-free Extend background tool calls from within-turn to cross-turn on a single process, since the mechanism already supports it: the run's abort signal never reaches the detached invoke (the graph forwards only configurable/ metadata to the tool-execute handler), so the floating promise keeps running past turn completion and its result stays in the in-process registry for a later turn to poll (get/list key only on user::conversation + id, never the dispatch run/turn). Guarantee no connection leak: ephemeral request-scoped MCP tools (runtime {{LIBRECHAT_BODY_*}} placeholders) capture their request-scoped store at creation and fall back to it, so config manipulation can't redirect them; their connection is torn down at request end. Tag such tools in createToolInstance and run them in the foreground instead of backgrounding them. Pooled/app-level MCP and structured tools are unaffected and survive cross-turn via their managed pools. Reword the model-facing contract (run_in_background, check_background_task, handle message, fileoverview) from within-turn to cross-turn on this server (not across restart/replica, which stays the durable follow-up). Tests: cross-turn poll retrieval; ephemeral MCP tool runs foreground. * 🐛 fix: Guard ephemeral MCP tag against a null server config createToolInstance can be reached with a null/stale capturedServerConfig (cached availableTools + getServerConfig returns null, as several MCP unit tests construct tools). The new unconditional requiresEphemeralUserConnection call then dereferenced config.source and threw during tool construction (CI: Tests api shard 2/3). Guard with the same serverConfig ? ... : false pattern the other callers use; a missing config is not request-scoped. * 🎨 fix: Deliver backgrounded tool artifacts on the poll turn A slow backgrounded MCP/action tool resolves after its dispatch turn is finalized: createToolEndCallback only appends to that turn's artifactPromises (already awaited) and writes to a closed stream, so the artifact (file/citation/ UI resource) was silently dropped — check_background_task recorded only the hasArtifact boolean. The cross-turn contract made this the common case. Hold the artifact on the task and deliver it through the LIVE poll turn's toolEndCallback the first time check_background_task collects that id (once, then cleared to free memory), attributed to the original tool. Same-turn and cross-turn now share this path since the model must poll to collect any result. Tests: registry claim-once; artifact delivered on poll not dispatch, idempotent. * ✨ feat: Agent-builder toggle for background tool calls + cap tool descriptions Add a per-MCP-tool "run in background" toggle in the agent builder, mirroring the programmatic/deferred pattern: gated on the admin `run_in_background` capability via useAgentCapabilities, read/written on tool_options[id] .run_in_background through useMCPToolOptions (per-tool + bulk mark-all), and rendered as a Zap toggle in MCPToolItem and McpSection with new locale keys. Also cap the section tool/server descriptions (McpSection, ToolSection, SkillSection) with max-h-40 overflow-y-auto so a long description scrolls instead of overflowing the dialog, matching MCPToolItem's existing cap. Tests: MCPToolItem renders/toggles the background button only when enabled. * 🧪 fix: Mock new background hook functions in McpSection spec * 🎨 fix: Restore background artifact when poll-turn delivery fails * 🛡️ fix: Harden background tool call edges from review findings - Error immediately (matching foreground) when a background-requested tool failed to load, instead of returning a success handle for a dead task - Exclude ephemeral request-scoped MCP tools at injection time so the model never sees a run_in_background param the executor would silently downgrade; flip the execute-time tag to fail closed on a missing server config - Source image-tool background exclusions from the shared imageGenTools set (adds missing stable-diffusion, an artifact-first live tool) instead of a hand-copied list - Add check_background_task to the eager-execution exclusion list: artifact collection is a one-shot claim that must not fire from a speculative snapshot the SDK may discard - Strip an imitated run_in_background arg on tools the executing agent never opted in (multi-agent history bleed), unless the tool's own schema declares the parameter - Truncate oversized stored results with an explicit marker via the shared truncateMiddle (moved to utils/text) instead of a silent slice - Document the at-most-once artifact delivery semantics honestly (the callback's downstream persistence is fire-and-forget, as in foreground) * ♻️ refactor: Deduplicate background tool-call plumbing and tighten types - Use the SDK's JsonSchemaType instead of a local duplicate; drop all as-unknown casts and type the poll-tool serializer explicitly - Drop derivable BackgroundTask state (progress, hasArtifact) and the dead `enabled` param/return on applyBackgroundToolCalls (guarded at the call site), which also skips the defs pass when nothing opted in - Fold the enable expression into synthesizeBackgroundToolOptions so the three load/added call sites can't drift - Throttle the registry's all-buckets sweep and always sweep the accessed bucket, so a hot poll loop is no longer O(total tasks server-wide); bound retained artifact memory with a size cap - Single-pass stripBackgroundFromToolDefinitions; pass metadata through to the poll-turn callback instead of a no-op reconstruction - Collapse the client's copy-pasted boolean option families into a keyed factory (also removes the shared-object mutation in the bulk toggles) and the six toggle-button copies into one OptionToggle component * 🧪 test: e2e coverage for cross-turn background tool calls Proves the full contract through the real pipeline (mock harness): an agent opts an MCP tool in via tool_options.run_in_background, the model dispatches it detached and receives the synthetic handle while the tool is still running (status=running in the rendered ack — the non-blocking guarantee without timing assertions), the tool completes after its turn finalized, and a later user turn recovers the task id from replayed history, polls check_background_task, and renders the collected result. - fake-mcp-server: slow_echo fixture tool (delayed echo) - fake-model: E2E_BACKGROUND_DISPATCH / E2E_BACKGROUND_COLLECT markers - e2e yaml: agents capabilities = defaults + run_in_background * 🔧 fix: Close two background capability gaps from review - Thread backgroundToolsAvailable through the OpenAI-compatible service (derived from app capabilities like codeEnvAvailable/statefulSessions), so agents with tool_options.run_in_background keep the feature on that route; fold the three capability derivations into one helper - Index ephemeral MCP servers by normalizeServerName when excluding tools from background injection: tool names embed the normalized server name while mcpConfig keys the original, so exotic server names previously escaped the injection-time exclusion * 🛂 fix: Fall back to configurable user identity for background task scoping The in-repo routes merge req into the tool-execute configurable, but external hosts of the exported OpenAI-compatible service inject their own loadTools and may not — tasks would then register under an empty user id, collapsing registry isolation to conversationId alone. Resolve the scoping id from req.user.id, then configurable.user_id / user, and cover the isolation with a foreign-user not_found test. * 🧹 chore: Apply repo import sorter to PR-touched files
7 tasks
LogicalAbsurd
pushed a commit
to LogicalAbsurd/LibreChat
that referenced
this pull request
Aug 27, 2026
…14197) * 🧵 feat: Background Tool Calls for Agents & Model Specs Opt-in, poll-based background tool execution. The model marks an eligible tool call with `run_in_background: true`; the host executor registers a task, returns a handle immediately (so the graph turn resolves), runs the tool as a detached promise, and the model retrieves the result via a new `check_background_task` poll tool. Host-side only — no `@librechat/agents` change. - Opt-in mirrors `deferred_tools`: admin capability `run_in_background` (off by default) + per-tool `tool_options.run_in_background`. - Model specs / ephemeral agents: `TModelSpec.runInBackground` / `TEphemeralAgent.run_in_background` synthesize per-tool options; both paths converge at `initializeAgent`. - In-process task registry: scoped per user+conversation, idempotent by toolCallId (safe across resume/replay), capped, TTL-swept. - Excludes direct-path / host-special / code-session tools. Subagents and push notifications are deferred follow-ups. * 🩹 fix: Harden background tool calls (Codex review) - Reliable per-agent execution gate: thread the injected `run_in_background` tool names from `initializeAgent` through `configurable.backgroundToolNames` (`toolRegistry` only reaches the executor for PTC/tool_search), fixing the silent no-op + unstripped-arg leak for ordinary event-driven tools. - Enforce the per-tool opt-in at execution (`backgroundToolSet.has(name)`) so a non-opted-in tool can't be backgrounded via an extra arg. - Gate the `check_background_task` interception on the run actually enabling background, so a user tool sharing that name still executes. - Forward `backgroundToolsAvailable` to added-convo (multi-convo) agents. - Exclude `web_search`/`file_search` from eligibility — their results are turned into user-visible attachments/citations only by the foreground toolEndCallback. * 🩹 fix: Address Codex round 2 on background tool calls - Idempotency scoped to run+turn: provider tool-call ids repeat across turns (e.g. `call_0`), so key the dedupe map by `runId::toolCallId` and sweep orphaned mappings — a later turn no longer collides with a retained task. - Artifacts preserved: a backgrounded tool's artifact is processed through the same `toolEndCallback` as the foreground path (images/files/citations no longer silently dropped), best-effort/guarded. - Forward the `run_in_background` capability to connected-agent discovery and subagent `processAgent` init, so a child agent's own event-driven tools work the same as when it runs as primary. - Strip the injected flag on foreground calls of background-capable tools (the model may emit it as `false`) so strict MCP/action schemas don't reject. - `check_background_task` list path returns metadata only (result_available / result_chars), never full results — prevents context overflow; the full result is returned only when a specific id is requested. * 🩹 fix: Address Codex round 3 on background tool calls - Exclude background-capable tools from eager execution (run.ts): a speculative eager dispatch of a `run_in_background` call could launch the detached task with partial/stale args, and that side effect can't be canceled. - Reserve the `check_background_task` name: overwrite a colliding user/MCP tool with the host poll schema (with a warning) so the advertised schema matches the executor's interception instead of hijacking a mismatched tool. - Don't inject background schemas into pure subagents (spawn-tool child graphs) whose tools don't reach the host interceptor; keep it for primary/added/ connected agents. Subagent background is the durable follow-up. - Thread `backgroundToolsAvailable` + `backgroundToolNames` through the OpenAI-compatible and Responses agent routes (was chat-only), so the same agent/model spec behaves consistently across surfaces. - Exclude image-generation built-ins (dalle/flux/gemini_image_gen/image_gen_oai/ image_edit_oai) — artifact-first tools whose files can't reliably attach to an already-saved turn when backgrounded. * 🩹 fix: Address Codex round 4 on background tool calls - Sanitize self-spawn subagent inputs: strip `run_in_background` + the `check_background_task` def from the parent AgentInputs reused for self-spawn, so the isolated child (direct/child-graph path) doesn't advertise a background schema it can't honor. The SDK resolver keeps a provided `agentInputs` even with `self: true`. - Exclude `check_background_task` from PTC (`run_tools_with_code`) tool definitions — it's host-only and not callable from generated code. - Parse stringified JSON args before deciding background dispatch and before stripping the flag, so string-delivered `run_in_background` is honored and never leaks to strict object-schema tools. - Skip injection for tools that already declare their own `run_in_background` param (would otherwise hijack/strip it), and for non-object (string-input) schemas (would otherwise rewrite the input contract). * 🩹 fix: Address Codex round 5 on background tool calls - check_background_task now parses stringified JSON args, so providers that deliver args as a string can retrieve a specific task by id (not just list). - Include agentId in the background dedupe key (`agentId::runId::toolCallId`): two agents in the same run emitting the same provider id (e.g. `call_0`) now launch independent tasks instead of colliding. - Self-spawn sanitization also strips the background entries from the reused toolRegistry (not just toolDefinitions), so a child using tool_search/deferred loading can't rediscover the host-only run_in_background / check_background_task. * 🩹 fix: Strip run_in_background from PTC target tool schemas (Codex round 6) The PTC path already filtered out the host-only check_background_task poll tool but still exposed target tool schemas with the injected `run_in_background` param (the shared toolRegistry entries were mutated by applyBackgroundToolCalls). PTC codegen doesn't go through the host background interceptor, so it could pass the flag to an MCP/action tool (strict-schema rejection or silent foreground with no poll). Sanitize the PTC toolDefs like the self-spawn path does. * 🩹 fix: Sanitize background from explicit subagent inputs (Codex round 7) A child agent reachable as a top-level/handoff agent is initialized WITH the background capability, then reused as an explicit subagent via buildSubagentConfigs. Round 4 only sanitized the self-spawn case; this now applies the same stripBackgroundFromToolDefinitions/Registry to explicit child agentInputs when `child.backgroundToolNames` is non-empty, so an isolated child graph doesn't advertise a run_in_background / check_background_task contract it can't honor. * 🩹 fix: Reap stuck/expired background tasks (Codex round 8) - get() now sweeps before returning, so repeatedly polling a known background_task_id can't keep an expired completed task (and its retained result, up to 100k chars) alive past the one-hour completed TTL. - sweep() now reaps `running` tasks older than a 30-min running TTL, marking them errored. Previously a detached call that never settled (hung network / lost MCP connection) held a running slot forever, exhausting the per-conversation cap and rejecting every later dispatch. * 🩹 fix: Evict oldest settled tasks instead of blocking at the cap (Codex round 9) Only the running-task cap gates dispatch now. The total-tasks cap (MAX_TASKS_PER_BUCKET) bounds memory but no longer rejects new background calls: when full, it evicts the oldest settled (completed/error) tasks to make room. Previously 200 quick background calls in one conversation would block all new dispatches for up to the completed-task TTL, since polling doesn't remove settled tasks. Running is already capped, so room always frees. * 📝 docs: Frame background tool calls as within-turn (Codex P1 contract) Codex escalated the request-lifecycle findings to P1 on the grounds that the advertised "poll later" contract can't be honored for genuinely long-running calls (request-scoped MCP connections + the run abort signal are torn down at turn end). Align the model-facing contract with what the same-run implementation actually delivers: the run_in_background param, check_background_task, and the dispatch handle now instruct the model to collect the result WITHIN THE SAME TURN (backgrounded work isn't guaranteed to survive past the turn). This is within-turn parallelism; cross-turn survival of long-running calls remains the deliberate durable subagent follow-up. Copy/comment-only; no behavior change. * ♻️ refactor: Cross-turn background tool calls, leak-free Extend background tool calls from within-turn to cross-turn on a single process, since the mechanism already supports it: the run's abort signal never reaches the detached invoke (the graph forwards only configurable/ metadata to the tool-execute handler), so the floating promise keeps running past turn completion and its result stays in the in-process registry for a later turn to poll (get/list key only on user::conversation + id, never the dispatch run/turn). Guarantee no connection leak: ephemeral request-scoped MCP tools (runtime {{LIBRECHAT_BODY_*}} placeholders) capture their request-scoped store at creation and fall back to it, so config manipulation can't redirect them; their connection is torn down at request end. Tag such tools in createToolInstance and run them in the foreground instead of backgrounding them. Pooled/app-level MCP and structured tools are unaffected and survive cross-turn via their managed pools. Reword the model-facing contract (run_in_background, check_background_task, handle message, fileoverview) from within-turn to cross-turn on this server (not across restart/replica, which stays the durable follow-up). Tests: cross-turn poll retrieval; ephemeral MCP tool runs foreground. * 🐛 fix: Guard ephemeral MCP tag against a null server config createToolInstance can be reached with a null/stale capturedServerConfig (cached availableTools + getServerConfig returns null, as several MCP unit tests construct tools). The new unconditional requiresEphemeralUserConnection call then dereferenced config.source and threw during tool construction (CI: Tests api shard 2/3). Guard with the same serverConfig ? ... : false pattern the other callers use; a missing config is not request-scoped. * 🎨 fix: Deliver backgrounded tool artifacts on the poll turn A slow backgrounded MCP/action tool resolves after its dispatch turn is finalized: createToolEndCallback only appends to that turn's artifactPromises (already awaited) and writes to a closed stream, so the artifact (file/citation/ UI resource) was silently dropped — check_background_task recorded only the hasArtifact boolean. The cross-turn contract made this the common case. Hold the artifact on the task and deliver it through the LIVE poll turn's toolEndCallback the first time check_background_task collects that id (once, then cleared to free memory), attributed to the original tool. Same-turn and cross-turn now share this path since the model must poll to collect any result. Tests: registry claim-once; artifact delivered on poll not dispatch, idempotent. * ✨ feat: Agent-builder toggle for background tool calls + cap tool descriptions Add a per-MCP-tool "run in background" toggle in the agent builder, mirroring the programmatic/deferred pattern: gated on the admin `run_in_background` capability via useAgentCapabilities, read/written on tool_options[id] .run_in_background through useMCPToolOptions (per-tool + bulk mark-all), and rendered as a Zap toggle in MCPToolItem and McpSection with new locale keys. Also cap the section tool/server descriptions (McpSection, ToolSection, SkillSection) with max-h-40 overflow-y-auto so a long description scrolls instead of overflowing the dialog, matching MCPToolItem's existing cap. Tests: MCPToolItem renders/toggles the background button only when enabled. * 🧪 fix: Mock new background hook functions in McpSection spec * 🎨 fix: Restore background artifact when poll-turn delivery fails * 🛡️ fix: Harden background tool call edges from review findings - Error immediately (matching foreground) when a background-requested tool failed to load, instead of returning a success handle for a dead task - Exclude ephemeral request-scoped MCP tools at injection time so the model never sees a run_in_background param the executor would silently downgrade; flip the execute-time tag to fail closed on a missing server config - Source image-tool background exclusions from the shared imageGenTools set (adds missing stable-diffusion, an artifact-first live tool) instead of a hand-copied list - Add check_background_task to the eager-execution exclusion list: artifact collection is a one-shot claim that must not fire from a speculative snapshot the SDK may discard - Strip an imitated run_in_background arg on tools the executing agent never opted in (multi-agent history bleed), unless the tool's own schema declares the parameter - Truncate oversized stored results with an explicit marker via the shared truncateMiddle (moved to utils/text) instead of a silent slice - Document the at-most-once artifact delivery semantics honestly (the callback's downstream persistence is fire-and-forget, as in foreground) * ♻️ refactor: Deduplicate background tool-call plumbing and tighten types - Use the SDK's JsonSchemaType instead of a local duplicate; drop all as-unknown casts and type the poll-tool serializer explicitly - Drop derivable BackgroundTask state (progress, hasArtifact) and the dead `enabled` param/return on applyBackgroundToolCalls (guarded at the call site), which also skips the defs pass when nothing opted in - Fold the enable expression into synthesizeBackgroundToolOptions so the three load/added call sites can't drift - Throttle the registry's all-buckets sweep and always sweep the accessed bucket, so a hot poll loop is no longer O(total tasks server-wide); bound retained artifact memory with a size cap - Single-pass stripBackgroundFromToolDefinitions; pass metadata through to the poll-turn callback instead of a no-op reconstruction - Collapse the client's copy-pasted boolean option families into a keyed factory (also removes the shared-object mutation in the bulk toggles) and the six toggle-button copies into one OptionToggle component * 🧪 test: e2e coverage for cross-turn background tool calls Proves the full contract through the real pipeline (mock harness): an agent opts an MCP tool in via tool_options.run_in_background, the model dispatches it detached and receives the synthetic handle while the tool is still running (status=running in the rendered ack — the non-blocking guarantee without timing assertions), the tool completes after its turn finalized, and a later user turn recovers the task id from replayed history, polls check_background_task, and renders the collected result. - fake-mcp-server: slow_echo fixture tool (delayed echo) - fake-model: E2E_BACKGROUND_DISPATCH / E2E_BACKGROUND_COLLECT markers - e2e yaml: agents capabilities = defaults + run_in_background * 🔧 fix: Close two background capability gaps from review - Thread backgroundToolsAvailable through the OpenAI-compatible service (derived from app capabilities like codeEnvAvailable/statefulSessions), so agents with tool_options.run_in_background keep the feature on that route; fold the three capability derivations into one helper - Index ephemeral MCP servers by normalizeServerName when excluding tools from background injection: tool names embed the normalized server name while mcpConfig keys the original, so exotic server names previously escaped the injection-time exclusion * 🛂 fix: Fall back to configurable user identity for background task scoping The in-repo routes merge req into the tool-execute configurable, but external hosts of the exported OpenAI-compatible service inject their own loadTools and may not — tasks would then register under an empty user id, collapsing registry isolation to conversationId alone. Resolve the scoping id from req.user.id, then configurable.user_id / user, and cover the isolation with a foreign-user not_found test. * 🧹 chore: Apply repo import sorter to PR-touched files
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Adds opt-in, poll-based background tool calls for agents and model specs. The model can mark an eligible tool call with an injected
run_in_background: trueparameter; the host executor registers a task, returns a synthetic handle immediately (so the graph superstep resolves and the turn continues), runs the real tool as a detached promise, and stores the result in an in-process registry. The model retrieves progress/result later via a newcheck_background_tasktool.A backgrounded call and the poll call are both synchronous from the graph's point of view, so no
@librechat/agentschange is required.Opt-in (mirrors
deferred_tools)AgentCapabilities.run_in_background(inendpoints.agents.capabilities). Off by default (experimental).tool_options[name].run_in_background.TModelSpec.runInBackgroundandTEphemeralAgent.run_in_backgroundsynthesize per-tooltool_optionsinload.ts/added.ts. Both paths converge atinitializeAgent, so the injection + execution logic is written once.When the capability is on and a tool is opted in, a
run_in_backgroundboolean is injected into that tool's schema and thecheck_background_taskpoll tool is registered for the run.How it works
initializeAgentrunsapplyBackgroundToolCallsafter all built-in tool registration: injects the param into eligible opted-in tools and registers the poll tool.createToolExecuteHandler(ON_TOOL_EXECUTE), a backgrounded call registers a task and returns{ background_task_id, status: "running", ... }immediately; the real tool runs detached and its result/error lands in the registry.check_background_taskis host-special-cased (likeread_file/skill) and reads the registry.The "background active" gate in the executor is
configurable.toolRegistry.has(check_background_task)— exactly the condition under which the model could have been shown the injected param.Registry
In-process, scoped per
user + conversation, idempotent bytoolCallId(safe across graph resume/replay — never double-dispatches), capped per conversation, and TTL-swept. The detached promise is.catch-guarded (no unhandled rejections).Scope / follow-ups (intentionally out of this PR)
injectedMessages) is a separate effort; queue/steer is not a prerequisite.read_file/skill/ memory /ask_user_question/ PTC /tool_search/ handoffs — they break on a synthetic (artifact-less) result or run off the event-driven path.mcp_allephemeral MCP servers: tool names aren't known at synthesis time, so they aren't marked.librechat.yaml/ model spec; saved agents viatool_options).Testing
background.spec.ts(25) — eligibility,run_in_backgrounddetect/strip, schema injection (non-mutating),applyBackgroundToolCallsgating, per-tool synthesis, registry (create/complete/fail/idempotency/capacity/scoping), andcheck_background_taskoutput.handlers.background.spec.ts(3) — drives the realcreateToolExecuteHandler: returns a handle immediately, runs the tool exactly once detached (without the injected flag), poll returns the result; idempotent on replay; synchronous when not requested.load.spec.ts(+2) — ephemeral flag and model-specrunInBackgroundsynthesizetool_optionsfor eligible tools only.tsc+eslintclean acrosspackages/apianddata-provider; no regressions in existing validation/handlers/load suites.@codex review