The short-circuit block in runTeam() called this.runAgent(), which emits
its own agent_start/agent_complete events and increments completedTaskCount.
The short-circuit block then emitted the same events again, and
buildTeamRunResult() incremented the count a second time.
Fix: call buildAgent() + agent.run() directly, bypassing runAgent().
Events and counting are handled once by the short-circuit block and
buildTeamRunResult() respectively.
Addresses all five review points from @JackChen-me on PR #70:
1. Extract shared keyword helpers into src/utils/keywords.ts so the
short-circuit selector and Scheduler.capability-match cannot drift.
Both orchestrator.ts and scheduler.ts now import the same module.
2. selectBestAgent now mirrors Scheduler.capability-match exactly,
including the asymmetric use of agent.model: agentKeywords includes
model, agentText does not. This restores parity with the documented
capability-match behaviour.
3. Remove isSimpleGoal and selectBestAgent from the public barrel
(src/index.ts). They remain exported from orchestrator.ts for unit
tests but are no longer part of the package API surface.
4. Forward the AbortSignal from runTeam(options) through the
short-circuit path. runAgent() now accepts an optional
{ abortSignal } argument; runTeam's short-circuit branch passes
the caller's signal so cancellation works for simple goals too.
5. Tighten the collaborate/coordinate complexity regexes so they only
fire on imperative directives ("collaborate with X", "coordinate
the team") and not on descriptive uses ("explain how pods
coordinate", "what is microservice collaboration").
Also fixes a pre-existing test failure in token-budget.test.ts:
"enforces orchestrator budget in runTeam" was using "Do work" as its
goal which now short-circuits, so the coordinator path the test was
exercising never ran. Switched to a multi-step goal.
Adds 60 new tests across short-circuit.test.ts and the new
keywords.test.ts covering all five fixes.
Co-Authored-By: Claude <noreply@anthropic.com>
When a goal is short (<200 chars) and contains no multi-step or
coordination signals, runTeam() now dispatches directly to the
best-matching agent — skipping the coordinator decomposition and
synthesis round-trips. This saves ~2 LLM calls worth of tokens
and latency for genuinely simple goals.
Complexity detection uses regex patterns for sequencing markers
(first...then, step N, numbered lists), coordination language
(collaborate, coordinate, work together), parallel execution
signals, and multi-deliverable patterns.
Agent selection reuses the same keyword-affinity scoring as the
capability-match scheduler strategy to pick the most relevant
agent from the team roster.
- Add isSimpleGoal() and selectBestAgent() (exported for testing)
- Add 35 unit tests covering heuristic edge cases and integration
- Update existing runTeam tests to use complex goals
Co-Authored-By: Claude <noreply@anthropic.com>
AgentPool now maintains a per-agent Semaphore(1) that serializes
concurrent run() calls targeting the same Agent. This prevents
shared-state races on Agent.state (status, messages, tokenUsage)
when multiple independent tasks are assigned to the same agent.
Lock acquisition order: per-agent lock first, then pool semaphore,
so queued tasks don't waste pool slots while waiting.
Fixes#61
Thread AbortSignal from the top-level API through RunContext to
executeQueue(), enabling graceful cancellation in Express, Next.js,
serverless, and CLI scenarios.
Changes:
- Added optional to RunContext interface
- now accepts
- now accepts
- executeQueue() checks signal.aborted before each dispatch round
and skips remaining tasks when cancelled
- Signal is forwarded to coordinator's run() and per-task pool.run()
so in-flight LLM calls are also cancelled
- Full backward compatibility: both methods work without options
The abort infrastructure already existed at lower layers
(AgentRunner, Agent, AgentPool) — this commit bridges the last gap
at the orchestrator level.
Co-authored-by: JasonOA888 <JasonOA888@users.noreply.github.com>
* feat(agent): add smart loop detection for stuck agents (#16)
Detect when agents repeat the same tool calls or text outputs in a
sliding window. Three modes: warn (inject nudge, terminate on 2nd hit),
terminate (immediate stop), or custom callback. Fully opt-in via
`loopDetection` on AgentConfig — zero overhead when unconfigured.
* fix(agent): support async onLoopDetected callbacks and prevent orphaned tool_use events
- Await onLoopDetected callback result so async functions work correctly
instead of silently falling through to 'continue'
- Move loop detection before yielding tool_use events so terminate mode
never emits tool_use without a matching tool_result
* fix(agent): reset loopWarned on recovery and rename maxRepeatedToolCalls to maxRepetitions
- Reset loopWarned flag when the agent stops repeating, so a future
loop gets a fresh warning cycle instead of immediate termination
- Rename maxRepeatedToolCalls → maxRepetitions since the threshold
applies to both tool call and text output repetition detection
* test(agent): add tests for async callback, warn recovery, and injected warning text
- Verify async onLoopDetected callback is awaited correctly
- Verify loopWarned resets after recovery, giving fresh warning cycle
- Verify WARNING TextBlock is injected into user message content
When both timeoutMs and a caller-provided abortSignal were set, the
timeout signal silently replaced the caller's signal. Now they are
combined via mergeAbortSignals() so either source can cancel the run.
Also removes dead array-handling branch in text-tool-extractor.ts
(extractJSONObjects only returns objects, never arrays).
Local models (Ollama, vLLM) sometimes return tool calls as text instead
of using the native tool_calls wire format. This adds a safety-net
extractor that parses tool calls from model text output when native
tool_calls is empty.
- Add text-tool-extractor with support for bare JSON, code fences,
and Hermes <tool_call> tags
- Wire fallback into OpenAI adapter chat() and stream() paths
- Add onWarning callback when model ignores configured tools
- Add timeoutMs on AgentConfig for per-run abort (local models can
be slow)
- Add 26 tests for extractor and fallback behavior
- Document local model compatibility in README
* feat(orchestrator): add onApproval callback for human-in-the-loop (#32)
Add an optional `onApproval` callback to OrchestratorConfig that gates
between task execution rounds. After each batch of parallel tasks
completes, the callback receives the completed tasks and the tasks about
to start, returning true to continue or false to abort gracefully.
Key changes:
- Add 'skipped' to TaskStatus for user-initiated abort (distinct from 'failed')
- Add skip(), skipRemaining(), cascadeSkip() to TaskQueue
- Add 'task_skipped' to OrchestratorEvent for progress monitoring
- Approval gate in executeQueue() with try/catch for callback errors
- Synthesis prompt now includes skipped tasks section
- 17 new tests covering queue skip operations and orchestrator integration
Closes#32
* docs: clarify onApproval contract and add missing test scenarios
- Document skip() cascade semantics, skipRemaining() in-flight constraint,
and onApproval trigger conditions / mutation warning
- Add concurrency safety comment on completedThisRound
- Note task_skipped as breaking union addition on OrchestratorEvent
- Add 3 test scenarios: single-batch no-callback, mixed success/failure
batch, and onProgress task_skipped event relay
* feat(agent): add beforeRun / afterRun lifecycle hooks (#31)
Add optional hook callbacks to AgentConfig for cross-cutting concerns
(guardrails, logging, token budgets) without modifying framework internals.
- beforeRun: receives prompt + agent config, can modify or throw to abort
- afterRun: receives AgentRunResult, can modify or throw to fail
- Works with all three execution modes: run(), prompt(), stream()
- 15 test cases covering modify, throw, async, composition, and history integrity
* fix(agent): preserve non-text content blocks in beforeRun hook
- applyHookContext now replaces only text blocks, keeping images and
tool results intact (was silently stripping them)
- Use backward loop instead of reverse() + find() for efficiency
- Clarify JSDoc that only `prompt` is applied from hook return value
- Add test for mixed-content user messages
* fix(agent): address review feedback on beforeRun/afterRun hooks
- Normalize stream done event to always yield AgentRunResult
- Move transitionTo('completed') after afterRun to fix state ordering
- Strip hook functions from BeforeRunHookContext.agent to avoid self-references
- Pass originalPrompt to applyHookContext to avoid redundant message scan
- Clarify afterRun JSDoc: not called when the run throws
- Add tests: error-path skip, outputSchema+afterRun, ctx.agent shape, multi-turn hooks
Add lightweight onTrace callback to OrchestratorConfig that emits
structured span events (llm_call, tool_call, task, agent) with timing,
token usage, and runId correlation. Zero overhead when not subscribed.
Closes#18
Use Number.isFinite() to sanitize maxRetries, retryDelayMs, and
retryBackoff before entering the retry loop. Prevents unbounded
retries from Infinity or broken loop bounds from NaN.
* feat: add task-level retry with exponential backoff
Add `maxRetries`, `retryDelayMs`, and `retryBackoff` to task config.
When a task fails and retries remain, the orchestrator waits with
exponential backoff and re-runs the task with a fresh agent conversation.
A `task_retry` event is emitted via `onProgress` for observability.
Cascade failure only occurs after all retries are exhausted.
Closes#30
* fix: address review — extract executeWithRetry, add delay cap, fix tests
- Extract `executeWithRetry()` as a testable exported function
- Add `computeRetryDelay()` with 30s max cap (prevents runaway backoff)
- Remove retry fields from `ParsedTaskSpec` (dead code for runTeam path)
- Deduplicate retry event emission (single code path for both error types)
- Injectable delay function for test determinism
- Rewrite tests to call the real `executeWithRetry`, not a copy
- 15 tests covering: success, retry+success, retry+failure, backoff
calculation, delay cap, delay function injection, no-retry default
* fix: clamp negative maxRetries/retryBackoff to safe values
- maxRetries clamped to >= 0 (negative values treated as no retry)
- retryBackoff clamped to >= 1 (prevents zero/negative delay oscillation)
- retryDelayMs clamped to >= 0
- Add tests for negative maxRetries and negative backoff
Addresses Codex review P1 on #37
* fix: accumulate token usage across retry attempts
Previously only the final attempt's tokenUsage was returned, causing
under-reporting of actual model consumption when retries occurred.
Now all attempts' token counts are summed in the returned result.
Addresses Codex review P2 (token usage) on #37
- Include error feedback user turn in mergedMessages to maintain
alternating user/assistant roles required by Anthropic API
- Use explicit undefined check instead of ?? for structured merge
to preserve null as a valid structured output value
When `outputSchema` is set on AgentConfig, the agent's final text output
is parsed as JSON, validated against the Zod schema, and exposed via
`result.structured`. On validation failure a single retry with error
feedback is attempted automatically.
Closes#29
crypto.randomUUID() is not globally available in Node 18. Import
randomUUID from node:crypto explicitly so the framework works on
all supported Node versions (>=18).
isTaskReady() rejects non-pending tasks on its first line, but
unblockDependents() passed blocked tasks directly to it. This meant
dependent tasks stayed blocked forever after their dependencies
completed, breaking any workflow with task dependencies.
Fix: pass a pending-status copy so isTaskReady only checks the
dependency condition.
Enable connecting to any OpenAI-compatible API (Ollama, vLLM, LM Studio,
etc.) by adding baseURL and apiKey fields to AgentConfig and
OrchestratorConfig, threaded through to adapter constructors.
- OpenAIAdapter and AnthropicAdapter accept optional baseURL
- createAdapter() forwards baseURL to both adapters, warns if used with copilot
- All execution paths (runAgent, runTeam coordinator, buildPool) merge defaults
- Fully backward compatible — omitting new fields preserves existing behavior
1. Fix header comment — document correct env var precedence
(apiKey → GITHUB_COPILOT_TOKEN → GITHUB_TOKEN → device flow)
2. Use application/x-www-form-urlencoded for device code endpoint
3. Use application/x-www-form-urlencoded for poll endpoint
4. Add mutex (promise-based) on #getSessionToken to prevent
concurrent token refreshes and duplicate device flow prompts
5. Add DeviceCodeCallback + CopilotAdapterOptions so callers can
control device flow output instead of hardcoded console.log
6. Extract shared OpenAI wire-format helpers into openai-common.ts,
imported by both openai.ts and copilot.ts (-142 lines net)
7. Update createAdapter JSDoc to mention copilot env vars
- Add CopilotAdapter with OAuth2 device flow authentication
- Token exchange via /copilot_internal/v2/token with caching
- Premium request multiplier system (getCopilotMultiplier)
- Full model metadata catalog (COPILOT_MODELS)
- Add 'copilot' to SupportedProvider and provider union types
- Add example: examples/05-copilot-test.ts
Ibrahim Kazimov
JackChen
MrAvalonApple
JackChen
EchoOfZion
Jason
Marcelo Ceccon
Deathwing