@botpress/evals 1.19.0-beta.6 → 1.19.0-beta.9

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
@@ -1 +1 @@
1
- {"version":3,"file":"sse-collector.d.ts","sourceRoot":"","sources":["../src/sse-collector.ts"],"names":[],"mappings":"AAAA;;;;;;;GAOG;AAEH,OAAO,KAAK,EAAE,IAAI,EAAE,MAAM,SAAS,CAAA;AAEnC,KAAK,SAAS,GAAG;IAAE,cAAc,EAAE,MAAM,CAAA;CAAE,GAAG;IAAE,UAAU,EAAE,MAAM,CAAA;CAAE,CAAA;AAEpE,UAAU,WAAW;IACnB,OAAO,EAAE,MAAM,CAAA;IACf;;;;;OAKG;IACH,WAAW,CAAC,EAAE,WAAW,CAAA;CAC1B;AAED,UAAU,mBAAoB,SAAQ,WAAW;IAC/C,MAAM,EAAE,SAAS,GAAG,WAAW,CAAA;CAChC;AAgBD,qBAAa,YAAY;IACvB,OAAO,CAAC,YAAY,CAAQ;IAC5B,OAAO,CAAC,OAAO,CAAwB;IAEvC,yEAAyE;IACzE,OAAO,CAAC,OAAO,CAA0B;IAEzC,iFAAiF;IACjF,OAAO,CAAC,aAAa,CAAoB;IAEzC,+EAA+E;IAC/E,OAAO,CAAC,eAAe,CAAoB;IAE3C,0CAA0C;IAC1C,OAAO,CAAC,OAAO,CAAkC;IAEjD,qDAAqD;IACrD,OAAO,CAAC,eAAe,CAA+B;IAEtD,8CAA8C;IAC9C,OAAO,CAAC,SAAS,CAAQ;IAEzB,yDAAyD;IACzD,OAAO,CAAC,WAAW,CAAqB;IAExC,sFAAsF;IACtF,OAAO,CAAC,aAAa,CAA4B;gBAErC,YAAY,EAAE,MAAM,EAAE,OAAO,GAAE,MAAM,CAAC,MAAM,EAAE,MAAM,CAAM;IAKtE;;;OAGG;IACG,OAAO,CAAC,MAAM,EAAE,SAAS,GAAG,OAAO,CAAC,IAAI,CAAC;IAoC/C;;;;OAIG;IACH,SAAS,IAAI,IAAI;IAYjB,OAAO,CAAC,MAAM,CAAC,QAAQ,CAAC,SAAS,CAAM;IAEvC;;;;;;;;OAQG;IACG,mBAAmB,CAAC,IAAI,EAAE,WAAW,GAAG,OAAO,CAAC,IAAI,CAAC;IAuC3D,OAAO,CAAC,MAAM,CAAC,QAAQ,CAAC,aAAa,CAAQ;IAE7C;;;;;;OAMG;IACH,OAAO,CAAC,mBAAmB;IA4B3B;;;;OAIG;IACG,eAAe,CAAC,IAAI,EAAE,MAAM,EAAE,IAAI,EAAE,mBAAmB,GAAG,OAAO,CAAC,IAAI,CAAC;IAyC7E;;OAEG;IACH,YAAY,IAAI,IAAI,EAAE;IAUtB;;OAEG;IACH,WAAW,IAAI,IAAI,EAAE;IAIrB;;OAEG;IACH,UAAU,IAAI,IAAI;IAgBlB,qDAAqD;IACrD,OAAO,CAAC,gBAAgB,CAAQ;IAChC,OAAO,CAAC,iBAAiB,CAAwB;IAEjD,OAAO,CAAC,MAAM,CAAC,QAAQ,CAAC,mBAAmB,CAAS;IAEpD;;;OAGG;IACH,OAAO,CAAC,gBAAgB;IA2BxB;;OAEG;YACW,cAAc;IA2B5B;;;OAGG;IACH,OAAO,CAAC,cAAc;IA+BtB;;OAEG;IACH,OAAO,CAAC,YAAY;IAsBpB;;OAEG;IACH,OAAO,CAAC,WAAW;IAInB;;OAEG;IACH,OAAO,CAAC,cAAc;IAQtB;;OAEG;IACH,OAAO,CAAC,aAAa;IAOrB;;;;;;;;;OASG;IACH,OAAO,CAAC,uBAAuB;IAc/B;;OAEG;IACH,OAAO,CAAC,yBAAyB;CAmBlC"}
1
+ {"version":3,"file":"sse-collector.d.ts","sourceRoot":"","sources":["../src/sse-collector.ts"],"names":[],"mappings":"AAAA;;;;;;;GAOG;AAEH,OAAO,KAAK,EAAE,IAAI,EAAE,MAAM,SAAS,CAAA;AAGnC,KAAK,SAAS,GAAG;IAAE,cAAc,EAAE,MAAM,CAAA;CAAE,GAAG;IAAE,UAAU,EAAE,MAAM,CAAA;CAAE,CAAA;AAEpE,UAAU,WAAW;IACnB,OAAO,EAAE,MAAM,CAAA;IACf;;;;;OAKG;IACH,WAAW,CAAC,EAAE,WAAW,CAAA;CAC1B;AAED,UAAU,mBAAoB,SAAQ,WAAW;IAC/C,MAAM,EAAE,SAAS,GAAG,WAAW,CAAA;CAChC;AAgBD,qBAAa,YAAY;IACvB,OAAO,CAAC,YAAY,CAAQ;IAC5B,OAAO,CAAC,OAAO,CAAwB;IAEvC,yEAAyE;IACzE,OAAO,CAAC,OAAO,CAA0B;IAEzC,iFAAiF;IACjF,OAAO,CAAC,aAAa,CAAoB;IAEzC,+EAA+E;IAC/E,OAAO,CAAC,eAAe,CAAoB;IAE3C,0CAA0C;IAC1C,OAAO,CAAC,OAAO,CAAkC;IAEjD,qDAAqD;IACrD,OAAO,CAAC,eAAe,CAA+B;IAEtD,8CAA8C;IAC9C,OAAO,CAAC,SAAS,CAAQ;IAEzB,yDAAyD;IACzD,OAAO,CAAC,WAAW,CAAqB;IAExC,sFAAsF;IACtF,OAAO,CAAC,aAAa,CAA4B;gBAErC,YAAY,EAAE,MAAM,EAAE,OAAO,GAAE,MAAM,CAAC,MAAM,EAAE,MAAM,CAAM;IAKtE;;;OAGG;IACG,OAAO,CAAC,MAAM,EAAE,SAAS,GAAG,OAAO,CAAC,IAAI,CAAC;IAyC/C;;;;OAIG;IACH,SAAS,IAAI,IAAI;IAYjB,OAAO,CAAC,MAAM,CAAC,QAAQ,CAAC,SAAS,CAAM;IAEvC;;;;;;;;OAQG;IACG,mBAAmB,CAAC,IAAI,EAAE,WAAW,GAAG,OAAO,CAAC,IAAI,CAAC;IAuC3D,OAAO,CAAC,MAAM,CAAC,QAAQ,CAAC,aAAa,CAAQ;IAE7C;;;;;;OAMG;IACH,OAAO,CAAC,mBAAmB;IA4B3B;;;;OAIG;IACG,eAAe,CAAC,IAAI,EAAE,MAAM,EAAE,IAAI,EAAE,mBAAmB,GAAG,OAAO,CAAC,IAAI,CAAC;IAyC7E;;OAEG;IACH,YAAY,IAAI,IAAI,EAAE;IAUtB;;OAEG;IACH,WAAW,IAAI,IAAI,EAAE;IAIrB;;OAEG;IACH,UAAU,IAAI,IAAI;IAgBlB,qDAAqD;IACrD,OAAO,CAAC,gBAAgB,CAAQ;IAChC,OAAO,CAAC,iBAAiB,CAAwB;IAEjD,OAAO,CAAC,MAAM,CAAC,QAAQ,CAAC,mBAAmB,CAAS;IAEpD;;;OAGG;IACH,OAAO,CAAC,gBAAgB;IA2BxB;;OAEG;YACW,cAAc;IA2B5B;;;OAGG;IACH,OAAO,CAAC,cAAc;IA+BtB;;OAEG;IACH,OAAO,CAAC,YAAY;IAsBpB;;OAEG;IACH,OAAO,CAAC,WAAW;IAInB;;OAEG;IACH,OAAO,CAAC,cAAc;IAQtB;;OAEG;IACH,OAAO,CAAC,aAAa;IAOrB;;;;;;;;;OASG;IACH,OAAO,CAAC,uBAAuB;IAc/B;;OAEG;IACH,OAAO,CAAC,yBAAyB;CAmBlC"}
@@ -1,3 +1,46 @@
1
+ // ../analytics/src/errors.ts
2
+ var AdkError = class extends Error {
3
+ /**
4
+ * Marker for cross-module-instance checks — `instanceof` breaks when a
5
+ * package is duplicated in the dependency graph.
6
+ * @internal
7
+ */
8
+ static __IS_ADK_BASE_ERROR = true;
9
+ /**
10
+ * Stable SCREAMING_SNAKE_CASE identifier, e.g. 'BOT_NOT_FOUND'. Subclasses
11
+ * pin `Code` to a literal union (`class EvalRunnerError extends
12
+ * AdkError<EvalErrorCode>`) so typos are compile errors and `code` narrows
13
+ * in switches.
14
+ */
15
+ code;
16
+ /**
17
+ * True for user/environment conditions the developer can fix themselves
18
+ * (validation, auth, connectivity). False for internal bugs. Defaults to
19
+ * false so unclassified errors surface as bugs until someone triages them.
20
+ */
21
+ expected;
22
+ /** Structured context. Sanitized before any telemetry leaves the machine. */
23
+ details;
24
+ /** Actionable next step shown to the user, e.g. "Run 'adk login' first." */
25
+ suggestion;
26
+ constructor(opts) {
27
+ super(opts.message, opts.cause !== void 0 ? { cause: opts.cause } : void 0);
28
+ this.name = this.constructor.name;
29
+ this.code = opts.code;
30
+ this.expected = opts.expected ?? false;
31
+ if (opts.details !== void 0) {
32
+ this.details = opts.details;
33
+ }
34
+ if (opts.suggestion !== void 0) {
35
+ this.suggestion = opts.suggestion;
36
+ }
37
+ }
38
+ };
39
+
40
+ // src/errors.ts
41
+ var EvalRunnerError = class extends AdkError {
42
+ };
43
+
1
44
  // src/sse-collector.ts
2
45
  function spanKey(span) {
3
46
  return `${span.id.trace}:${span.id.span}`;
@@ -42,10 +85,15 @@ var SSECollector = class _SSECollector {
42
85
  signal: this.abortController.signal
43
86
  });
44
87
  if (!response.ok) {
45
- throw new Error(`SSE connection failed: ${response.status} ${response.statusText}`);
88
+ throw new EvalRunnerError({
89
+ code: "SSE_CONNECT_FAILED",
90
+ message: `SSE connection failed: ${response.status} ${response.statusText}`,
91
+ expected: true,
92
+ suggestion: "Make sure the dev server is running (`adk dev`) and reachable."
93
+ });
46
94
  }
47
95
  if (!response.body) {
48
- throw new Error("SSE response has no body");
96
+ throw new EvalRunnerError({ code: "SSE_NO_BODY", message: "SSE response has no body" });
49
97
  }
50
98
  this.connected = true;
51
99
  this._consumeStream(response.body).catch((err) => {
@@ -1,7 +1,7 @@
1
1
  {
2
2
  "version": 3,
3
- "sources": ["../src/sse-collector.ts"],
4
- "sourcesContent": ["/**\n * SSECollector \u2014 connects to the ADK dev server's SSE trace stream,\n * accumulates spans, and provides promise-based wait methods for\n * turn completion and workflow signals.\n *\n * Uses fetch() + async iteration with manual SSE line parsing\n * (Bun does not have native EventSource).\n */\n\nimport type { Span } from './trace'\n\ntype SSEFilter = { conversationId: string } | { workflowId: string }\n\ninterface WaitOptions {\n timeout: number\n /**\n * Optional abort signal \u2014 when fired, the wait rejects immediately rather\n * than burning the full timeout. Used by the runner so Stop unblocks\n * pending workflow assertions (otherwise a single negative-result wait\n * can hold the suite hostage for `idleTimeout` per assertion).\n */\n abortSignal?: AbortSignal\n}\n\ninterface WorkflowWaitOptions extends WaitOptions {\n signal: 'entered' | 'completed'\n}\n\n/**\n * Unique key for span deduplication.\n */\nfunction spanKey(span: Span): string {\n return `${span.id.trace}:${span.id.span}`\n}\n\n/**\n * Type-safe accessor for span.data fields.\n */\nfunction spanData(span: Span): Record<string, unknown> {\n return span.data && typeof span.data === 'object' ? (span.data as Record<string, unknown>) : {}\n}\n\nexport class SSECollector {\n private devServerUrl: string\n private headers: Record<string, string>\n\n /** All spans accumulated since connect(), keyed by spanKey for dedup. */\n private spanMap = new Map<string, Span>()\n\n /** Index into spanMap insertion order marking where the current turn started. */\n private turnStartKeys = new Set<string>()\n\n /** Handler span IDs seen before the current turn (for startTurn filtering). */\n private seenHandlerKeys = new Set<string>()\n\n /** Callbacks waiting for span updates. */\n private waiters: Array<(span: Span) => void> = []\n\n /** AbortController for the active SSE connection. */\n private abortController: AbortController | null = null\n\n /** Whether the background loop is running. */\n private connected = false\n\n /** Error from the SSE stream if it died unexpectedly. */\n private streamError: Error | null = null\n\n /** Set while a settle phase is in flight; called by disconnect() to short-circuit. */\n private _cancelSettle: (() => void) | null = null\n\n constructor(devServerUrl: string, headers: Record<string, string> = {}) {\n this.devServerUrl = devServerUrl\n this.headers = headers\n }\n\n /**\n * Open a streaming SSE connection filtered by conversationId or workflowId.\n * Starts a background loop that parses SSE events and accumulates spans.\n */\n async connect(filter: SSEFilter): Promise<void> {\n const attributeName = 'conversationId' in filter ? 'conversationId' : 'workflowId'\n const attributeValue = 'conversationId' in filter ? filter.conversationId : filter.workflowId\n\n const url = `${this.devServerUrl}/api/traces/stream?attributeName=${encodeURIComponent(attributeName)}&attributeValue=${encodeURIComponent(attributeValue)}`\n\n this.abortController = new AbortController()\n\n const response = await fetch(url, {\n headers: { ...this.headers, Accept: 'text/event-stream' },\n signal: this.abortController.signal,\n })\n\n if (!response.ok) {\n throw new Error(`SSE connection failed: ${response.status} ${response.statusText}`)\n }\n\n if (!response.body) {\n throw new Error('SSE response has no body')\n }\n\n this.connected = true\n\n // Start background parsing loop\n this._consumeStream(response.body).catch((err) => {\n if (this.connected) {\n this.streamError = err instanceof Error ? err : new Error(String(err))\n }\n })\n\n // Wait for the initial snapshot to arrive before returning.\n // This ensures that connect() resolves only after the snapshot\n // has been processed so callers can immediately check spans.\n await this._waitForSnapshot()\n }\n\n /**\n * Mark the beginning of a new turn.\n * Records which handler.conversation span IDs have been seen so far,\n * so waitForTurnComplete() only resolves on NEW handler completions.\n */\n startTurn(): void {\n // Snapshot the current set of span keys as the turn boundary\n this.turnStartKeys = new Set(this.spanMap.keys())\n\n // Record all handler.conversation spans seen so far\n for (const [key, span] of this.spanMap) {\n if (span.name === 'handler.conversation') {\n this.seenHandlerKeys.add(key)\n }\n }\n }\n\n private static readonly SETTLE_MS = 100\n\n /**\n * Wait for a handler.conversation span to reach terminal status (ok or error).\n * Only considers spans NOT seen before the last startTurn() call.\n *\n * After the handler closes, waits for a quiet period (no new spans for SETTLE_MS)\n * to ensure all child spans (tool calls, messages, state saves) have arrived.\n * The runtime's HttpSpanExporter fires POSTs without awaiting them, so child\n * spans may arrive slightly after the handler close event.\n */\n async waitForTurnComplete(opts: WaitOptions): Promise<void> {\n // Check if a completed handler already exists (e.g. from snapshot)\n if (this._hasNewCompletedHandler()) {\n return this._settleAfterHandler()\n }\n if (opts.abortSignal?.aborted) {\n throw new Error('waitForTurnComplete aborted')\n }\n\n return new Promise<void>((resolve, reject) => {\n const timer = setTimeout(() => {\n this._removeWaiter(waiter)\n opts.abortSignal?.removeEventListener('abort', onAbort)\n const reason = this.streamError\n ? `SSE stream died: ${this.streamError.message}`\n : 'no handler.conversation span completed'\n reject(new Error(`waitForTurnComplete timed out after ${opts.timeout}ms \u2014 ${reason}`))\n }, opts.timeout)\n\n const onAbort = () => {\n clearTimeout(timer)\n this._removeWaiter(waiter)\n reject(new Error('waitForTurnComplete aborted'))\n }\n opts.abortSignal?.addEventListener('abort', onAbort, { once: true })\n\n const waiter = (_span: Span) => {\n if (this._hasNewCompletedHandler()) {\n clearTimeout(timer)\n this._removeWaiter(waiter)\n opts.abortSignal?.removeEventListener('abort', onAbort)\n this._settleAfterHandler().then(resolve, reject)\n }\n }\n\n this.waiters.push(waiter)\n })\n }\n\n private static readonly MAX_SETTLE_MS = 2_000\n\n /**\n * After the handler span closes, wait for spans to stop arriving.\n * Resolves when no new spans arrive for SETTLE_MS, or after MAX_SETTLE_MS hard cap.\n *\n * Exposes a cancellation hook via _cancelSettle so disconnect() can short-\n * circuit an in-flight settle phase and avoid leaking timers for up to 2s.\n */\n private _settleAfterHandler(): Promise<void> {\n return new Promise<void>((resolve) => {\n let resolved = false\n const done = () => {\n if (resolved) return\n resolved = true\n clearTimeout(hardCap)\n clearTimeout(settleTimer)\n this._removeWaiter(settleWaiter)\n this._cancelSettle = null\n resolve()\n }\n\n const hardCap = setTimeout(done, SSECollector.MAX_SETTLE_MS)\n\n let settleTimer = setTimeout(done, SSECollector.SETTLE_MS)\n\n const settleWaiter = () => {\n if (resolved) return\n clearTimeout(settleTimer)\n settleTimer = setTimeout(done, SSECollector.SETTLE_MS)\n }\n\n this.waiters.push(settleWaiter)\n this._cancelSettle = done\n })\n }\n\n /**\n * Wait for a handler.workflow span with matching workflow name.\n * - signal 'entered': resolves when the span appears (any status)\n * - signal 'completed': resolves when the span has status !== 'running'\n */\n async waitForWorkflow(name: string, opts: WorkflowWaitOptions): Promise<void> {\n // Check if condition is already met\n if (this._matchesWorkflowCondition(name, opts.signal)) {\n return\n }\n if (opts.abortSignal?.aborted) {\n throw new Error(`waitForWorkflow(\"${name}\", { signal: \"${opts.signal}\" }) aborted`)\n }\n\n return new Promise<void>((resolve, reject) => {\n const timer = setTimeout(() => {\n this._removeWaiter(waiter)\n opts.abortSignal?.removeEventListener('abort', onAbort)\n const reason = this.streamError ? `SSE stream died: ${this.streamError.message}` : 'span not found'\n reject(\n new Error(\n `waitForWorkflow(\"${name}\", { signal: \"${opts.signal}\" }) timed out after ${opts.timeout}ms \u2014 ${reason}`\n )\n )\n }, opts.timeout)\n\n const onAbort = () => {\n clearTimeout(timer)\n this._removeWaiter(waiter)\n reject(new Error(`waitForWorkflow(\"${name}\", { signal: \"${opts.signal}\" }) aborted`))\n }\n opts.abortSignal?.addEventListener('abort', onAbort, { once: true })\n\n const waiter = (_span: Span) => {\n if (this._matchesWorkflowCondition(name, opts.signal)) {\n clearTimeout(timer)\n this._removeWaiter(waiter)\n opts.abortSignal?.removeEventListener('abort', onAbort)\n resolve()\n }\n }\n\n this.waiters.push(waiter)\n })\n }\n\n /**\n * Return spans accumulated since the last startTurn() call.\n */\n getTurnSpans(): Span[] {\n const result: Span[] = []\n for (const [key, span] of this.spanMap) {\n if (!this.turnStartKeys.has(key)) {\n result.push(span)\n }\n }\n return result\n }\n\n /**\n * Return all spans accumulated since connect().\n */\n getAllSpans(): Span[] {\n return Array.from(this.spanMap.values())\n }\n\n /**\n * Close the SSE connection and clean up.\n */\n disconnect(): void {\n this.connected = false\n this.abortController?.abort()\n this.abortController = null\n // Short-circuit an in-flight settle phase so its hardCap / settleTimer\n // don't fire after disconnect.\n this._cancelSettle?.()\n this._cancelSettle = null\n // Waiters hold setTimeout references that will reject on their own;\n // clearing the array just prevents further notifications.\n this.waiters = []\n this.snapshotResolvers = []\n }\n\n // \u2500\u2500 Private \u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\n\n /** Whether we have received the initial snapshot. */\n private snapshotReceived = false\n private snapshotResolvers: Array<() => void> = []\n\n private static readonly SNAPSHOT_TIMEOUT_MS = 10_000\n\n /**\n * Returns a promise that resolves once the snapshot event has been processed,\n * or rejects after a timeout to prevent connect() from hanging indefinitely.\n */\n private _waitForSnapshot(): Promise<void> {\n if (this.snapshotReceived) {\n return Promise.resolve()\n }\n return new Promise<void>((resolve, reject) => {\n let settled = false\n const resolver = () => {\n if (settled) return\n settled = true\n clearTimeout(timer)\n resolve()\n }\n const timer = setTimeout(() => {\n if (settled) return\n settled = true\n const idx = this.snapshotResolvers.indexOf(resolver)\n if (idx !== -1) this.snapshotResolvers.splice(idx, 1)\n reject(\n new Error(\n `SSE snapshot not received within ${SSECollector.SNAPSHOT_TIMEOUT_MS}ms \u2014 is the dev server healthy?`\n )\n )\n }, SSECollector.SNAPSHOT_TIMEOUT_MS)\n this.snapshotResolvers.push(resolver)\n })\n }\n\n /**\n * Consume the ReadableStream body, parse SSE lines, and dispatch events.\n */\n private async _consumeStream(body: ReadableStream<Uint8Array>): Promise<void> {\n const decoder = new TextDecoder()\n let buffer = ''\n\n const reader = body.getReader()\n\n try {\n while (this.connected) {\n const { done, value } = await reader.read()\n if (done) break\n\n buffer += decoder.decode(value, { stream: true })\n\n // Split on double-newline boundaries to get complete SSE events\n const parts = buffer.split('\\n\\n')\n // The last part may be incomplete \u2014 keep it in the buffer\n buffer = parts.pop() ?? ''\n\n for (const part of parts) {\n this._parseSSEBlock(part)\n }\n }\n } finally {\n reader.releaseLock()\n }\n }\n\n /**\n * Parse a single SSE event block (text between \\n\\n boundaries).\n * Extracts event name and data payload.\n */\n private _parseSSEBlock(block: string): void {\n const lines = block.split('\\n')\n let eventName = ''\n let dataLine = ''\n\n for (const line of lines) {\n if (line.startsWith('event:')) {\n eventName = line.slice('event:'.length).trim()\n } else if (line.startsWith('data:')) {\n const value = line.slice('data:'.length).trim()\n dataLine = dataLine ? `${dataLine}\\n${value}` : value\n }\n // id: and retry: lines are ignored\n }\n\n if (!eventName || !dataLine) {\n // Could be a retry: line or empty block \u2014 skip\n if (!this.snapshotReceived && block.includes('retry:')) {\n // The retry line comes before the snapshot; don't resolve yet\n }\n return\n }\n\n try {\n const payload = JSON.parse(dataLine)\n this._handleEvent(eventName, payload)\n } catch {\n // Malformed JSON \u2014 skip\n }\n }\n\n /**\n * Dispatch a parsed SSE event to the appropriate handler.\n */\n private _handleEvent(eventName: string, payload: unknown): void {\n if (eventName === 'snapshot') {\n const data = payload as { spans: Span[] }\n const spans = Array.isArray(data.spans) ? data.spans : []\n for (const span of spans) {\n this._upsertSpan(span)\n }\n this.snapshotReceived = true\n for (const resolve of this.snapshotResolvers) {\n resolve()\n }\n this.snapshotResolvers = []\n } else if (eventName === 'update') {\n const data = payload as { span: Span }\n if (data.span) {\n this._upsertSpan(data.span)\n this._notifyWaiters(data.span)\n }\n }\n // keepalive events are intentionally ignored\n }\n\n /**\n * Insert or replace a span in the map (dedup by trace+span ID).\n */\n private _upsertSpan(span: Span): void {\n this.spanMap.set(spanKey(span), span)\n }\n\n /**\n * Notify all waiters that a new span was received.\n */\n private _notifyWaiters(span: Span): void {\n // Copy the array since waiters may remove themselves during iteration\n const current = [...this.waiters]\n for (const waiter of current) {\n waiter(span)\n }\n }\n\n /**\n * Remove a waiter callback from the list.\n */\n private _removeWaiter(fn: (span: Span) => void): void {\n const idx = this.waiters.indexOf(fn)\n if (idx !== -1) {\n this.waiters.splice(idx, 1)\n }\n }\n\n /**\n * Check if there is a handler.conversation span that:\n * 1. Was NOT seen before startTurn()\n * 2. Has terminal status (ok or error)\n *\n * Marks the matched key as consumed so the same handler span cannot\n * satisfy both the current turn's settle phase AND the next turn's\n * waitForTurnComplete check (which would otherwise resolve instantly\n * against stale data instead of waiting for the new turn's handler).\n */\n private _hasNewCompletedHandler(): boolean {\n for (const [key, span] of this.spanMap) {\n if (\n span.name === 'handler.conversation' &&\n !this.seenHandlerKeys.has(key) &&\n (span.status === 'ok' || span.status === 'error')\n ) {\n this.seenHandlerKeys.add(key)\n return true\n }\n }\n return false\n }\n\n /**\n * Check if a handler.workflow span matches the given condition.\n */\n private _matchesWorkflowCondition(name: string, signal: 'entered' | 'completed'): boolean {\n for (const span of this.spanMap.values()) {\n if (span.name !== 'handler.workflow') continue\n\n const d = spanData(span)\n const wfName = d['workflow.name'] || d['workflowName']\n if (wfName !== name) continue\n\n if (signal === 'entered') {\n // Any status means the workflow was entered\n return true\n }\n\n if (signal === 'completed') {\n return span.status === 'ok' || span.status === 'error'\n }\n }\n return false\n }\n}\n"],
5
- "mappings": ";AA+BA,SAAS,QAAQ,MAAoB;AACnC,SAAO,GAAG,KAAK,GAAG,KAAK,IAAI,KAAK,GAAG,IAAI;AACzC;AAKA,SAAS,SAAS,MAAqC;AACrD,SAAO,KAAK,QAAQ,OAAO,KAAK,SAAS,WAAY,KAAK,OAAmC,CAAC;AAChG;AAEO,IAAM,eAAN,MAAM,cAAa;AAAA,EAChB;AAAA,EACA;AAAA;AAAA,EAGA,UAAU,oBAAI,IAAkB;AAAA;AAAA,EAGhC,gBAAgB,oBAAI,IAAY;AAAA;AAAA,EAGhC,kBAAkB,oBAAI,IAAY;AAAA;AAAA,EAGlC,UAAuC,CAAC;AAAA;AAAA,EAGxC,kBAA0C;AAAA;AAAA,EAG1C,YAAY;AAAA;AAAA,EAGZ,cAA4B;AAAA;AAAA,EAG5B,gBAAqC;AAAA,EAE7C,YAAY,cAAsB,UAAkC,CAAC,GAAG;AACtE,SAAK,eAAe;AACpB,SAAK,UAAU;AAAA,EACjB;AAAA;AAAA;AAAA;AAAA;AAAA,EAMA,MAAM,QAAQ,QAAkC;AAC9C,UAAM,gBAAgB,oBAAoB,SAAS,mBAAmB;AACtE,UAAM,iBAAiB,oBAAoB,SAAS,OAAO,iBAAiB,OAAO;AAEnF,UAAM,MAAM,GAAG,KAAK,YAAY,oCAAoC,mBAAmB,aAAa,CAAC,mBAAmB,mBAAmB,cAAc,CAAC;AAE1J,SAAK,kBAAkB,IAAI,gBAAgB;AAE3C,UAAM,WAAW,MAAM,MAAM,KAAK;AAAA,MAChC,SAAS,EAAE,GAAG,KAAK,SAAS,QAAQ,oBAAoB;AAAA,MACxD,QAAQ,KAAK,gBAAgB;AAAA,IAC/B,CAAC;AAED,QAAI,CAAC,SAAS,IAAI;AAChB,YAAM,IAAI,MAAM,0BAA0B,SAAS,MAAM,IAAI,SAAS,UAAU,EAAE;AAAA,IACpF;AAEA,QAAI,CAAC,SAAS,MAAM;AAClB,YAAM,IAAI,MAAM,0BAA0B;AAAA,IAC5C;AAEA,SAAK,YAAY;AAGjB,SAAK,eAAe,SAAS,IAAI,EAAE,MAAM,CAAC,QAAQ;AAChD,UAAI,KAAK,WAAW;AAClB,aAAK,cAAc,eAAe,QAAQ,MAAM,IAAI,MAAM,OAAO,GAAG,CAAC;AAAA,MACvE;AAAA,IACF,CAAC;AAKD,UAAM,KAAK,iBAAiB;AAAA,EAC9B;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAOA,YAAkB;AAEhB,SAAK,gBAAgB,IAAI,IAAI,KAAK,QAAQ,KAAK,CAAC;AAGhD,eAAW,CAAC,KAAK,IAAI,KAAK,KAAK,SAAS;AACtC,UAAI,KAAK,SAAS,wBAAwB;AACxC,aAAK,gBAAgB,IAAI,GAAG;AAAA,MAC9B;AAAA,IACF;AAAA,EACF;AAAA,EAEA,OAAwB,YAAY;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAWpC,MAAM,oBAAoB,MAAkC;AAE1D,QAAI,KAAK,wBAAwB,GAAG;AAClC,aAAO,KAAK,oBAAoB;AAAA,IAClC;AACA,QAAI,KAAK,aAAa,SAAS;AAC7B,YAAM,IAAI,MAAM,6BAA6B;AAAA,IAC/C;AAEA,WAAO,IAAI,QAAc,CAAC,SAAS,WAAW;AAC5C,YAAM,QAAQ,WAAW,MAAM;AAC7B,aAAK,cAAc,MAAM;AACzB,aAAK,aAAa,oBAAoB,SAAS,OAAO;AACtD,cAAM,SAAS,KAAK,cAChB,oBAAoB,KAAK,YAAY,OAAO,KAC5C;AACJ,eAAO,IAAI,MAAM,uCAAuC,KAAK,OAAO,aAAQ,MAAM,EAAE,CAAC;AAAA,MACvF,GAAG,KAAK,OAAO;AAEf,YAAM,UAAU,MAAM;AACpB,qBAAa,KAAK;AAClB,aAAK,cAAc,MAAM;AACzB,eAAO,IAAI,MAAM,6BAA6B,CAAC;AAAA,MACjD;AACA,WAAK,aAAa,iBAAiB,SAAS,SAAS,EAAE,MAAM,KAAK,CAAC;AAEnE,YAAM,SAAS,CAAC,UAAgB;AAC9B,YAAI,KAAK,wBAAwB,GAAG;AAClC,uBAAa,KAAK;AAClB,eAAK,cAAc,MAAM;AACzB,eAAK,aAAa,oBAAoB,SAAS,OAAO;AACtD,eAAK,oBAAoB,EAAE,KAAK,SAAS,MAAM;AAAA,QACjD;AAAA,MACF;AAEA,WAAK,QAAQ,KAAK,MAAM;AAAA,IAC1B,CAAC;AAAA,EACH;AAAA,EAEA,OAAwB,gBAAgB;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAShC,sBAAqC;AAC3C,WAAO,IAAI,QAAc,CAAC,YAAY;AACpC,UAAI,WAAW;AACf,YAAM,OAAO,MAAM;AACjB,YAAI,SAAU;AACd,mBAAW;AACX,qBAAa,OAAO;AACpB,qBAAa,WAAW;AACxB,aAAK,cAAc,YAAY;AAC/B,aAAK,gBAAgB;AACrB,gBAAQ;AAAA,MACV;AAEA,YAAM,UAAU,WAAW,MAAM,cAAa,aAAa;AAE3D,UAAI,cAAc,WAAW,MAAM,cAAa,SAAS;AAEzD,YAAM,eAAe,MAAM;AACzB,YAAI,SAAU;AACd,qBAAa,WAAW;AACxB,sBAAc,WAAW,MAAM,cAAa,SAAS;AAAA,MACvD;AAEA,WAAK,QAAQ,KAAK,YAAY;AAC9B,WAAK,gBAAgB;AAAA,IACvB,CAAC;AAAA,EACH;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAOA,MAAM,gBAAgB,MAAc,MAA0C;AAE5E,QAAI,KAAK,0BAA0B,MAAM,KAAK,MAAM,GAAG;AACrD;AAAA,IACF;AACA,QAAI,KAAK,aAAa,SAAS;AAC7B,YAAM,IAAI,MAAM,oBAAoB,IAAI,iBAAiB,KAAK,MAAM,cAAc;AAAA,IACpF;AAEA,WAAO,IAAI,QAAc,CAAC,SAAS,WAAW;AAC5C,YAAM,QAAQ,WAAW,MAAM;AAC7B,aAAK,cAAc,MAAM;AACzB,aAAK,aAAa,oBAAoB,SAAS,OAAO;AACtD,cAAM,SAAS,KAAK,cAAc,oBAAoB,KAAK,YAAY,OAAO,KAAK;AACnF;AAAA,UACE,IAAI;AAAA,YACF,oBAAoB,IAAI,iBAAiB,KAAK,MAAM,wBAAwB,KAAK,OAAO,aAAQ,MAAM;AAAA,UACxG;AAAA,QACF;AAAA,MACF,GAAG,KAAK,OAAO;AAEf,YAAM,UAAU,MAAM;AACpB,qBAAa,KAAK;AAClB,aAAK,cAAc,MAAM;AACzB,eAAO,IAAI,MAAM,oBAAoB,IAAI,iBAAiB,KAAK,MAAM,cAAc,CAAC;AAAA,MACtF;AACA,WAAK,aAAa,iBAAiB,SAAS,SAAS,EAAE,MAAM,KAAK,CAAC;AAEnE,YAAM,SAAS,CAAC,UAAgB;AAC9B,YAAI,KAAK,0BAA0B,MAAM,KAAK,MAAM,GAAG;AACrD,uBAAa,KAAK;AAClB,eAAK,cAAc,MAAM;AACzB,eAAK,aAAa,oBAAoB,SAAS,OAAO;AACtD,kBAAQ;AAAA,QACV;AAAA,MACF;AAEA,WAAK,QAAQ,KAAK,MAAM;AAAA,IAC1B,CAAC;AAAA,EACH;AAAA;AAAA;AAAA;AAAA,EAKA,eAAuB;AACrB,UAAM,SAAiB,CAAC;AACxB,eAAW,CAAC,KAAK,IAAI,KAAK,KAAK,SAAS;AACtC,UAAI,CAAC,KAAK,cAAc,IAAI,GAAG,GAAG;AAChC,eAAO,KAAK,IAAI;AAAA,MAClB;AAAA,IACF;AACA,WAAO;AAAA,EACT;AAAA;AAAA;AAAA;AAAA,EAKA,cAAsB;AACpB,WAAO,MAAM,KAAK,KAAK,QAAQ,OAAO,CAAC;AAAA,EACzC;AAAA;AAAA;AAAA;AAAA,EAKA,aAAmB;AACjB,SAAK,YAAY;AACjB,SAAK,iBAAiB,MAAM;AAC5B,SAAK,kBAAkB;AAGvB,SAAK,gBAAgB;AACrB,SAAK,gBAAgB;AAGrB,SAAK,UAAU,CAAC;AAChB,SAAK,oBAAoB,CAAC;AAAA,EAC5B;AAAA;AAAA;AAAA,EAKQ,mBAAmB;AAAA,EACnB,oBAAuC,CAAC;AAAA,EAEhD,OAAwB,sBAAsB;AAAA;AAAA;AAAA;AAAA;AAAA,EAMtC,mBAAkC;AACxC,QAAI,KAAK,kBAAkB;AACzB,aAAO,QAAQ,QAAQ;AAAA,IACzB;AACA,WAAO,IAAI,QAAc,CAAC,SAAS,WAAW;AAC5C,UAAI,UAAU;AACd,YAAM,WAAW,MAAM;AACrB,YAAI,QAAS;AACb,kBAAU;AACV,qBAAa,KAAK;AAClB,gBAAQ;AAAA,MACV;AACA,YAAM,QAAQ,WAAW,MAAM;AAC7B,YAAI,QAAS;AACb,kBAAU;AACV,cAAM,MAAM,KAAK,kBAAkB,QAAQ,QAAQ;AACnD,YAAI,QAAQ,GAAI,MAAK,kBAAkB,OAAO,KAAK,CAAC;AACpD;AAAA,UACE,IAAI;AAAA,YACF,oCAAoC,cAAa,mBAAmB;AAAA,UACtE;AAAA,QACF;AAAA,MACF,GAAG,cAAa,mBAAmB;AACnC,WAAK,kBAAkB,KAAK,QAAQ;AAAA,IACtC,CAAC;AAAA,EACH;AAAA;AAAA;AAAA;AAAA,EAKA,MAAc,eAAe,MAAiD;AAC5E,UAAM,UAAU,IAAI,YAAY;AAChC,QAAI,SAAS;AAEb,UAAM,SAAS,KAAK,UAAU;AAE9B,QAAI;AACF,aAAO,KAAK,WAAW;AACrB,cAAM,EAAE,MAAM,MAAM,IAAI,MAAM,OAAO,KAAK;AAC1C,YAAI,KAAM;AAEV,kBAAU,QAAQ,OAAO,OAAO,EAAE,QAAQ,KAAK,CAAC;AAGhD,cAAM,QAAQ,OAAO,MAAM,MAAM;AAEjC,iBAAS,MAAM,IAAI,KAAK;AAExB,mBAAW,QAAQ,OAAO;AACxB,eAAK,eAAe,IAAI;AAAA,QAC1B;AAAA,MACF;AAAA,IACF,UAAE;AACA,aAAO,YAAY;AAAA,IACrB;AAAA,EACF;AAAA;AAAA;AAAA;AAAA;AAAA,EAMQ,eAAe,OAAqB;AAC1C,UAAM,QAAQ,MAAM,MAAM,IAAI;AAC9B,QAAI,YAAY;AAChB,QAAI,WAAW;AAEf,eAAW,QAAQ,OAAO;AACxB,UAAI,KAAK,WAAW,QAAQ,GAAG;AAC7B,oBAAY,KAAK,MAAM,SAAS,MAAM,EAAE,KAAK;AAAA,MAC/C,WAAW,KAAK,WAAW,OAAO,GAAG;AACnC,cAAM,QAAQ,KAAK,MAAM,QAAQ,MAAM,EAAE,KAAK;AAC9C,mBAAW,WAAW,GAAG,QAAQ;AAAA,EAAK,KAAK,KAAK;AAAA,MAClD;AAAA,IAEF;AAEA,QAAI,CAAC,aAAa,CAAC,UAAU;AAE3B,UAAI,CAAC,KAAK,oBAAoB,MAAM,SAAS,QAAQ,GAAG;AAAA,MAExD;AACA;AAAA,IACF;AAEA,QAAI;AACF,YAAM,UAAU,KAAK,MAAM,QAAQ;AACnC,WAAK,aAAa,WAAW,OAAO;AAAA,IACtC,QAAQ;AAAA,IAER;AAAA,EACF;AAAA;AAAA;AAAA;AAAA,EAKQ,aAAa,WAAmB,SAAwB;AAC9D,QAAI,cAAc,YAAY;AAC5B,YAAM,OAAO;AACb,YAAM,QAAQ,MAAM,QAAQ,KAAK,KAAK,IAAI,KAAK,QAAQ,CAAC;AACxD,iBAAW,QAAQ,OAAO;AACxB,aAAK,YAAY,IAAI;AAAA,MACvB;AACA,WAAK,mBAAmB;AACxB,iBAAW,WAAW,KAAK,mBAAmB;AAC5C,gBAAQ;AAAA,MACV;AACA,WAAK,oBAAoB,CAAC;AAAA,IAC5B,WAAW,cAAc,UAAU;AACjC,YAAM,OAAO;AACb,UAAI,KAAK,MAAM;AACb,aAAK,YAAY,KAAK,IAAI;AAC1B,aAAK,eAAe,KAAK,IAAI;AAAA,MAC/B;AAAA,IACF;AAAA,EAEF;AAAA;AAAA;AAAA;AAAA,EAKQ,YAAY,MAAkB;AACpC,SAAK,QAAQ,IAAI,QAAQ,IAAI,GAAG,IAAI;AAAA,EACtC;AAAA;AAAA;AAAA;AAAA,EAKQ,eAAe,MAAkB;AAEvC,UAAM,UAAU,CAAC,GAAG,KAAK,OAAO;AAChC,eAAW,UAAU,SAAS;AAC5B,aAAO,IAAI;AAAA,IACb;AAAA,EACF;AAAA;AAAA;AAAA;AAAA,EAKQ,cAAc,IAAgC;AACpD,UAAM,MAAM,KAAK,QAAQ,QAAQ,EAAE;AACnC,QAAI,QAAQ,IAAI;AACd,WAAK,QAAQ,OAAO,KAAK,CAAC;AAAA,IAC5B;AAAA,EACF;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAYQ,0BAAmC;AACzC,eAAW,CAAC,KAAK,IAAI,KAAK,KAAK,SAAS;AACtC,UACE,KAAK,SAAS,0BACd,CAAC,KAAK,gBAAgB,IAAI,GAAG,MAC5B,KAAK,WAAW,QAAQ,KAAK,WAAW,UACzC;AACA,aAAK,gBAAgB,IAAI,GAAG;AAC5B,eAAO;AAAA,MACT;AAAA,IACF;AACA,WAAO;AAAA,EACT;AAAA;AAAA;AAAA;AAAA,EAKQ,0BAA0B,MAAc,QAA0C;AACxF,eAAW,QAAQ,KAAK,QAAQ,OAAO,GAAG;AACxC,UAAI,KAAK,SAAS,mBAAoB;AAEtC,YAAM,IAAI,SAAS,IAAI;AACvB,YAAM,SAAS,EAAE,eAAe,KAAK,EAAE,cAAc;AACrD,UAAI,WAAW,KAAM;AAErB,UAAI,WAAW,WAAW;AAExB,eAAO;AAAA,MACT;AAEA,UAAI,WAAW,aAAa;AAC1B,eAAO,KAAK,WAAW,QAAQ,KAAK,WAAW;AAAA,MACjD;AAAA,IACF;AACA,WAAO;AAAA,EACT;AACF;",
3
+ "sources": ["../../analytics/src/errors.ts", "../src/errors.ts", "../src/sse-collector.ts"],
4
+ "sourcesContent": ["/**\n * Shared base error for ADK packages. Lives in `@botpress/analytics` because\n * it's the dependency-light package every surface already shares (cli, adk,\n * ui) \u2014 so all typed errors can genuinely inherit one base instead of\n * mirroring its shape.\n *\n * The policy lives in docs/ERROR-HANDLING.md. The short version: every error\n * we *intend* to throw should carry a stable `code` (so PostHog error\n * tracking can slice issues by code) and an `expected` flag separating\n * user/environment conditions (bad input, not logged in, network down \u2014 render\n * a friendly message, low alert priority) from internal bugs (invariant\n * violations \u2014 full stack, page someone).\n *\n * The runtime package is out of scope and keeps its own `Errors.AbstractError`\n * hierarchy. The adk typed families (`DependencyError`, `Agent0*Error`) extend\n * this base; new code should start here.\n */\nexport class AdkError<Code extends string = string> extends Error {\n /**\n * Marker for cross-module-instance checks \u2014 `instanceof` breaks when a\n * package is duplicated in the dependency graph.\n * @internal\n */\n public static readonly __IS_ADK_BASE_ERROR = true\n\n /**\n * Stable SCREAMING_SNAKE_CASE identifier, e.g. 'BOT_NOT_FOUND'. Subclasses\n * pin `Code` to a literal union (`class EvalRunnerError extends\n * AdkError<EvalErrorCode>`) so typos are compile errors and `code` narrows\n * in switches.\n */\n readonly code: Code\n /**\n * True for user/environment conditions the developer can fix themselves\n * (validation, auth, connectivity). False for internal bugs. Defaults to\n * false so unclassified errors surface as bugs until someone triages them.\n */\n readonly expected: boolean\n /** Structured context. Sanitized before any telemetry leaves the machine. */\n readonly details?: Record<string, unknown>\n /** Actionable next step shown to the user, e.g. \"Run 'adk login' first.\" */\n readonly suggestion?: string\n\n constructor(opts: {\n code: Code\n message: string\n expected?: boolean\n details?: Record<string, unknown>\n suggestion?: string\n cause?: unknown\n }) {\n super(opts.message, opts.cause !== undefined ? { cause: opts.cause } : undefined)\n this.name = this.constructor.name\n this.code = opts.code\n this.expected = opts.expected ?? false\n // Conditional assignment keeps the optional fields absent-not-undefined,\n // which downstream tsconfigs with exactOptionalPropertyTypes require.\n if (opts.details !== undefined) {\n this.details = opts.details\n }\n if (opts.suggestion !== undefined) {\n this.suggestion = opts.suggestion\n }\n }\n}\n\n/**\n * Duck-typed guard that survives duplicated module instances in the\n * dependency graph, where `instanceof AdkError` returns false for an error\n * constructed from another copy of this module. No manual prototype walking\n * needed: static properties are inherited through the constructor chain, so\n * a subclass's constructor resolves the marker in one lookup.\n */\nexport function isAdkError(error: unknown): error is AdkError {\n if (error instanceof AdkError) return true\n if (!(error instanceof Error)) return false\n return (error.constructor as unknown as Record<string, unknown> | undefined)?.__IS_ADK_BASE_ERROR === true\n}\n", "import { AdkError } from '@botpress/analytics'\n\n/**\n * Stable error codes for the eval engine. Typed as a literal union so a typo'd\n * code is a compile error and PostHog issue-slicing stays consistent.\n * Conventions per docs/ERROR-HANDLING.md.\n */\nexport const EVAL_ERROR_CODES = [\n 'EVAL_LOAD_FAILED',\n 'EVAL_FILE_EMPTY',\n 'EVAL_DUPLICATE_NAME',\n 'EVAL_SEED_NO_CONVERSATION',\n 'EVAL_NO_CONVERSATION_ID',\n 'EVAL_TURN_CONFIG_INVALID',\n 'SSE_CONNECT_FAILED',\n 'SSE_NO_BODY',\n 'CHAT_CLIENT_MISSING',\n 'CHAT_NOT_CONNECTED',\n 'CHAT_INTEGRATION_MISSING',\n 'CHAT_CHANNEL_UNBOUND',\n] as const\n\nexport type EvalErrorCode = (typeof EVAL_ERROR_CODES)[number]\n\n/**\n * Typed error for the eval engine. Named EvalRunnerError (not EvalError) to\n * avoid shadowing the JS built-in `EvalError`.\n */\nexport class EvalRunnerError extends AdkError<EvalErrorCode> {}\n", "/**\n * SSECollector \u2014 connects to the ADK dev server's SSE trace stream,\n * accumulates spans, and provides promise-based wait methods for\n * turn completion and workflow signals.\n *\n * Uses fetch() + async iteration with manual SSE line parsing\n * (Bun does not have native EventSource).\n */\n\nimport type { Span } from './trace'\nimport { EvalRunnerError } from './errors'\n\ntype SSEFilter = { conversationId: string } | { workflowId: string }\n\ninterface WaitOptions {\n timeout: number\n /**\n * Optional abort signal \u2014 when fired, the wait rejects immediately rather\n * than burning the full timeout. Used by the runner so Stop unblocks\n * pending workflow assertions (otherwise a single negative-result wait\n * can hold the suite hostage for `idleTimeout` per assertion).\n */\n abortSignal?: AbortSignal\n}\n\ninterface WorkflowWaitOptions extends WaitOptions {\n signal: 'entered' | 'completed'\n}\n\n/**\n * Unique key for span deduplication.\n */\nfunction spanKey(span: Span): string {\n return `${span.id.trace}:${span.id.span}`\n}\n\n/**\n * Type-safe accessor for span.data fields.\n */\nfunction spanData(span: Span): Record<string, unknown> {\n return span.data && typeof span.data === 'object' ? (span.data as Record<string, unknown>) : {}\n}\n\nexport class SSECollector {\n private devServerUrl: string\n private headers: Record<string, string>\n\n /** All spans accumulated since connect(), keyed by spanKey for dedup. */\n private spanMap = new Map<string, Span>()\n\n /** Index into spanMap insertion order marking where the current turn started. */\n private turnStartKeys = new Set<string>()\n\n /** Handler span IDs seen before the current turn (for startTurn filtering). */\n private seenHandlerKeys = new Set<string>()\n\n /** Callbacks waiting for span updates. */\n private waiters: Array<(span: Span) => void> = []\n\n /** AbortController for the active SSE connection. */\n private abortController: AbortController | null = null\n\n /** Whether the background loop is running. */\n private connected = false\n\n /** Error from the SSE stream if it died unexpectedly. */\n private streamError: Error | null = null\n\n /** Set while a settle phase is in flight; called by disconnect() to short-circuit. */\n private _cancelSettle: (() => void) | null = null\n\n constructor(devServerUrl: string, headers: Record<string, string> = {}) {\n this.devServerUrl = devServerUrl\n this.headers = headers\n }\n\n /**\n * Open a streaming SSE connection filtered by conversationId or workflowId.\n * Starts a background loop that parses SSE events and accumulates spans.\n */\n async connect(filter: SSEFilter): Promise<void> {\n const attributeName = 'conversationId' in filter ? 'conversationId' : 'workflowId'\n const attributeValue = 'conversationId' in filter ? filter.conversationId : filter.workflowId\n\n const url = `${this.devServerUrl}/api/traces/stream?attributeName=${encodeURIComponent(attributeName)}&attributeValue=${encodeURIComponent(attributeValue)}`\n\n this.abortController = new AbortController()\n\n const response = await fetch(url, {\n headers: { ...this.headers, Accept: 'text/event-stream' },\n signal: this.abortController.signal,\n })\n\n if (!response.ok) {\n throw new EvalRunnerError({\n code: 'SSE_CONNECT_FAILED',\n message: `SSE connection failed: ${response.status} ${response.statusText}`,\n expected: true,\n suggestion: 'Make sure the dev server is running (`adk dev`) and reachable.',\n })\n }\n\n if (!response.body) {\n throw new EvalRunnerError({ code: 'SSE_NO_BODY', message: 'SSE response has no body' })\n }\n\n this.connected = true\n\n // Start background parsing loop\n this._consumeStream(response.body).catch((err) => {\n if (this.connected) {\n this.streamError = err instanceof Error ? err : new Error(String(err))\n }\n })\n\n // Wait for the initial snapshot to arrive before returning.\n // This ensures that connect() resolves only after the snapshot\n // has been processed so callers can immediately check spans.\n await this._waitForSnapshot()\n }\n\n /**\n * Mark the beginning of a new turn.\n * Records which handler.conversation span IDs have been seen so far,\n * so waitForTurnComplete() only resolves on NEW handler completions.\n */\n startTurn(): void {\n // Snapshot the current set of span keys as the turn boundary\n this.turnStartKeys = new Set(this.spanMap.keys())\n\n // Record all handler.conversation spans seen so far\n for (const [key, span] of this.spanMap) {\n if (span.name === 'handler.conversation') {\n this.seenHandlerKeys.add(key)\n }\n }\n }\n\n private static readonly SETTLE_MS = 100\n\n /**\n * Wait for a handler.conversation span to reach terminal status (ok or error).\n * Only considers spans NOT seen before the last startTurn() call.\n *\n * After the handler closes, waits for a quiet period (no new spans for SETTLE_MS)\n * to ensure all child spans (tool calls, messages, state saves) have arrived.\n * The runtime's HttpSpanExporter fires POSTs without awaiting them, so child\n * spans may arrive slightly after the handler close event.\n */\n async waitForTurnComplete(opts: WaitOptions): Promise<void> {\n // Check if a completed handler already exists (e.g. from snapshot)\n if (this._hasNewCompletedHandler()) {\n return this._settleAfterHandler()\n }\n if (opts.abortSignal?.aborted) {\n throw new Error('waitForTurnComplete aborted')\n }\n\n return new Promise<void>((resolve, reject) => {\n const timer = setTimeout(() => {\n this._removeWaiter(waiter)\n opts.abortSignal?.removeEventListener('abort', onAbort)\n const reason = this.streamError\n ? `SSE stream died: ${this.streamError.message}`\n : 'no handler.conversation span completed'\n reject(new Error(`waitForTurnComplete timed out after ${opts.timeout}ms \u2014 ${reason}`))\n }, opts.timeout)\n\n const onAbort = () => {\n clearTimeout(timer)\n this._removeWaiter(waiter)\n reject(new Error('waitForTurnComplete aborted'))\n }\n opts.abortSignal?.addEventListener('abort', onAbort, { once: true })\n\n const waiter = (_span: Span) => {\n if (this._hasNewCompletedHandler()) {\n clearTimeout(timer)\n this._removeWaiter(waiter)\n opts.abortSignal?.removeEventListener('abort', onAbort)\n this._settleAfterHandler().then(resolve, reject)\n }\n }\n\n this.waiters.push(waiter)\n })\n }\n\n private static readonly MAX_SETTLE_MS = 2_000\n\n /**\n * After the handler span closes, wait for spans to stop arriving.\n * Resolves when no new spans arrive for SETTLE_MS, or after MAX_SETTLE_MS hard cap.\n *\n * Exposes a cancellation hook via _cancelSettle so disconnect() can short-\n * circuit an in-flight settle phase and avoid leaking timers for up to 2s.\n */\n private _settleAfterHandler(): Promise<void> {\n return new Promise<void>((resolve) => {\n let resolved = false\n const done = () => {\n if (resolved) return\n resolved = true\n clearTimeout(hardCap)\n clearTimeout(settleTimer)\n this._removeWaiter(settleWaiter)\n this._cancelSettle = null\n resolve()\n }\n\n const hardCap = setTimeout(done, SSECollector.MAX_SETTLE_MS)\n\n let settleTimer = setTimeout(done, SSECollector.SETTLE_MS)\n\n const settleWaiter = () => {\n if (resolved) return\n clearTimeout(settleTimer)\n settleTimer = setTimeout(done, SSECollector.SETTLE_MS)\n }\n\n this.waiters.push(settleWaiter)\n this._cancelSettle = done\n })\n }\n\n /**\n * Wait for a handler.workflow span with matching workflow name.\n * - signal 'entered': resolves when the span appears (any status)\n * - signal 'completed': resolves when the span has status !== 'running'\n */\n async waitForWorkflow(name: string, opts: WorkflowWaitOptions): Promise<void> {\n // Check if condition is already met\n if (this._matchesWorkflowCondition(name, opts.signal)) {\n return\n }\n if (opts.abortSignal?.aborted) {\n throw new Error(`waitForWorkflow(\"${name}\", { signal: \"${opts.signal}\" }) aborted`)\n }\n\n return new Promise<void>((resolve, reject) => {\n const timer = setTimeout(() => {\n this._removeWaiter(waiter)\n opts.abortSignal?.removeEventListener('abort', onAbort)\n const reason = this.streamError ? `SSE stream died: ${this.streamError.message}` : 'span not found'\n reject(\n new Error(\n `waitForWorkflow(\"${name}\", { signal: \"${opts.signal}\" }) timed out after ${opts.timeout}ms \u2014 ${reason}`\n )\n )\n }, opts.timeout)\n\n const onAbort = () => {\n clearTimeout(timer)\n this._removeWaiter(waiter)\n reject(new Error(`waitForWorkflow(\"${name}\", { signal: \"${opts.signal}\" }) aborted`))\n }\n opts.abortSignal?.addEventListener('abort', onAbort, { once: true })\n\n const waiter = (_span: Span) => {\n if (this._matchesWorkflowCondition(name, opts.signal)) {\n clearTimeout(timer)\n this._removeWaiter(waiter)\n opts.abortSignal?.removeEventListener('abort', onAbort)\n resolve()\n }\n }\n\n this.waiters.push(waiter)\n })\n }\n\n /**\n * Return spans accumulated since the last startTurn() call.\n */\n getTurnSpans(): Span[] {\n const result: Span[] = []\n for (const [key, span] of this.spanMap) {\n if (!this.turnStartKeys.has(key)) {\n result.push(span)\n }\n }\n return result\n }\n\n /**\n * Return all spans accumulated since connect().\n */\n getAllSpans(): Span[] {\n return Array.from(this.spanMap.values())\n }\n\n /**\n * Close the SSE connection and clean up.\n */\n disconnect(): void {\n this.connected = false\n this.abortController?.abort()\n this.abortController = null\n // Short-circuit an in-flight settle phase so its hardCap / settleTimer\n // don't fire after disconnect.\n this._cancelSettle?.()\n this._cancelSettle = null\n // Waiters hold setTimeout references that will reject on their own;\n // clearing the array just prevents further notifications.\n this.waiters = []\n this.snapshotResolvers = []\n }\n\n // \u2500\u2500 Private \u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\n\n /** Whether we have received the initial snapshot. */\n private snapshotReceived = false\n private snapshotResolvers: Array<() => void> = []\n\n private static readonly SNAPSHOT_TIMEOUT_MS = 10_000\n\n /**\n * Returns a promise that resolves once the snapshot event has been processed,\n * or rejects after a timeout to prevent connect() from hanging indefinitely.\n */\n private _waitForSnapshot(): Promise<void> {\n if (this.snapshotReceived) {\n return Promise.resolve()\n }\n return new Promise<void>((resolve, reject) => {\n let settled = false\n const resolver = () => {\n if (settled) return\n settled = true\n clearTimeout(timer)\n resolve()\n }\n const timer = setTimeout(() => {\n if (settled) return\n settled = true\n const idx = this.snapshotResolvers.indexOf(resolver)\n if (idx !== -1) this.snapshotResolvers.splice(idx, 1)\n reject(\n new Error(\n `SSE snapshot not received within ${SSECollector.SNAPSHOT_TIMEOUT_MS}ms \u2014 is the dev server healthy?`\n )\n )\n }, SSECollector.SNAPSHOT_TIMEOUT_MS)\n this.snapshotResolvers.push(resolver)\n })\n }\n\n /**\n * Consume the ReadableStream body, parse SSE lines, and dispatch events.\n */\n private async _consumeStream(body: ReadableStream<Uint8Array>): Promise<void> {\n const decoder = new TextDecoder()\n let buffer = ''\n\n const reader = body.getReader()\n\n try {\n while (this.connected) {\n const { done, value } = await reader.read()\n if (done) break\n\n buffer += decoder.decode(value, { stream: true })\n\n // Split on double-newline boundaries to get complete SSE events\n const parts = buffer.split('\\n\\n')\n // The last part may be incomplete \u2014 keep it in the buffer\n buffer = parts.pop() ?? ''\n\n for (const part of parts) {\n this._parseSSEBlock(part)\n }\n }\n } finally {\n reader.releaseLock()\n }\n }\n\n /**\n * Parse a single SSE event block (text between \\n\\n boundaries).\n * Extracts event name and data payload.\n */\n private _parseSSEBlock(block: string): void {\n const lines = block.split('\\n')\n let eventName = ''\n let dataLine = ''\n\n for (const line of lines) {\n if (line.startsWith('event:')) {\n eventName = line.slice('event:'.length).trim()\n } else if (line.startsWith('data:')) {\n const value = line.slice('data:'.length).trim()\n dataLine = dataLine ? `${dataLine}\\n${value}` : value\n }\n // id: and retry: lines are ignored\n }\n\n if (!eventName || !dataLine) {\n // Could be a retry: line or empty block \u2014 skip\n if (!this.snapshotReceived && block.includes('retry:')) {\n // The retry line comes before the snapshot; don't resolve yet\n }\n return\n }\n\n try {\n const payload = JSON.parse(dataLine)\n this._handleEvent(eventName, payload)\n } catch {\n // Malformed JSON \u2014 skip\n }\n }\n\n /**\n * Dispatch a parsed SSE event to the appropriate handler.\n */\n private _handleEvent(eventName: string, payload: unknown): void {\n if (eventName === 'snapshot') {\n const data = payload as { spans: Span[] }\n const spans = Array.isArray(data.spans) ? data.spans : []\n for (const span of spans) {\n this._upsertSpan(span)\n }\n this.snapshotReceived = true\n for (const resolve of this.snapshotResolvers) {\n resolve()\n }\n this.snapshotResolvers = []\n } else if (eventName === 'update') {\n const data = payload as { span: Span }\n if (data.span) {\n this._upsertSpan(data.span)\n this._notifyWaiters(data.span)\n }\n }\n // keepalive events are intentionally ignored\n }\n\n /**\n * Insert or replace a span in the map (dedup by trace+span ID).\n */\n private _upsertSpan(span: Span): void {\n this.spanMap.set(spanKey(span), span)\n }\n\n /**\n * Notify all waiters that a new span was received.\n */\n private _notifyWaiters(span: Span): void {\n // Copy the array since waiters may remove themselves during iteration\n const current = [...this.waiters]\n for (const waiter of current) {\n waiter(span)\n }\n }\n\n /**\n * Remove a waiter callback from the list.\n */\n private _removeWaiter(fn: (span: Span) => void): void {\n const idx = this.waiters.indexOf(fn)\n if (idx !== -1) {\n this.waiters.splice(idx, 1)\n }\n }\n\n /**\n * Check if there is a handler.conversation span that:\n * 1. Was NOT seen before startTurn()\n * 2. Has terminal status (ok or error)\n *\n * Marks the matched key as consumed so the same handler span cannot\n * satisfy both the current turn's settle phase AND the next turn's\n * waitForTurnComplete check (which would otherwise resolve instantly\n * against stale data instead of waiting for the new turn's handler).\n */\n private _hasNewCompletedHandler(): boolean {\n for (const [key, span] of this.spanMap) {\n if (\n span.name === 'handler.conversation' &&\n !this.seenHandlerKeys.has(key) &&\n (span.status === 'ok' || span.status === 'error')\n ) {\n this.seenHandlerKeys.add(key)\n return true\n }\n }\n return false\n }\n\n /**\n * Check if a handler.workflow span matches the given condition.\n */\n private _matchesWorkflowCondition(name: string, signal: 'entered' | 'completed'): boolean {\n for (const span of this.spanMap.values()) {\n if (span.name !== 'handler.workflow') continue\n\n const d = spanData(span)\n const wfName = d['workflow.name'] || d['workflowName']\n if (wfName !== name) continue\n\n if (signal === 'entered') {\n // Any status means the workflow was entered\n return true\n }\n\n if (signal === 'completed') {\n return span.status === 'ok' || span.status === 'error'\n }\n }\n return false\n }\n}\n"],
5
+ "mappings": ";AAiBO,IAAM,WAAN,cAAqD,MAAM;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAMhE,OAAuB,sBAAsB;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAQpC;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAMA;AAAA;AAAA,EAEA;AAAA;AAAA,EAEA;AAAA,EAET,YAAY,MAOT;AACD,UAAM,KAAK,SAAS,KAAK,UAAU,SAAY,EAAE,OAAO,KAAK,MAAM,IAAI,MAAS;AAChF,SAAK,OAAO,KAAK,YAAY;AAC7B,SAAK,OAAO,KAAK;AACjB,SAAK,WAAW,KAAK,YAAY;AAGjC,QAAI,KAAK,YAAY,QAAW;AAC9B,WAAK,UAAU,KAAK;AAAA,IACtB;AACA,QAAI,KAAK,eAAe,QAAW;AACjC,WAAK,aAAa,KAAK;AAAA,IACzB;AAAA,EACF;AACF;;;ACpCO,IAAM,kBAAN,cAA8B,SAAwB;AAAC;;;ACI9D,SAAS,QAAQ,MAAoB;AACnC,SAAO,GAAG,KAAK,GAAG,KAAK,IAAI,KAAK,GAAG,IAAI;AACzC;AAKA,SAAS,SAAS,MAAqC;AACrD,SAAO,KAAK,QAAQ,OAAO,KAAK,SAAS,WAAY,KAAK,OAAmC,CAAC;AAChG;AAEO,IAAM,eAAN,MAAM,cAAa;AAAA,EAChB;AAAA,EACA;AAAA;AAAA,EAGA,UAAU,oBAAI,IAAkB;AAAA;AAAA,EAGhC,gBAAgB,oBAAI,IAAY;AAAA;AAAA,EAGhC,kBAAkB,oBAAI,IAAY;AAAA;AAAA,EAGlC,UAAuC,CAAC;AAAA;AAAA,EAGxC,kBAA0C;AAAA;AAAA,EAG1C,YAAY;AAAA;AAAA,EAGZ,cAA4B;AAAA;AAAA,EAG5B,gBAAqC;AAAA,EAE7C,YAAY,cAAsB,UAAkC,CAAC,GAAG;AACtE,SAAK,eAAe;AACpB,SAAK,UAAU;AAAA,EACjB;AAAA;AAAA;AAAA;AAAA;AAAA,EAMA,MAAM,QAAQ,QAAkC;AAC9C,UAAM,gBAAgB,oBAAoB,SAAS,mBAAmB;AACtE,UAAM,iBAAiB,oBAAoB,SAAS,OAAO,iBAAiB,OAAO;AAEnF,UAAM,MAAM,GAAG,KAAK,YAAY,oCAAoC,mBAAmB,aAAa,CAAC,mBAAmB,mBAAmB,cAAc,CAAC;AAE1J,SAAK,kBAAkB,IAAI,gBAAgB;AAE3C,UAAM,WAAW,MAAM,MAAM,KAAK;AAAA,MAChC,SAAS,EAAE,GAAG,KAAK,SAAS,QAAQ,oBAAoB;AAAA,MACxD,QAAQ,KAAK,gBAAgB;AAAA,IAC/B,CAAC;AAED,QAAI,CAAC,SAAS,IAAI;AAChB,YAAM,IAAI,gBAAgB;AAAA,QACxB,MAAM;AAAA,QACN,SAAS,0BAA0B,SAAS,MAAM,IAAI,SAAS,UAAU;AAAA,QACzE,UAAU;AAAA,QACV,YAAY;AAAA,MACd,CAAC;AAAA,IACH;AAEA,QAAI,CAAC,SAAS,MAAM;AAClB,YAAM,IAAI,gBAAgB,EAAE,MAAM,eAAe,SAAS,2BAA2B,CAAC;AAAA,IACxF;AAEA,SAAK,YAAY;AAGjB,SAAK,eAAe,SAAS,IAAI,EAAE,MAAM,CAAC,QAAQ;AAChD,UAAI,KAAK,WAAW;AAClB,aAAK,cAAc,eAAe,QAAQ,MAAM,IAAI,MAAM,OAAO,GAAG,CAAC;AAAA,MACvE;AAAA,IACF,CAAC;AAKD,UAAM,KAAK,iBAAiB;AAAA,EAC9B;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAOA,YAAkB;AAEhB,SAAK,gBAAgB,IAAI,IAAI,KAAK,QAAQ,KAAK,CAAC;AAGhD,eAAW,CAAC,KAAK,IAAI,KAAK,KAAK,SAAS;AACtC,UAAI,KAAK,SAAS,wBAAwB;AACxC,aAAK,gBAAgB,IAAI,GAAG;AAAA,MAC9B;AAAA,IACF;AAAA,EACF;AAAA,EAEA,OAAwB,YAAY;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAWpC,MAAM,oBAAoB,MAAkC;AAE1D,QAAI,KAAK,wBAAwB,GAAG;AAClC,aAAO,KAAK,oBAAoB;AAAA,IAClC;AACA,QAAI,KAAK,aAAa,SAAS;AAC7B,YAAM,IAAI,MAAM,6BAA6B;AAAA,IAC/C;AAEA,WAAO,IAAI,QAAc,CAAC,SAAS,WAAW;AAC5C,YAAM,QAAQ,WAAW,MAAM;AAC7B,aAAK,cAAc,MAAM;AACzB,aAAK,aAAa,oBAAoB,SAAS,OAAO;AACtD,cAAM,SAAS,KAAK,cAChB,oBAAoB,KAAK,YAAY,OAAO,KAC5C;AACJ,eAAO,IAAI,MAAM,uCAAuC,KAAK,OAAO,aAAQ,MAAM,EAAE,CAAC;AAAA,MACvF,GAAG,KAAK,OAAO;AAEf,YAAM,UAAU,MAAM;AACpB,qBAAa,KAAK;AAClB,aAAK,cAAc,MAAM;AACzB,eAAO,IAAI,MAAM,6BAA6B,CAAC;AAAA,MACjD;AACA,WAAK,aAAa,iBAAiB,SAAS,SAAS,EAAE,MAAM,KAAK,CAAC;AAEnE,YAAM,SAAS,CAAC,UAAgB;AAC9B,YAAI,KAAK,wBAAwB,GAAG;AAClC,uBAAa,KAAK;AAClB,eAAK,cAAc,MAAM;AACzB,eAAK,aAAa,oBAAoB,SAAS,OAAO;AACtD,eAAK,oBAAoB,EAAE,KAAK,SAAS,MAAM;AAAA,QACjD;AAAA,MACF;AAEA,WAAK,QAAQ,KAAK,MAAM;AAAA,IAC1B,CAAC;AAAA,EACH;AAAA,EAEA,OAAwB,gBAAgB;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAShC,sBAAqC;AAC3C,WAAO,IAAI,QAAc,CAAC,YAAY;AACpC,UAAI,WAAW;AACf,YAAM,OAAO,MAAM;AACjB,YAAI,SAAU;AACd,mBAAW;AACX,qBAAa,OAAO;AACpB,qBAAa,WAAW;AACxB,aAAK,cAAc,YAAY;AAC/B,aAAK,gBAAgB;AACrB,gBAAQ;AAAA,MACV;AAEA,YAAM,UAAU,WAAW,MAAM,cAAa,aAAa;AAE3D,UAAI,cAAc,WAAW,MAAM,cAAa,SAAS;AAEzD,YAAM,eAAe,MAAM;AACzB,YAAI,SAAU;AACd,qBAAa,WAAW;AACxB,sBAAc,WAAW,MAAM,cAAa,SAAS;AAAA,MACvD;AAEA,WAAK,QAAQ,KAAK,YAAY;AAC9B,WAAK,gBAAgB;AAAA,IACvB,CAAC;AAAA,EACH;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAOA,MAAM,gBAAgB,MAAc,MAA0C;AAE5E,QAAI,KAAK,0BAA0B,MAAM,KAAK,MAAM,GAAG;AACrD;AAAA,IACF;AACA,QAAI,KAAK,aAAa,SAAS;AAC7B,YAAM,IAAI,MAAM,oBAAoB,IAAI,iBAAiB,KAAK,MAAM,cAAc;AAAA,IACpF;AAEA,WAAO,IAAI,QAAc,CAAC,SAAS,WAAW;AAC5C,YAAM,QAAQ,WAAW,MAAM;AAC7B,aAAK,cAAc,MAAM;AACzB,aAAK,aAAa,oBAAoB,SAAS,OAAO;AACtD,cAAM,SAAS,KAAK,cAAc,oBAAoB,KAAK,YAAY,OAAO,KAAK;AACnF;AAAA,UACE,IAAI;AAAA,YACF,oBAAoB,IAAI,iBAAiB,KAAK,MAAM,wBAAwB,KAAK,OAAO,aAAQ,MAAM;AAAA,UACxG;AAAA,QACF;AAAA,MACF,GAAG,KAAK,OAAO;AAEf,YAAM,UAAU,MAAM;AACpB,qBAAa,KAAK;AAClB,aAAK,cAAc,MAAM;AACzB,eAAO,IAAI,MAAM,oBAAoB,IAAI,iBAAiB,KAAK,MAAM,cAAc,CAAC;AAAA,MACtF;AACA,WAAK,aAAa,iBAAiB,SAAS,SAAS,EAAE,MAAM,KAAK,CAAC;AAEnE,YAAM,SAAS,CAAC,UAAgB;AAC9B,YAAI,KAAK,0BAA0B,MAAM,KAAK,MAAM,GAAG;AACrD,uBAAa,KAAK;AAClB,eAAK,cAAc,MAAM;AACzB,eAAK,aAAa,oBAAoB,SAAS,OAAO;AACtD,kBAAQ;AAAA,QACV;AAAA,MACF;AAEA,WAAK,QAAQ,KAAK,MAAM;AAAA,IAC1B,CAAC;AAAA,EACH;AAAA;AAAA;AAAA;AAAA,EAKA,eAAuB;AACrB,UAAM,SAAiB,CAAC;AACxB,eAAW,CAAC,KAAK,IAAI,KAAK,KAAK,SAAS;AACtC,UAAI,CAAC,KAAK,cAAc,IAAI,GAAG,GAAG;AAChC,eAAO,KAAK,IAAI;AAAA,MAClB;AAAA,IACF;AACA,WAAO;AAAA,EACT;AAAA;AAAA;AAAA;AAAA,EAKA,cAAsB;AACpB,WAAO,MAAM,KAAK,KAAK,QAAQ,OAAO,CAAC;AAAA,EACzC;AAAA;AAAA;AAAA;AAAA,EAKA,aAAmB;AACjB,SAAK,YAAY;AACjB,SAAK,iBAAiB,MAAM;AAC5B,SAAK,kBAAkB;AAGvB,SAAK,gBAAgB;AACrB,SAAK,gBAAgB;AAGrB,SAAK,UAAU,CAAC;AAChB,SAAK,oBAAoB,CAAC;AAAA,EAC5B;AAAA;AAAA;AAAA,EAKQ,mBAAmB;AAAA,EACnB,oBAAuC,CAAC;AAAA,EAEhD,OAAwB,sBAAsB;AAAA;AAAA;AAAA;AAAA;AAAA,EAMtC,mBAAkC;AACxC,QAAI,KAAK,kBAAkB;AACzB,aAAO,QAAQ,QAAQ;AAAA,IACzB;AACA,WAAO,IAAI,QAAc,CAAC,SAAS,WAAW;AAC5C,UAAI,UAAU;AACd,YAAM,WAAW,MAAM;AACrB,YAAI,QAAS;AACb,kBAAU;AACV,qBAAa,KAAK;AAClB,gBAAQ;AAAA,MACV;AACA,YAAM,QAAQ,WAAW,MAAM;AAC7B,YAAI,QAAS;AACb,kBAAU;AACV,cAAM,MAAM,KAAK,kBAAkB,QAAQ,QAAQ;AACnD,YAAI,QAAQ,GAAI,MAAK,kBAAkB,OAAO,KAAK,CAAC;AACpD;AAAA,UACE,IAAI;AAAA,YACF,oCAAoC,cAAa,mBAAmB;AAAA,UACtE;AAAA,QACF;AAAA,MACF,GAAG,cAAa,mBAAmB;AACnC,WAAK,kBAAkB,KAAK,QAAQ;AAAA,IACtC,CAAC;AAAA,EACH;AAAA;AAAA;AAAA;AAAA,EAKA,MAAc,eAAe,MAAiD;AAC5E,UAAM,UAAU,IAAI,YAAY;AAChC,QAAI,SAAS;AAEb,UAAM,SAAS,KAAK,UAAU;AAE9B,QAAI;AACF,aAAO,KAAK,WAAW;AACrB,cAAM,EAAE,MAAM,MAAM,IAAI,MAAM,OAAO,KAAK;AAC1C,YAAI,KAAM;AAEV,kBAAU,QAAQ,OAAO,OAAO,EAAE,QAAQ,KAAK,CAAC;AAGhD,cAAM,QAAQ,OAAO,MAAM,MAAM;AAEjC,iBAAS,MAAM,IAAI,KAAK;AAExB,mBAAW,QAAQ,OAAO;AACxB,eAAK,eAAe,IAAI;AAAA,QAC1B;AAAA,MACF;AAAA,IACF,UAAE;AACA,aAAO,YAAY;AAAA,IACrB;AAAA,EACF;AAAA;AAAA;AAAA;AAAA;AAAA,EAMQ,eAAe,OAAqB;AAC1C,UAAM,QAAQ,MAAM,MAAM,IAAI;AAC9B,QAAI,YAAY;AAChB,QAAI,WAAW;AAEf,eAAW,QAAQ,OAAO;AACxB,UAAI,KAAK,WAAW,QAAQ,GAAG;AAC7B,oBAAY,KAAK,MAAM,SAAS,MAAM,EAAE,KAAK;AAAA,MAC/C,WAAW,KAAK,WAAW,OAAO,GAAG;AACnC,cAAM,QAAQ,KAAK,MAAM,QAAQ,MAAM,EAAE,KAAK;AAC9C,mBAAW,WAAW,GAAG,QAAQ;AAAA,EAAK,KAAK,KAAK;AAAA,MAClD;AAAA,IAEF;AAEA,QAAI,CAAC,aAAa,CAAC,UAAU;AAE3B,UAAI,CAAC,KAAK,oBAAoB,MAAM,SAAS,QAAQ,GAAG;AAAA,MAExD;AACA;AAAA,IACF;AAEA,QAAI;AACF,YAAM,UAAU,KAAK,MAAM,QAAQ;AACnC,WAAK,aAAa,WAAW,OAAO;AAAA,IACtC,QAAQ;AAAA,IAER;AAAA,EACF;AAAA;AAAA;AAAA;AAAA,EAKQ,aAAa,WAAmB,SAAwB;AAC9D,QAAI,cAAc,YAAY;AAC5B,YAAM,OAAO;AACb,YAAM,QAAQ,MAAM,QAAQ,KAAK,KAAK,IAAI,KAAK,QAAQ,CAAC;AACxD,iBAAW,QAAQ,OAAO;AACxB,aAAK,YAAY,IAAI;AAAA,MACvB;AACA,WAAK,mBAAmB;AACxB,iBAAW,WAAW,KAAK,mBAAmB;AAC5C,gBAAQ;AAAA,MACV;AACA,WAAK,oBAAoB,CAAC;AAAA,IAC5B,WAAW,cAAc,UAAU;AACjC,YAAM,OAAO;AACb,UAAI,KAAK,MAAM;AACb,aAAK,YAAY,KAAK,IAAI;AAC1B,aAAK,eAAe,KAAK,IAAI;AAAA,MAC/B;AAAA,IACF;AAAA,EAEF;AAAA;AAAA;AAAA;AAAA,EAKQ,YAAY,MAAkB;AACpC,SAAK,QAAQ,IAAI,QAAQ,IAAI,GAAG,IAAI;AAAA,EACtC;AAAA;AAAA;AAAA;AAAA,EAKQ,eAAe,MAAkB;AAEvC,UAAM,UAAU,CAAC,GAAG,KAAK,OAAO;AAChC,eAAW,UAAU,SAAS;AAC5B,aAAO,IAAI;AAAA,IACb;AAAA,EACF;AAAA;AAAA;AAAA;AAAA,EAKQ,cAAc,IAAgC;AACpD,UAAM,MAAM,KAAK,QAAQ,QAAQ,EAAE;AACnC,QAAI,QAAQ,IAAI;AACd,WAAK,QAAQ,OAAO,KAAK,CAAC;AAAA,IAC5B;AAAA,EACF;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA;AAAA,EAYQ,0BAAmC;AACzC,eAAW,CAAC,KAAK,IAAI,KAAK,KAAK,SAAS;AACtC,UACE,KAAK,SAAS,0BACd,CAAC,KAAK,gBAAgB,IAAI,GAAG,MAC5B,KAAK,WAAW,QAAQ,KAAK,WAAW,UACzC;AACA,aAAK,gBAAgB,IAAI,GAAG;AAC5B,eAAO;AAAA,MACT;AAAA,IACF;AACA,WAAO;AAAA,EACT;AAAA;AAAA;AAAA;AAAA,EAKQ,0BAA0B,MAAc,QAA0C;AACxF,eAAW,QAAQ,KAAK,QAAQ,OAAO,GAAG;AACxC,UAAI,KAAK,SAAS,mBAAoB;AAEtC,YAAM,IAAI,SAAS,IAAI;AACvB,YAAM,SAAS,EAAE,eAAe,KAAK,EAAE,cAAc;AACrD,UAAI,WAAW,KAAM;AAErB,UAAI,WAAW,WAAW;AAExB,eAAO;AAAA,MACT;AAEA,UAAI,WAAW,aAAa;AAC1B,eAAO,KAAK,WAAW,QAAQ,KAAK,WAAW;AAAA,MACjD;AAAA,IACF;AACA,WAAO;AAAA,EACT;AACF;",
6
6
  "names": []
7
7
  }
@@ -1 +1 @@
1
- {"version":3,"file":"transformer.d.ts","sourceRoot":"","sources":["../src/transformer.ts"],"names":[],"mappings":"AAAA;;;GAGG;AAEH,OAAO,KAAK,EAAE,IAAI,EAAE,MAAM,SAAS,CAAA;AACnC,OAAO,KAAK,EAA2B,QAAQ,EAAgB,MAAM,SAAS,CAAA;AA4B9E;;;;;;;;;GASG;AACH,wBAAgB,cAAc,CAAC,KAAK,EAAE,IAAI,EAAE,GAAG,QAAQ,CAQtD"}
1
+ {"version":3,"file":"transformer.d.ts","sourceRoot":"","sources":["../src/transformer.ts"],"names":[],"mappings":"AAAA;;;GAGG;AAEH,OAAO,KAAK,EAAE,IAAI,EAAE,MAAM,SAAS,CAAA;AACnC,OAAO,KAAK,EAA2B,QAAQ,EAAgB,MAAM,SAAS,CAAA;AA8B9E;;;;;;;;;GASG;AACH,wBAAgB,cAAc,CAAC,KAAK,EAAE,IAAI,EAAE,GAAG,QAAQ,CAQtD"}
@@ -1,7 +1,7 @@
1
1
  {
2
2
  "version": 3,
3
3
  "sources": ["../src/transformer.ts"],
4
- "sourcesContent": ["/**\n * Transformer: projects raw trace spans into grader-friendly TurnData.\n * Pure function \u2014 no side effects, no network calls.\n */\n\nimport type { Span } from './trace'\nimport type { StateMutation, ToolCall, TurnData, WorkflowSpan } from './types'\n\n/** Safe accessor for span.data as a string-keyed record. */\nfunction data(span: Span): Record<string, unknown> {\n return (span.data && typeof span.data === 'object' ? span.data : {}) as Record<string, unknown>\n}\n\n/** Attempt to parse a JSON string. Returns parsed result, raw value on parse failure, or fallback if value is nullish. */\nfunction tryParseJson(value: unknown): unknown {\n if (value == null) return undefined\n if (typeof value !== 'string') return value\n try {\n return JSON.parse(value)\n } catch {\n return value\n }\n}\n\n/** Parse JSON with a strict fallback \u2014 returns fallback on parse failure instead of the raw string. */\nfunction parseJsonOrFallback<T>(value: unknown, fallback: T): T {\n if (typeof value !== 'string') return fallback\n try {\n return JSON.parse(value) as T\n } catch {\n return fallback\n }\n}\n\n/**\n * Transform raw trace spans into structured TurnData for graders.\n *\n * Extraction logic:\n * - messages: from `chat.sendMessage` spans, ordered by startedAt\n * - toolCalls: from `autonomous.tool` spans with terminal status, deduped by span id, ordered by endedAt\n * - stateMutations: from `state.save` spans where changed_keys has length > 0\n * - workflowSpans: from `handler.workflow` spans\n * - handler: from `handler.conversation` or `handler.workflow` span with terminal status\n */\nexport function transformSpans(spans: Span[]): TurnData {\n return {\n messages: extractMessages(spans),\n toolCalls: extractToolCalls(spans),\n stateMutations: extractStateMutations(spans),\n workflowSpans: extractWorkflowSpans(spans),\n ...extractHandler(spans),\n }\n}\n\nfunction extractMessages(spans: Span[]): string[] {\n return spans\n .filter((s) => s.name === 'chat.sendMessage' && data(s)['message.preview'])\n .sort((a, b) => a.timing.startedAt - b.timing.startedAt)\n .map((s) => String(data(s)['message.preview']))\n}\n\nfunction extractToolCalls(spans: Span[]): ToolCall[] {\n const toolSpans = spans.filter(\n (s) => s.name === 'autonomous.tool' && s.status !== 'running' && data(s)['autonomous.tool.name']\n )\n\n // Deduplicate by span id\n const seen = new Set<string>()\n const unique = toolSpans.filter((s) => {\n if (seen.has(s.id.span)) return false\n seen.add(s.id.span)\n return true\n })\n\n return unique\n .sort((a, b) => (a.timing.endedAt ?? Infinity) - (b.timing.endedAt ?? Infinity))\n .map((s) => {\n const d = data(s)\n const input = parseJsonOrFallback<Record<string, unknown>>(d['autonomous.tool.input'], {})\n return {\n name: String(d['autonomous.tool.name']),\n input,\n output: String(d['autonomous.tool.output'] ?? ''),\n status: String(d['autonomous.tool.status'] ?? 'unknown'),\n }\n })\n}\n\nfunction extractStateMutations(spans: Span[]): StateMutation[] {\n return spans\n .filter((s) => {\n if (s.name !== 'state.save') return false\n const d = data(s)\n const changedKeys = d['state.changed_keys']\n return Array.isArray(changedKeys) && changedKeys.length > 0\n })\n .map((s) => {\n const d = data(s)\n return {\n type: String(d['type'] ?? 'unknown'),\n changedKeys: (d['state.changed_keys'] as string[]) ?? [],\n previous: tryParseJson(d['state.previous_value']),\n current: tryParseJson(d['state.value']),\n swappedToFile: d['swapped_to_file'] === true,\n }\n })\n}\n\nfunction extractWorkflowSpans(spans: Span[]): WorkflowSpan[] {\n return spans\n .filter((s) => s.name === 'handler.workflow')\n .map((s) => {\n const d = data(s)\n const result: WorkflowSpan = {\n name: String(d['workflow.name'] || d['workflowName'] || ''),\n status: s.status as WorkflowSpan['status'],\n }\n const statusFinal = d['workflow.status.final']\n if (statusFinal) {\n result.statusFinal = String(statusFinal)\n }\n return result\n })\n}\n\nfunction extractHandler(spans: Span[]): { handlerDuration: number; handlerStatus: 'ok' | 'error' } {\n const handler = spans.find(\n (s) => (s.name === 'handler.conversation' || s.name === 'handler.workflow') && s.status !== 'running'\n )\n\n if (!handler) {\n return { handlerDuration: 0, handlerStatus: 'ok' }\n }\n\n return {\n handlerDuration: handler.timing.duration ?? 0,\n handlerStatus: handler.status === 'error' ? 'error' : 'ok',\n }\n}\n"],
5
- "mappings": ";AASA,SAAS,KAAK,MAAqC;AACjD,SAAQ,KAAK,QAAQ,OAAO,KAAK,SAAS,WAAW,KAAK,OAAO,CAAC;AACpE;AAGA,SAAS,aAAa,OAAyB;AAC7C,MAAI,SAAS,KAAM,QAAO;AAC1B,MAAI,OAAO,UAAU,SAAU,QAAO;AACtC,MAAI;AACF,WAAO,KAAK,MAAM,KAAK;AAAA,EACzB,QAAQ;AACN,WAAO;AAAA,EACT;AACF;AAGA,SAAS,oBAAuB,OAAgB,UAAgB;AAC9D,MAAI,OAAO,UAAU,SAAU,QAAO;AACtC,MAAI;AACF,WAAO,KAAK,MAAM,KAAK;AAAA,EACzB,QAAQ;AACN,WAAO;AAAA,EACT;AACF;AAYO,SAAS,eAAe,OAAyB;AACtD,SAAO;AAAA,IACL,UAAU,gBAAgB,KAAK;AAAA,IAC/B,WAAW,iBAAiB,KAAK;AAAA,IACjC,gBAAgB,sBAAsB,KAAK;AAAA,IAC3C,eAAe,qBAAqB,KAAK;AAAA,IACzC,GAAG,eAAe,KAAK;AAAA,EACzB;AACF;AAEA,SAAS,gBAAgB,OAAyB;AAChD,SAAO,MACJ,OAAO,CAAC,MAAM,EAAE,SAAS,sBAAsB,KAAK,CAAC,EAAE,iBAAiB,CAAC,EACzE,KAAK,CAAC,GAAG,MAAM,EAAE,OAAO,YAAY,EAAE,OAAO,SAAS,EACtD,IAAI,CAAC,MAAM,OAAO,KAAK,CAAC,EAAE,iBAAiB,CAAC,CAAC;AAClD;AAEA,SAAS,iBAAiB,OAA2B;AACnD,QAAM,YAAY,MAAM;AAAA,IACtB,CAAC,MAAM,EAAE,SAAS,qBAAqB,EAAE,WAAW,aAAa,KAAK,CAAC,EAAE,sBAAsB;AAAA,EACjG;AAGA,QAAM,OAAO,oBAAI,IAAY;AAC7B,QAAM,SAAS,UAAU,OAAO,CAAC,MAAM;AACrC,QAAI,KAAK,IAAI,EAAE,GAAG,IAAI,EAAG,QAAO;AAChC,SAAK,IAAI,EAAE,GAAG,IAAI;AAClB,WAAO;AAAA,EACT,CAAC;AAED,SAAO,OACJ,KAAK,CAAC,GAAG,OAAO,EAAE,OAAO,WAAW,aAAa,EAAE,OAAO,WAAW,SAAS,EAC9E,IAAI,CAAC,MAAM;AACV,UAAM,IAAI,KAAK,CAAC;AAChB,UAAM,QAAQ,oBAA6C,EAAE,uBAAuB,GAAG,CAAC,CAAC;AACzF,WAAO;AAAA,MACL,MAAM,OAAO,EAAE,sBAAsB,CAAC;AAAA,MACtC;AAAA,MACA,QAAQ,OAAO,EAAE,wBAAwB,KAAK,EAAE;AAAA,MAChD,QAAQ,OAAO,EAAE,wBAAwB,KAAK,SAAS;AAAA,IACzD;AAAA,EACF,CAAC;AACL;AAEA,SAAS,sBAAsB,OAAgC;AAC7D,SAAO,MACJ,OAAO,CAAC,MAAM;AACb,QAAI,EAAE,SAAS,aAAc,QAAO;AACpC,UAAM,IAAI,KAAK,CAAC;AAChB,UAAM,cAAc,EAAE,oBAAoB;AAC1C,WAAO,MAAM,QAAQ,WAAW,KAAK,YAAY,SAAS;AAAA,EAC5D,CAAC,EACA,IAAI,CAAC,MAAM;AACV,UAAM,IAAI,KAAK,CAAC;AAChB,WAAO;AAAA,MACL,MAAM,OAAO,EAAE,MAAM,KAAK,SAAS;AAAA,MACnC,aAAc,EAAE,oBAAoB,KAAkB,CAAC;AAAA,MACvD,UAAU,aAAa,EAAE,sBAAsB,CAAC;AAAA,MAChD,SAAS,aAAa,EAAE,aAAa,CAAC;AAAA,MACtC,eAAe,EAAE,iBAAiB,MAAM;AAAA,IAC1C;AAAA,EACF,CAAC;AACL;AAEA,SAAS,qBAAqB,OAA+B;AAC3D,SAAO,MACJ,OAAO,CAAC,MAAM,EAAE,SAAS,kBAAkB,EAC3C,IAAI,CAAC,MAAM;AACV,UAAM,IAAI,KAAK,CAAC;AAChB,UAAM,SAAuB;AAAA,MAC3B,MAAM,OAAO,EAAE,eAAe,KAAK,EAAE,cAAc,KAAK,EAAE;AAAA,MAC1D,QAAQ,EAAE;AAAA,IACZ;AACA,UAAM,cAAc,EAAE,uBAAuB;AAC7C,QAAI,aAAa;AACf,aAAO,cAAc,OAAO,WAAW;AAAA,IACzC;AACA,WAAO;AAAA,EACT,CAAC;AACL;AAEA,SAAS,eAAe,OAA2E;AACjG,QAAM,UAAU,MAAM;AAAA,IACpB,CAAC,OAAO,EAAE,SAAS,0BAA0B,EAAE,SAAS,uBAAuB,EAAE,WAAW;AAAA,EAC9F;AAEA,MAAI,CAAC,SAAS;AACZ,WAAO,EAAE,iBAAiB,GAAG,eAAe,KAAK;AAAA,EACnD;AAEA,SAAO;AAAA,IACL,iBAAiB,QAAQ,OAAO,YAAY;AAAA,IAC5C,eAAe,QAAQ,WAAW,UAAU,UAAU;AAAA,EACxD;AACF;",
4
+ "sourcesContent": ["/**\n * Transformer: projects raw trace spans into grader-friendly TurnData.\n * Pure function \u2014 no side effects, no network calls.\n */\n\nimport type { Span } from './trace'\nimport type { StateMutation, ToolCall, TurnData, WorkflowSpan } from './types'\n\n/** Safe accessor for span.data as a string-keyed record. */\nfunction data(span: Span): Record<string, unknown> {\n return (span.data && typeof span.data === 'object' ? span.data : {}) as Record<string, unknown>\n}\n\n/** Attempt to parse a JSON string. Returns parsed result, raw value on parse failure, or fallback if value is nullish. */\nfunction tryParseJson(value: unknown): unknown {\n if (value == null) return undefined\n if (typeof value !== 'string') return value\n try {\n return JSON.parse(value)\n } catch {\n // Not JSON \u2014 fall back to the raw value, per this function's contract\n return value\n }\n}\n\n/** Parse JSON with a strict fallback \u2014 returns fallback on parse failure instead of the raw string. */\nfunction parseJsonOrFallback<T>(value: unknown, fallback: T): T {\n if (typeof value !== 'string') return fallback\n try {\n return JSON.parse(value) as T\n } catch {\n // Not JSON \u2014 return the caller's fallback, per this function's contract\n return fallback\n }\n}\n\n/**\n * Transform raw trace spans into structured TurnData for graders.\n *\n * Extraction logic:\n * - messages: from `chat.sendMessage` spans, ordered by startedAt\n * - toolCalls: from `autonomous.tool` spans with terminal status, deduped by span id, ordered by endedAt\n * - stateMutations: from `state.save` spans where changed_keys has length > 0\n * - workflowSpans: from `handler.workflow` spans\n * - handler: from `handler.conversation` or `handler.workflow` span with terminal status\n */\nexport function transformSpans(spans: Span[]): TurnData {\n return {\n messages: extractMessages(spans),\n toolCalls: extractToolCalls(spans),\n stateMutations: extractStateMutations(spans),\n workflowSpans: extractWorkflowSpans(spans),\n ...extractHandler(spans),\n }\n}\n\nfunction extractMessages(spans: Span[]): string[] {\n return spans\n .filter((s) => s.name === 'chat.sendMessage' && data(s)['message.preview'])\n .sort((a, b) => a.timing.startedAt - b.timing.startedAt)\n .map((s) => String(data(s)['message.preview']))\n}\n\nfunction extractToolCalls(spans: Span[]): ToolCall[] {\n const toolSpans = spans.filter(\n (s) => s.name === 'autonomous.tool' && s.status !== 'running' && data(s)['autonomous.tool.name']\n )\n\n // Deduplicate by span id\n const seen = new Set<string>()\n const unique = toolSpans.filter((s) => {\n if (seen.has(s.id.span)) return false\n seen.add(s.id.span)\n return true\n })\n\n return unique\n .sort((a, b) => (a.timing.endedAt ?? Infinity) - (b.timing.endedAt ?? Infinity))\n .map((s) => {\n const d = data(s)\n const input = parseJsonOrFallback<Record<string, unknown>>(d['autonomous.tool.input'], {})\n return {\n name: String(d['autonomous.tool.name']),\n input,\n output: String(d['autonomous.tool.output'] ?? ''),\n status: String(d['autonomous.tool.status'] ?? 'unknown'),\n }\n })\n}\n\nfunction extractStateMutations(spans: Span[]): StateMutation[] {\n return spans\n .filter((s) => {\n if (s.name !== 'state.save') return false\n const d = data(s)\n const changedKeys = d['state.changed_keys']\n return Array.isArray(changedKeys) && changedKeys.length > 0\n })\n .map((s) => {\n const d = data(s)\n return {\n type: String(d['type'] ?? 'unknown'),\n changedKeys: (d['state.changed_keys'] as string[]) ?? [],\n previous: tryParseJson(d['state.previous_value']),\n current: tryParseJson(d['state.value']),\n swappedToFile: d['swapped_to_file'] === true,\n }\n })\n}\n\nfunction extractWorkflowSpans(spans: Span[]): WorkflowSpan[] {\n return spans\n .filter((s) => s.name === 'handler.workflow')\n .map((s) => {\n const d = data(s)\n const result: WorkflowSpan = {\n name: String(d['workflow.name'] || d['workflowName'] || ''),\n status: s.status as WorkflowSpan['status'],\n }\n const statusFinal = d['workflow.status.final']\n if (statusFinal) {\n result.statusFinal = String(statusFinal)\n }\n return result\n })\n}\n\nfunction extractHandler(spans: Span[]): { handlerDuration: number; handlerStatus: 'ok' | 'error' } {\n const handler = spans.find(\n (s) => (s.name === 'handler.conversation' || s.name === 'handler.workflow') && s.status !== 'running'\n )\n\n if (!handler) {\n return { handlerDuration: 0, handlerStatus: 'ok' }\n }\n\n return {\n handlerDuration: handler.timing.duration ?? 0,\n handlerStatus: handler.status === 'error' ? 'error' : 'ok',\n }\n}\n"],
5
+ "mappings": ";AASA,SAAS,KAAK,MAAqC;AACjD,SAAQ,KAAK,QAAQ,OAAO,KAAK,SAAS,WAAW,KAAK,OAAO,CAAC;AACpE;AAGA,SAAS,aAAa,OAAyB;AAC7C,MAAI,SAAS,KAAM,QAAO;AAC1B,MAAI,OAAO,UAAU,SAAU,QAAO;AACtC,MAAI;AACF,WAAO,KAAK,MAAM,KAAK;AAAA,EACzB,QAAQ;AAEN,WAAO;AAAA,EACT;AACF;AAGA,SAAS,oBAAuB,OAAgB,UAAgB;AAC9D,MAAI,OAAO,UAAU,SAAU,QAAO;AACtC,MAAI;AACF,WAAO,KAAK,MAAM,KAAK;AAAA,EACzB,QAAQ;AAEN,WAAO;AAAA,EACT;AACF;AAYO,SAAS,eAAe,OAAyB;AACtD,SAAO;AAAA,IACL,UAAU,gBAAgB,KAAK;AAAA,IAC/B,WAAW,iBAAiB,KAAK;AAAA,IACjC,gBAAgB,sBAAsB,KAAK;AAAA,IAC3C,eAAe,qBAAqB,KAAK;AAAA,IACzC,GAAG,eAAe,KAAK;AAAA,EACzB;AACF;AAEA,SAAS,gBAAgB,OAAyB;AAChD,SAAO,MACJ,OAAO,CAAC,MAAM,EAAE,SAAS,sBAAsB,KAAK,CAAC,EAAE,iBAAiB,CAAC,EACzE,KAAK,CAAC,GAAG,MAAM,EAAE,OAAO,YAAY,EAAE,OAAO,SAAS,EACtD,IAAI,CAAC,MAAM,OAAO,KAAK,CAAC,EAAE,iBAAiB,CAAC,CAAC;AAClD;AAEA,SAAS,iBAAiB,OAA2B;AACnD,QAAM,YAAY,MAAM;AAAA,IACtB,CAAC,MAAM,EAAE,SAAS,qBAAqB,EAAE,WAAW,aAAa,KAAK,CAAC,EAAE,sBAAsB;AAAA,EACjG;AAGA,QAAM,OAAO,oBAAI,IAAY;AAC7B,QAAM,SAAS,UAAU,OAAO,CAAC,MAAM;AACrC,QAAI,KAAK,IAAI,EAAE,GAAG,IAAI,EAAG,QAAO;AAChC,SAAK,IAAI,EAAE,GAAG,IAAI;AAClB,WAAO;AAAA,EACT,CAAC;AAED,SAAO,OACJ,KAAK,CAAC,GAAG,OAAO,EAAE,OAAO,WAAW,aAAa,EAAE,OAAO,WAAW,SAAS,EAC9E,IAAI,CAAC,MAAM;AACV,UAAM,IAAI,KAAK,CAAC;AAChB,UAAM,QAAQ,oBAA6C,EAAE,uBAAuB,GAAG,CAAC,CAAC;AACzF,WAAO;AAAA,MACL,MAAM,OAAO,EAAE,sBAAsB,CAAC;AAAA,MACtC;AAAA,MACA,QAAQ,OAAO,EAAE,wBAAwB,KAAK,EAAE;AAAA,MAChD,QAAQ,OAAO,EAAE,wBAAwB,KAAK,SAAS;AAAA,IACzD;AAAA,EACF,CAAC;AACL;AAEA,SAAS,sBAAsB,OAAgC;AAC7D,SAAO,MACJ,OAAO,CAAC,MAAM;AACb,QAAI,EAAE,SAAS,aAAc,QAAO;AACpC,UAAM,IAAI,KAAK,CAAC;AAChB,UAAM,cAAc,EAAE,oBAAoB;AAC1C,WAAO,MAAM,QAAQ,WAAW,KAAK,YAAY,SAAS;AAAA,EAC5D,CAAC,EACA,IAAI,CAAC,MAAM;AACV,UAAM,IAAI,KAAK,CAAC;AAChB,WAAO;AAAA,MACL,MAAM,OAAO,EAAE,MAAM,KAAK,SAAS;AAAA,MACnC,aAAc,EAAE,oBAAoB,KAAkB,CAAC;AAAA,MACvD,UAAU,aAAa,EAAE,sBAAsB,CAAC;AAAA,MAChD,SAAS,aAAa,EAAE,aAAa,CAAC;AAAA,MACtC,eAAe,EAAE,iBAAiB,MAAM;AAAA,IAC1C;AAAA,EACF,CAAC;AACL;AAEA,SAAS,qBAAqB,OAA+B;AAC3D,SAAO,MACJ,OAAO,CAAC,MAAM,EAAE,SAAS,kBAAkB,EAC3C,IAAI,CAAC,MAAM;AACV,UAAM,IAAI,KAAK,CAAC;AAChB,UAAM,SAAuB;AAAA,MAC3B,MAAM,OAAO,EAAE,eAAe,KAAK,EAAE,cAAc,KAAK,EAAE;AAAA,MAC1D,QAAQ,EAAE;AAAA,IACZ;AACA,UAAM,cAAc,EAAE,uBAAuB;AAC7C,QAAI,aAAa;AACf,aAAO,cAAc,OAAO,WAAW;AAAA,IACzC;AACA,WAAO;AAAA,EACT,CAAC;AACL;AAEA,SAAS,eAAe,OAA2E;AACjG,QAAM,UAAU,MAAM;AAAA,IACpB,CAAC,OAAO,EAAE,SAAS,0BAA0B,EAAE,SAAS,uBAAuB,EAAE,WAAW;AAAA,EAC9F;AAEA,MAAI,CAAC,SAAS;AACZ,WAAO,EAAE,iBAAiB,GAAG,eAAe,KAAK;AAAA,EACnD;AAEA,SAAO;AAAA,IACL,iBAAiB,QAAQ,OAAO,YAAY;AAAA,IAC5C,eAAe,QAAQ,WAAW,UAAU,UAAU;AAAA,EACxD;AACF;",
6
6
  "names": []
7
7
  }
package/dist/types.d.ts CHANGED
@@ -56,6 +56,8 @@ export interface EvalReport {
56
56
  pass: boolean;
57
57
  duration: number;
58
58
  error?: string;
59
+ /** Stable EvalErrorCode when `error` came from a typed EvalRunnerError — lets consumers distinguish setup/config failures from bot failures. */
60
+ errorCode?: string;
59
61
  }
60
62
  export type EvalProgressEvent = {
61
63
  type: 'suite_start';
@@ -120,6 +122,14 @@ export interface EvalLogger {
120
122
  warn: (message: string) => void;
121
123
  error: (message: string) => void;
122
124
  }
125
+ /**
126
+ * Fallback when no logger is injected (programmatic use outside the CLI).
127
+ * console rather than a no-op so warnings stay visible — a silent default
128
+ * would hide judge failures and skipped pre-flights, the exact swallow
129
+ * anti-pattern docs/ERROR-HANDLING.md bans. The CLI always injects its own
130
+ * structured logger via EvalRunnerConfig.logger.
131
+ */
132
+ export declare const defaultLogger: EvalLogger;
123
133
  export interface EvalRunnerConfig {
124
134
  credentials: {
125
135
  token: string;
@@ -141,6 +151,13 @@ export interface EvalRunnerConfig {
141
151
  onProgress?: (event: EvalProgressEvent) => void;
142
152
  /** Logger for warnings and errors. Defaults to console. */
143
153
  logger?: EvalLogger;
154
+ /**
155
+ * Called for unexpected (internal-bug) eval failures that runEval folds
156
+ * into report.error — they never propagate to the CLI's command boundary,
157
+ * so without this hook they'd be invisible to error tracking. The CLI binds
158
+ * this to telemetry.captureException; evals itself stays telemetry-free.
159
+ */
160
+ onException?: (error: unknown, properties?: Record<string, unknown>) => void;
144
161
  /**
145
162
  * Optional abort signal — when fired, the suite stops starting new evals.
146
163
  * The currently running eval continues to completion (LLM calls in flight
@@ -1 +1 @@
1
- {"version":3,"file":"types.d.ts","sourceRoot":"","sources":["../src/types.ts"],"names":[],"mappings":"AAAA;;;;;GAKG;AAEH,OAAO,EACL,IAAI,EACJ,KAAK,cAAc,EACnB,KAAK,SAAS,EACd,KAAK,gBAAgB,EACrB,KAAK,cAAc,EACnB,KAAK,iBAAiB,EACtB,KAAK,iBAAiB,EACtB,KAAK,aAAa,EAClB,KAAK,cAAc,EACnB,KAAK,iBAAiB,EACtB,KAAK,eAAe,EACpB,KAAK,aAAa,EAClB,KAAK,eAAe,GACrB,MAAM,cAAc,CAAA;AAErB,YAAY,EAAE,IAAI,EAAE,UAAU,EAAE,MAAM,SAAS,CAAA;AAE/C,MAAM,WAAW,oBAAoB;IACnC,IAAI,EAAE;QAAE,EAAE,EAAE,MAAM,CAAA;KAAE,CAAA;IACpB,kBAAkB,EAAE,CAAC,KAAK,EAAE,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,KAAK,OAAO,CAAC;QAAE,YAAY,EAAE;YAAE,EAAE,EAAE,MAAM,CAAA;SAAE,CAAA;KAAE,CAAC,CAAA;IAGjG,aAAa,EAAE,CAAC,KAAK,EAAE,GAAG,KAAK,OAAO,CAAC,OAAO,CAAC,CAAA;IAC/C,WAAW,EAAE,CAAC,KAAK,EAAE,GAAG,KAAK,OAAO,CAAC,OAAO,CAAC,CAAA;CAC9C;AAED,MAAM,WAAW,UAAU;IACzB,OAAO,EAAE,CAAC,KAAK,EAAE;QAAE,SAAS,EAAE,MAAM,CAAA;KAAE,KAAK,OAAO,CAAC,oBAAoB,CAAC,CAAA;CACzE;AAID,MAAM,WAAW,YAAY;IAC3B,SAAS,EAAE,MAAM,CAAA;IACjB,IAAI,EAAE,OAAO,CAAA;IACb,QAAQ,EAAE,MAAM,CAAA;IAChB,MAAM,EAAE,MAAM,CAAA;CACf;AAID,MAAM,WAAW,QAAQ;IACvB,IAAI,EAAE,MAAM,CAAA;IACZ,KAAK,EAAE,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAA;IAC9B,MAAM,EAAE,MAAM,CAAA;IACd,MAAM,EAAE,MAAM,CAAA;CACf;AAID,MAAM,WAAW,UAAU;IACzB,UAAU,EAAE,MAAM,CAAA;IAClB,WAAW,EAAE,MAAM,CAAA;IACnB,WAAW,EAAE,MAAM,CAAA;IACnB,UAAU,EAAE,YAAY,EAAE,CAAA;IAC1B,IAAI,EAAE,OAAO,CAAA;IACb,qDAAqD;IACrD,WAAW,EAAE,MAAM,CAAA;IACnB,wDAAwD;IACxD,YAAY,EAAE,MAAM,CAAA;CACrB;AAID,MAAM,WAAW,UAAU;IACzB,IAAI,EAAE,MAAM,CAAA;IACZ,WAAW,CAAC,EAAE,MAAM,CAAA;IACpB,IAAI,CAAC,EAAE,YAAY,GAAG,YAAY,CAAA;IAClC,IAAI,CAAC,EAAE,MAAM,EAAE,CAAA;IACf,KAAK,EAAE,UAAU,EAAE,CAAA;IACnB,iBAAiB,EAAE,YAAY,EAAE,CAAA;IACjC,IAAI,EAAE,OAAO,CAAA;IACb,QAAQ,EAAE,MAAM,CAAA;IAChB,KAAK,CAAC,EAAE,MAAM,CAAA;CACf;AAID,MAAM,MAAM,iBAAiB,GACzB;IAAE,IAAI,EAAE,aAAa,CAAC;IAAC,UAAU,EAAE,MAAM,CAAA;CAAE,GAC3C;IAAE,IAAI,EAAE,YAAY,CAAC;IAAC,QAAQ,EAAE,MAAM,CAAC;IAAC,KAAK,EAAE,MAAM,CAAC;IAAC,UAAU,EAAE,MAAM,CAAA;CAAE,GAC3E;IACE,IAAI,EAAE,YAAY,CAAA;IAClB,QAAQ,EAAE,MAAM,CAAA;IAChB,SAAS,EAAE,MAAM,CAAA;IACjB,UAAU,EAAE,MAAM,CAAA;IAClB,UAAU,EAAE,MAAM,CAAA;IAClB,WAAW,EAAE,MAAM,CAAA;CACpB,GACD;IACE,IAAI,EAAE,eAAe,CAAA;IACrB,QAAQ,EAAE,MAAM,CAAA;IAChB,SAAS,EAAE,MAAM,CAAA;IACjB,UAAU,EAAE,MAAM,CAAA;IAClB,UAAU,EAAE,MAAM,CAAA;IAClB,UAAU,EAAE,UAAU,CAAA;CACvB,GACD;IAAE,IAAI,EAAE,eAAe,CAAC;IAAC,QAAQ,EAAE,MAAM,CAAC;IAAC,KAAK,EAAE,MAAM,CAAC;IAAC,MAAM,EAAE,UAAU,CAAA;CAAE,GAC9E;IAAE,IAAI,EAAE,gBAAgB,CAAC;IAAC,MAAM,EAAE,aAAa,CAAC;IAAC,KAAK,CAAC,EAAE,MAAM,CAAA;CAAE,CAAA;AAIrE,MAAM,WAAW,aAAa;IAC5B,EAAE,EAAE,MAAM,CAAA;IACV,SAAS,EAAE,MAAM,CAAA;IACjB,KAAK,EAAE,UAAU,EAAE,CAAA;IACnB,MAAM,EAAE,MAAM,CAAA;IACd,MAAM,EAAE,MAAM,CAAA;IACd,KAAK,EAAE,MAAM,CAAA;IACb,QAAQ,EAAE,MAAM,CAAA;IAChB,MAAM,CAAC,EAAE;QACP,KAAK,CAAC,EAAE,MAAM,EAAE,CAAA;QAChB,IAAI,CAAC,EAAE,MAAM,EAAE,CAAA;QACf,IAAI,CAAC,EAAE,YAAY,GAAG,YAAY,CAAA;KACnC,CAAA;IACD;;;;OAIG;IACH,OAAO,CAAC,EAAE,OAAO,CAAA;CAClB;AAID,MAAM,WAAW,aAAa;IAC5B,SAAS,EAAE,MAAM,CAAA;IACjB,KAAK,EAAE,MAAM,CAAA;IACb,KAAK,EAAE,MAAM,CAAA;IACb,MAAM,EAAE,MAAM,CAAA;CACf;AAID,MAAM,WAAW,UAAU;IACzB,IAAI,EAAE,CAAC,OAAO,EAAE,MAAM,KAAK,IAAI,CAAA;IAC/B,IAAI,EAAE,CAAC,OAAO,EAAE,MAAM,KAAK,IAAI,CAAA;IAC/B,KAAK,EAAE,CAAC,OAAO,EAAE,MAAM,KAAK,IAAI,CAAA;CACjC;AAID,MAAM,WAAW,gBAAgB;IAC/B,WAAW,EAAE;QACX,KAAK,EAAE,MAAM,CAAA;QACb,MAAM,EAAE,MAAM,CAAA;QACd,KAAK,EAAE,MAAM,CAAA;QACb,WAAW,CAAC,EAAE,MAAM,CAAA;QACpB,SAAS,CAAC,EAAE,MAAM,CAAA;KACnB,CAAA;IACD,SAAS,EAAE,MAAM,CAAA;IACjB,oGAAoG;IACpG,YAAY,CAAC,EAAE,MAAM,CAAA;IACrB,kGAAkG;IAClG,gBAAgB,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,MAAM,CAAC,CAAA;IACzC;;;OAGG;IACH,UAAU,CAAC,EAAE,UAAU,CAAA;IACvB,UAAU,CAAC,EAAE,CAAC,KAAK,EAAE,iBAAiB,KAAK,IAAI,CAAA;IAC/C,2DAA2D;IAC3D,MAAM,CAAC,EAAE,UAAU,CAAA;IACnB;;;;OAIG;IACH,MAAM,CAAC,EAAE,WAAW,CAAA;IACpB,4FAA4F;IAC5F,WAAW,CAAC,EAAE;QACZ,WAAW,CAAC,EAAE,MAAM,CAAA;QACpB,mIAAmI;QACnI,kBAAkB,CAAC,EAAE,MAAM,CAAA;QAC3B,wFAAwF;QACxF,UAAU,CAAC,EAAE,MAAM,CAAA;KACpB,CAAA;CACF;AAID,MAAM,WAAW,UAAU;IACzB,KAAK,CAAC,EAAE,MAAM,EAAE,CAAA;IAChB,IAAI,CAAC,EAAE,MAAM,EAAE,CAAA;IACf,IAAI,CAAC,EAAE,YAAY,GAAG,YAAY,CAAA;CACnC;AAID,MAAM,WAAW,aAAa;IAC5B,IAAI,EAAE,MAAM,CAAA;IACZ,WAAW,EAAE,MAAM,EAAE,CAAA;IACrB,QAAQ,EAAE,OAAO,CAAA;IACjB,OAAO,EAAE,OAAO,CAAA;IAChB,aAAa,EAAE,OAAO,CAAA;CACvB;AAED,MAAM,WAAW,YAAY;IAC3B,IAAI,EAAE,MAAM,CAAA;IACZ,MAAM,EAAE,SAAS,GAAG,IAAI,GAAG,OAAO,CAAA;IAClC,WAAW,CAAC,EAAE,MAAM,CAAA;CACrB;AAED,MAAM,WAAW,QAAQ;IACvB,QAAQ,EAAE,MAAM,EAAE,CAAA;IAClB,SAAS,EAAE,QAAQ,EAAE,CAAA;IACrB,cAAc,EAAE,aAAa,EAAE,CAAA;IAC/B,aAAa,EAAE,YAAY,EAAE,CAAA;IAC7B,eAAe,EAAE,MAAM,CAAA;IACvB,aAAa,EAAE,IAAI,GAAG,OAAO,CAAA;CAC9B"}
1
+ {"version":3,"file":"types.d.ts","sourceRoot":"","sources":["../src/types.ts"],"names":[],"mappings":"AAAA;;;;;GAKG;AAEH,OAAO,EACL,IAAI,EACJ,KAAK,cAAc,EACnB,KAAK,SAAS,EACd,KAAK,gBAAgB,EACrB,KAAK,cAAc,EACnB,KAAK,iBAAiB,EACtB,KAAK,iBAAiB,EACtB,KAAK,aAAa,EAClB,KAAK,cAAc,EACnB,KAAK,iBAAiB,EACtB,KAAK,eAAe,EACpB,KAAK,aAAa,EAClB,KAAK,eAAe,GACrB,MAAM,cAAc,CAAA;AAErB,YAAY,EAAE,IAAI,EAAE,UAAU,EAAE,MAAM,SAAS,CAAA;AAE/C,MAAM,WAAW,oBAAoB;IACnC,IAAI,EAAE;QAAE,EAAE,EAAE,MAAM,CAAA;KAAE,CAAA;IACpB,kBAAkB,EAAE,CAAC,KAAK,EAAE,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,KAAK,OAAO,CAAC;QAAE,YAAY,EAAE;YAAE,EAAE,EAAE,MAAM,CAAA;SAAE,CAAA;KAAE,CAAC,CAAA;IAGjG,aAAa,EAAE,CAAC,KAAK,EAAE,GAAG,KAAK,OAAO,CAAC,OAAO,CAAC,CAAA;IAC/C,WAAW,EAAE,CAAC,KAAK,EAAE,GAAG,KAAK,OAAO,CAAC,OAAO,CAAC,CAAA;CAC9C;AAED,MAAM,WAAW,UAAU;IACzB,OAAO,EAAE,CAAC,KAAK,EAAE;QAAE,SAAS,EAAE,MAAM,CAAA;KAAE,KAAK,OAAO,CAAC,oBAAoB,CAAC,CAAA;CACzE;AAID,MAAM,WAAW,YAAY;IAC3B,SAAS,EAAE,MAAM,CAAA;IACjB,IAAI,EAAE,OAAO,CAAA;IACb,QAAQ,EAAE,MAAM,CAAA;IAChB,MAAM,EAAE,MAAM,CAAA;CACf;AAID,MAAM,WAAW,QAAQ;IACvB,IAAI,EAAE,MAAM,CAAA;IACZ,KAAK,EAAE,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,CAAA;IAC9B,MAAM,EAAE,MAAM,CAAA;IACd,MAAM,EAAE,MAAM,CAAA;CACf;AAID,MAAM,WAAW,UAAU;IACzB,UAAU,EAAE,MAAM,CAAA;IAClB,WAAW,EAAE,MAAM,CAAA;IACnB,WAAW,EAAE,MAAM,CAAA;IACnB,UAAU,EAAE,YAAY,EAAE,CAAA;IAC1B,IAAI,EAAE,OAAO,CAAA;IACb,qDAAqD;IACrD,WAAW,EAAE,MAAM,CAAA;IACnB,wDAAwD;IACxD,YAAY,EAAE,MAAM,CAAA;CACrB;AAID,MAAM,WAAW,UAAU;IACzB,IAAI,EAAE,MAAM,CAAA;IACZ,WAAW,CAAC,EAAE,MAAM,CAAA;IACpB,IAAI,CAAC,EAAE,YAAY,GAAG,YAAY,CAAA;IAClC,IAAI,CAAC,EAAE,MAAM,EAAE,CAAA;IACf,KAAK,EAAE,UAAU,EAAE,CAAA;IACnB,iBAAiB,EAAE,YAAY,EAAE,CAAA;IACjC,IAAI,EAAE,OAAO,CAAA;IACb,QAAQ,EAAE,MAAM,CAAA;IAChB,KAAK,CAAC,EAAE,MAAM,CAAA;IACd,gJAAgJ;IAChJ,SAAS,CAAC,EAAE,MAAM,CAAA;CACnB;AAID,MAAM,MAAM,iBAAiB,GACzB;IAAE,IAAI,EAAE,aAAa,CAAC;IAAC,UAAU,EAAE,MAAM,CAAA;CAAE,GAC3C;IAAE,IAAI,EAAE,YAAY,CAAC;IAAC,QAAQ,EAAE,MAAM,CAAC;IAAC,KAAK,EAAE,MAAM,CAAC;IAAC,UAAU,EAAE,MAAM,CAAA;CAAE,GAC3E;IACE,IAAI,EAAE,YAAY,CAAA;IAClB,QAAQ,EAAE,MAAM,CAAA;IAChB,SAAS,EAAE,MAAM,CAAA;IACjB,UAAU,EAAE,MAAM,CAAA;IAClB,UAAU,EAAE,MAAM,CAAA;IAClB,WAAW,EAAE,MAAM,CAAA;CACpB,GACD;IACE,IAAI,EAAE,eAAe,CAAA;IACrB,QAAQ,EAAE,MAAM,CAAA;IAChB,SAAS,EAAE,MAAM,CAAA;IACjB,UAAU,EAAE,MAAM,CAAA;IAClB,UAAU,EAAE,MAAM,CAAA;IAClB,UAAU,EAAE,UAAU,CAAA;CACvB,GACD;IAAE,IAAI,EAAE,eAAe,CAAC;IAAC,QAAQ,EAAE,MAAM,CAAC;IAAC,KAAK,EAAE,MAAM,CAAC;IAAC,MAAM,EAAE,UAAU,CAAA;CAAE,GAC9E;IAAE,IAAI,EAAE,gBAAgB,CAAC;IAAC,MAAM,EAAE,aAAa,CAAC;IAAC,KAAK,CAAC,EAAE,MAAM,CAAA;CAAE,CAAA;AAIrE,MAAM,WAAW,aAAa;IAC5B,EAAE,EAAE,MAAM,CAAA;IACV,SAAS,EAAE,MAAM,CAAA;IACjB,KAAK,EAAE,UAAU,EAAE,CAAA;IACnB,MAAM,EAAE,MAAM,CAAA;IACd,MAAM,EAAE,MAAM,CAAA;IACd,KAAK,EAAE,MAAM,CAAA;IACb,QAAQ,EAAE,MAAM,CAAA;IAChB,MAAM,CAAC,EAAE;QACP,KAAK,CAAC,EAAE,MAAM,EAAE,CAAA;QAChB,IAAI,CAAC,EAAE,MAAM,EAAE,CAAA;QACf,IAAI,CAAC,EAAE,YAAY,GAAG,YAAY,CAAA;KACnC,CAAA;IACD;;;;OAIG;IACH,OAAO,CAAC,EAAE,OAAO,CAAA;CAClB;AAID,MAAM,WAAW,aAAa;IAC5B,SAAS,EAAE,MAAM,CAAA;IACjB,KAAK,EAAE,MAAM,CAAA;IACb,KAAK,EAAE,MAAM,CAAA;IACb,MAAM,EAAE,MAAM,CAAA;CACf;AAID,MAAM,WAAW,UAAU;IACzB,IAAI,EAAE,CAAC,OAAO,EAAE,MAAM,KAAK,IAAI,CAAA;IAC/B,IAAI,EAAE,CAAC,OAAO,EAAE,MAAM,KAAK,IAAI,CAAA;IAC/B,KAAK,EAAE,CAAC,OAAO,EAAE,MAAM,KAAK,IAAI,CAAA;CACjC;AAED;;;;;;GAMG;AACH,eAAO,MAAM,aAAa,EAAE,UAAoB,CAAA;AAIhD,MAAM,WAAW,gBAAgB;IAC/B,WAAW,EAAE;QACX,KAAK,EAAE,MAAM,CAAA;QACb,MAAM,EAAE,MAAM,CAAA;QACd,KAAK,EAAE,MAAM,CAAA;QACb,WAAW,CAAC,EAAE,MAAM,CAAA;QACpB,SAAS,CAAC,EAAE,MAAM,CAAA;KACnB,CAAA;IACD,SAAS,EAAE,MAAM,CAAA;IACjB,oGAAoG;IACpG,YAAY,CAAC,EAAE,MAAM,CAAA;IACrB,kGAAkG;IAClG,gBAAgB,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,MAAM,CAAC,CAAA;IACzC;;;OAGG;IACH,UAAU,CAAC,EAAE,UAAU,CAAA;IACvB,UAAU,CAAC,EAAE,CAAC,KAAK,EAAE,iBAAiB,KAAK,IAAI,CAAA;IAC/C,2DAA2D;IAC3D,MAAM,CAAC,EAAE,UAAU,CAAA;IACnB;;;;;OAKG;IACH,WAAW,CAAC,EAAE,CAAC,KAAK,EAAE,OAAO,EAAE,UAAU,CAAC,EAAE,MAAM,CAAC,MAAM,EAAE,OAAO,CAAC,KAAK,IAAI,CAAA;IAC5E;;;;OAIG;IACH,MAAM,CAAC,EAAE,WAAW,CAAA;IACpB,4FAA4F;IAC5F,WAAW,CAAC,EAAE;QACZ,WAAW,CAAC,EAAE,MAAM,CAAA;QACpB,mIAAmI;QACnI,kBAAkB,CAAC,EAAE,MAAM,CAAA;QAC3B,wFAAwF;QACxF,UAAU,CAAC,EAAE,MAAM,CAAA;KACpB,CAAA;CACF;AAID,MAAM,WAAW,UAAU;IACzB,KAAK,CAAC,EAAE,MAAM,EAAE,CAAA;IAChB,IAAI,CAAC,EAAE,MAAM,EAAE,CAAA;IACf,IAAI,CAAC,EAAE,YAAY,GAAG,YAAY,CAAA;CACnC;AAID,MAAM,WAAW,aAAa;IAC5B,IAAI,EAAE,MAAM,CAAA;IACZ,WAAW,EAAE,MAAM,EAAE,CAAA;IACrB,QAAQ,EAAE,OAAO,CAAA;IACjB,OAAO,EAAE,OAAO,CAAA;IAChB,aAAa,EAAE,OAAO,CAAA;CACvB;AAED,MAAM,WAAW,YAAY;IAC3B,IAAI,EAAE,MAAM,CAAA;IACZ,MAAM,EAAE,SAAS,GAAG,IAAI,GAAG,OAAO,CAAA;IAClC,WAAW,CAAC,EAAE,MAAM,CAAA;CACrB;AAED,MAAM,WAAW,QAAQ;IACvB,QAAQ,EAAE,MAAM,EAAE,CAAA;IAClB,SAAS,EAAE,QAAQ,EAAE,CAAA;IACrB,cAAc,EAAE,aAAa,EAAE,CAAA;IAC/B,aAAa,EAAE,YAAY,EAAE,CAAA;IAC7B,eAAe,EAAE,MAAM,CAAA;IACvB,aAAa,EAAE,IAAI,GAAG,OAAO,CAAA;CAC9B"}
package/dist/types.js CHANGED
@@ -19,7 +19,11 @@ var Eval = class {
19
19
  if (def.options !== void 0) this.options = def.options;
20
20
  }
21
21
  };
22
+
23
+ // src/types.ts
24
+ var defaultLogger = console;
22
25
  export {
23
- Eval
26
+ Eval,
27
+ defaultLogger
24
28
  };
25
29
  //# sourceMappingURL=types.js.map
package/dist/types.js.map CHANGED
@@ -1,7 +1,7 @@
1
1
  {
2
2
  "version": 3,
3
- "sources": ["../src/definition.ts"],
4
- "sourcesContent": ["export type MatchOperator =\n | string\n | { equals: unknown }\n | { contains: string }\n | { not_contains: string }\n | { matches: string }\n | { in: unknown[] }\n | { exists: boolean }\n | { gte: number }\n | { lte: number }\n\nexport type ResponseAssertion =\n | { contains: string }\n | { not_contains: string }\n | { matches: string }\n | { llm_judge: string }\n// | { similar_to: string } TODO: uncomment this when it gets implemented\n\nexport type ToolAssertion =\n | { called: string; params?: Record<string, MatchOperator> }\n | { not_called: string }\n | { call_order: string[] }\n\nexport interface StateAssertion {\n path: string\n equals?: unknown\n changed?: boolean\n}\n\nexport interface WorkflowAssertion {\n name: string\n entered?: boolean\n completed?: boolean\n}\n\nexport type NumericOperator = { equals: number } | { gte: number } | { lte: number }\n\nexport interface TimingAssertion {\n response_time: NumericOperator\n}\n\nexport interface TurnAssertions {\n response?: ResponseAssertion[]\n tools?: ToolAssertion[]\n state?: StateAssertion[]\n workflow?: WorkflowAssertion[]\n timing?: TimingAssertion[]\n}\n\nexport interface ConversationTurn {\n user?: string\n event?: { type: string; payload?: Record<string, unknown> }\n expectSilence?: boolean\n assert?: TurnAssertions\n}\n\nexport interface OutcomeAssertions {\n state?: StateAssertion[]\n workflow?: WorkflowAssertion[]\n}\n\nexport interface EvalSetup {\n state?: {\n bot?: Record<string, unknown>\n user?: Record<string, unknown>\n conversation?: Record<string, unknown>\n }\n workflow?: {\n trigger: string\n input?: Record<string, unknown>\n }\n}\n\nexport interface EvalDefinition {\n name: string\n description?: string\n tags?: string[]\n type?: 'capability' | 'regression'\n setup?: EvalSetup\n conversation: ConversationTurn[]\n outcome?: OutcomeAssertions\n options?: {\n idleTimeout?: number\n judgePassThreshold?: number\n }\n}\n\nexport class Eval implements EvalDefinition {\n readonly name: string\n readonly description?: string\n readonly tags?: string[]\n readonly type?: 'capability' | 'regression'\n readonly setup?: EvalSetup\n readonly conversation: ConversationTurn[]\n readonly outcome?: OutcomeAssertions\n readonly options?: {\n idleTimeout?: number\n judgePassThreshold?: number\n }\n\n constructor(def: EvalDefinition) {\n this.name = def.name\n this.conversation = def.conversation\n if (def.description !== undefined) this.description = def.description\n if (def.tags !== undefined) this.tags = def.tags\n if (def.type !== undefined) this.type = def.type\n if (def.setup !== undefined) this.setup = def.setup\n if (def.outcome !== undefined) this.outcome = def.outcome\n if (def.options !== undefined) this.options = def.options\n }\n}\n"],
5
- "mappings": ";AAuFO,IAAM,OAAN,MAAqC;AAAA,EACjC;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EAKT,YAAY,KAAqB;AAC/B,SAAK,OAAO,IAAI;AAChB,SAAK,eAAe,IAAI;AACxB,QAAI,IAAI,gBAAgB,OAAW,MAAK,cAAc,IAAI;AAC1D,QAAI,IAAI,SAAS,OAAW,MAAK,OAAO,IAAI;AAC5C,QAAI,IAAI,SAAS,OAAW,MAAK,OAAO,IAAI;AAC5C,QAAI,IAAI,UAAU,OAAW,MAAK,QAAQ,IAAI;AAC9C,QAAI,IAAI,YAAY,OAAW,MAAK,UAAU,IAAI;AAClD,QAAI,IAAI,YAAY,OAAW,MAAK,UAAU,IAAI;AAAA,EACpD;AACF;",
3
+ "sources": ["../src/definition.ts", "../src/types.ts"],
4
+ "sourcesContent": ["export type MatchOperator =\n | string\n | { equals: unknown }\n | { contains: string }\n | { not_contains: string }\n | { matches: string }\n | { in: unknown[] }\n | { exists: boolean }\n | { gte: number }\n | { lte: number }\n\nexport type ResponseAssertion =\n | { contains: string }\n | { not_contains: string }\n | { matches: string }\n | { llm_judge: string }\n// | { similar_to: string } TODO: uncomment this when it gets implemented\n\nexport type ToolAssertion =\n | { called: string; params?: Record<string, MatchOperator> }\n | { not_called: string }\n | { call_order: string[] }\n\nexport interface StateAssertion {\n path: string\n equals?: unknown\n changed?: boolean\n}\n\nexport interface WorkflowAssertion {\n name: string\n entered?: boolean\n completed?: boolean\n}\n\nexport type NumericOperator = { equals: number } | { gte: number } | { lte: number }\n\nexport interface TimingAssertion {\n response_time: NumericOperator\n}\n\nexport interface TurnAssertions {\n response?: ResponseAssertion[]\n tools?: ToolAssertion[]\n state?: StateAssertion[]\n workflow?: WorkflowAssertion[]\n timing?: TimingAssertion[]\n}\n\nexport interface ConversationTurn {\n user?: string\n event?: { type: string; payload?: Record<string, unknown> }\n expectSilence?: boolean\n assert?: TurnAssertions\n}\n\nexport interface OutcomeAssertions {\n state?: StateAssertion[]\n workflow?: WorkflowAssertion[]\n}\n\nexport interface EvalSetup {\n state?: {\n bot?: Record<string, unknown>\n user?: Record<string, unknown>\n conversation?: Record<string, unknown>\n }\n workflow?: {\n trigger: string\n input?: Record<string, unknown>\n }\n}\n\nexport interface EvalDefinition {\n name: string\n description?: string\n tags?: string[]\n type?: 'capability' | 'regression'\n setup?: EvalSetup\n conversation: ConversationTurn[]\n outcome?: OutcomeAssertions\n options?: {\n idleTimeout?: number\n judgePassThreshold?: number\n }\n}\n\nexport class Eval implements EvalDefinition {\n readonly name: string\n readonly description?: string\n readonly tags?: string[]\n readonly type?: 'capability' | 'regression'\n readonly setup?: EvalSetup\n readonly conversation: ConversationTurn[]\n readonly outcome?: OutcomeAssertions\n readonly options?: {\n idleTimeout?: number\n judgePassThreshold?: number\n }\n\n constructor(def: EvalDefinition) {\n this.name = def.name\n this.conversation = def.conversation\n if (def.description !== undefined) this.description = def.description\n if (def.tags !== undefined) this.tags = def.tags\n if (def.type !== undefined) this.type = def.type\n if (def.setup !== undefined) this.setup = def.setup\n if (def.outcome !== undefined) this.outcome = def.outcome\n if (def.options !== undefined) this.options = def.options\n }\n}\n", "/**\n * Eval type definitions.\n *\n * User-facing types live in @botpress/evals so eval authors can update eval\n * tooling without tying the authoring API to the agent runtime package.\n */\n\nexport {\n Eval,\n type EvalDefinition,\n type EvalSetup,\n type ConversationTurn,\n type TurnAssertions,\n type OutcomeAssertions,\n type ResponseAssertion,\n type ToolAssertion,\n type StateAssertion,\n type WorkflowAssertion,\n type TimingAssertion,\n type MatchOperator,\n type NumericOperator,\n} from './definition'\n\nexport type { Span, SpanStatus } from './trace'\n\nexport interface ChatClientConnection {\n user: { id: string }\n createConversation: (props: Record<string, unknown>) => Promise<{ conversation: { id: string } }>\n // Broad parameter types keep this compatible with @botpress/chat's narrower\n // generated payload unions while avoiding a dependency on those definitions.\n createMessage: (props: any) => Promise<unknown>\n createEvent: (props: any) => Promise<unknown>\n}\n\nexport interface ChatClient {\n connect: (props: { webhookId: string }) => Promise<ChatClientConnection>\n}\n\n// --- Grader result (universal currency for all graders) ---\n\nexport interface GraderResult {\n assertion: string\n pass: boolean\n expected: string\n actual: string\n}\n\n// --- Tool call extracted from traces ---\n\nexport interface ToolCall {\n name: string\n input: Record<string, unknown>\n output: string\n status: string\n}\n\n// --- Turn report ---\n\nexport interface TurnReport {\n turnNumber: number\n userMessage: string\n botResponse: string\n assertions: GraderResult[]\n pass: boolean\n /** Time spent waiting for the bot to respond (ms) */\n botDuration: number\n /** Time spent fetching traces + running graders (ms) */\n evalDuration: number\n}\n\n// --- Single eval report ---\n\nexport interface EvalReport {\n name: string\n description?: string\n type?: 'capability' | 'regression'\n tags?: string[]\n turns: TurnReport[]\n outcomeAssertions: GraderResult[]\n pass: boolean\n duration: number\n error?: string\n /** Stable EvalErrorCode when `error` came from a typed EvalRunnerError \u2014 lets consumers distinguish setup/config failures from bot failures. */\n errorCode?: string\n}\n\n// --- Progress event for real-time UI updates ---\n\nexport type EvalProgressEvent =\n | { type: 'suite_start'; totalEvals: number }\n | { type: 'eval_start'; evalName: string; index: number; totalTurns: number }\n | {\n type: 'turn_start'\n evalName: string\n evalIndex: number\n turnNumber: number\n totalTurns: number\n userMessage: string\n }\n | {\n type: 'turn_complete'\n evalName: string\n evalIndex: number\n turnNumber: number\n totalTurns: number\n turnReport: TurnReport\n }\n | { type: 'eval_complete'; evalName: string; index: number; report: EvalReport }\n | { type: 'suite_complete'; report: EvalRunReport; error?: string }\n\n// --- Eval run report (full suite) ---\n\nexport interface EvalRunReport {\n id: string\n timestamp: string\n evals: EvalReport[]\n passed: number\n failed: number\n total: number\n duration: number\n filter?: {\n names?: string[]\n tags?: string[]\n type?: 'capability' | 'regression'\n }\n /**\n * True when the suite was aborted mid-run (e.g. user clicked Stop). The\n * persisted report is partial \u2014 `evals` reflects only the completed evals,\n * and `total` matches that count, not the original suite size.\n */\n aborted?: boolean\n}\n\n// --- Bot connection info ---\n\nexport interface BotConnection {\n webhookId: string\n botId: string\n token: string\n apiUrl: string\n}\n\n// --- Logger interface ---\n\nexport interface EvalLogger {\n info: (message: string) => void\n warn: (message: string) => void\n error: (message: string) => void\n}\n\n/**\n * Fallback when no logger is injected (programmatic use outside the CLI).\n * console rather than a no-op so warnings stay visible \u2014 a silent default\n * would hide judge failures and skipped pre-flights, the exact swallow\n * anti-pattern docs/ERROR-HANDLING.md bans. The CLI always injects its own\n * structured logger via EvalRunnerConfig.logger.\n */\nexport const defaultLogger: EvalLogger = console\n\n// --- Eval runner configuration ---\n\nexport interface EvalRunnerConfig {\n credentials: {\n token: string\n apiUrl: string\n botId: string\n workspaceId?: string\n webhookId?: string\n }\n agentPath: string\n /** URL of the dev server providing traces and config. Defaults to the discovered DevConsole URL. */\n devServerUrl?: string\n /** Headers for dev-server requests, e.g. X-Agent-Path when routed through the singleton proxy. */\n devServerHeaders?: Record<string, string>\n /**\n * Optional host-provided chat client. The CLI passes its bundled chat client\n * here so the eval engine does not depend on @botpress/adk.\n */\n chatClient?: ChatClient\n onProgress?: (event: EvalProgressEvent) => void\n /** Logger for warnings and errors. Defaults to console. */\n logger?: EvalLogger\n /**\n * Called for unexpected (internal-bug) eval failures that runEval folds\n * into report.error \u2014 they never propagate to the CLI's command boundary,\n * so without this hook they'd be invisible to error tracking. The CLI binds\n * this to telemetry.captureException; evals itself stays telemetry-free.\n */\n onException?: (error: unknown, properties?: Record<string, unknown>) => void\n /**\n * Optional abort signal \u2014 when fired, the suite stops starting new evals.\n * The currently running eval continues to completion (LLM calls in flight\n * cannot always be cancelled), but no further evals are started.\n */\n signal?: AbortSignal\n /** Bot-level eval options from agent.config.ts. Cascades: eval > agent config > default. */\n evalOptions?: {\n idleTimeout?: number\n /** LLM judge pass threshold (integer, 1-5). Values outside this range are clamped. Cascades: eval > agent config > default (3). */\n judgePassThreshold?: number\n /** Model to use for llm_judge assertions (e.g. 'openai:gpt-4o'). Defaults to 'fast'. */\n judgeModel?: string\n }\n}\n\n// --- Eval suite filter ---\n\nexport interface EvalFilter {\n names?: string[]\n tags?: string[]\n type?: 'capability' | 'regression'\n}\n\n// --- Transformer types (trace-driven eval engine) ---\n\nexport interface StateMutation {\n type: string // 'bot', 'user', 'conversation'\n changedKeys: string[]\n previous: unknown\n current: unknown\n swappedToFile: boolean // data['swapped_to_file'] === true\n}\n\nexport interface WorkflowSpan {\n name: string\n status: 'running' | 'ok' | 'error'\n statusFinal?: string // 'completed', 'continue', 'failed'\n}\n\nexport interface TurnData {\n messages: string[]\n toolCalls: ToolCall[]\n stateMutations: StateMutation[]\n workflowSpans: WorkflowSpan[]\n handlerDuration: number\n handlerStatus: 'ok' | 'error'\n}\n"],
5
+ "mappings": ";AAuFO,IAAM,OAAN,MAAqC;AAAA,EACjC;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EAKT,YAAY,KAAqB;AAC/B,SAAK,OAAO,IAAI;AAChB,SAAK,eAAe,IAAI;AACxB,QAAI,IAAI,gBAAgB,OAAW,MAAK,cAAc,IAAI;AAC1D,QAAI,IAAI,SAAS,OAAW,MAAK,OAAO,IAAI;AAC5C,QAAI,IAAI,SAAS,OAAW,MAAK,OAAO,IAAI;AAC5C,QAAI,IAAI,UAAU,OAAW,MAAK,QAAQ,IAAI;AAC9C,QAAI,IAAI,YAAY,OAAW,MAAK,UAAU,IAAI;AAClD,QAAI,IAAI,YAAY,OAAW,MAAK,UAAU,IAAI;AAAA,EACpD;AACF;;;AC+CO,IAAM,gBAA4B;",
6
6
  "names": []
7
7
  }
package/package.json CHANGED
@@ -1,6 +1,6 @@
1
1
  {
2
2
  "name": "@botpress/evals",
3
- "version": "1.19.0-beta.6",
3
+ "version": "1.19.0-beta.9",
4
4
  "description": "Evaluation definitions and runner for ADK-based Botpress agents",
5
5
  "keywords": [
6
6
  "adk",