@tangle-network/agent-eval 0.115.3 → 0.117.0
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.
- package/CHANGELOG.md +55 -0
- package/dist/analyst/index.d.ts +16 -11
- package/dist/analyst/index.js +33 -25
- package/dist/analyst/index.js.map +1 -1
- package/dist/{analyze-runs-BYHg6Irm.d.ts → analyze-runs--2x39HZ7.d.ts} +3 -3
- package/dist/{baseline-DsNteOgR.d.ts → baseline-DKq3gJpP.d.ts} +6 -3
- package/dist/belief-state/index.d.ts +6 -6
- package/dist/belief-state/index.js +1 -1
- package/dist/benchmarks/index.d.ts +12 -5
- package/dist/benchmarks/index.js +11 -10
- package/dist/builder-eval/index.d.ts +4 -4
- package/dist/builder-eval/index.js +1 -1
- package/dist/{calibration-Dz8TQV4y.d.ts → calibration-C8MTS7cw.d.ts} +2 -2
- package/dist/campaign/index.d.ts +247 -34
- package/dist/campaign/index.js +33 -13
- package/dist/chunk-3YYRZDON.js +45 -0
- package/dist/chunk-3YYRZDON.js.map +1 -0
- package/dist/{chunk-RPDDVKI7.js → chunk-4JLWXDYA.js} +2 -2
- package/dist/{chunk-WSBUZMBU.js → chunk-CCZIVI3F.js} +54 -115
- package/dist/chunk-CCZIVI3F.js.map +1 -0
- package/dist/{chunk-J6P6PK2R.js → chunk-FQNLDL4D.js} +3 -3
- package/dist/{chunk-ONM6PEAE.js → chunk-GQCZRZ7L.js} +2 -2
- package/dist/chunk-HHWE3POT.js +94 -0
- package/dist/chunk-HHWE3POT.js.map +1 -0
- package/dist/{chunk-ADYLPOSX.js → chunk-HQPHZGL6.js} +1112 -135
- package/dist/chunk-HQPHZGL6.js.map +1 -0
- package/dist/{chunk-FAOEFFRT.js → chunk-IDZTTFRR.js} +390 -78
- package/dist/chunk-IDZTTFRR.js.map +1 -0
- package/dist/{chunk-3LXTCTWL.js → chunk-JSDVRFAP.js} +2 -2
- package/dist/{chunk-MHNQWM4I.js → chunk-LQUTGLOZ.js} +5 -1
- package/dist/chunk-LQUTGLOZ.js.map +1 -0
- package/dist/{chunk-4D5RVB3W.js → chunk-LTVG32KX.js} +30 -5
- package/dist/chunk-LTVG32KX.js.map +1 -0
- package/dist/{chunk-5S5NJ63F.js → chunk-MGEHEHSN.js} +807 -15
- package/dist/chunk-MGEHEHSN.js.map +1 -0
- package/dist/{chunk-GY4SYVPJ.js → chunk-NJC7U437.js} +97 -25
- package/dist/chunk-NJC7U437.js.map +1 -0
- package/dist/{chunk-NYFUT3B3.js → chunk-ODVOOEWQ.js} +31 -10
- package/dist/chunk-ODVOOEWQ.js.map +1 -0
- package/dist/{chunk-LNQEP766.js → chunk-S2F4J57L.js} +44 -4
- package/dist/chunk-S2F4J57L.js.map +1 -0
- package/dist/chunk-VCTY3W6J.js +798 -0
- package/dist/chunk-VCTY3W6J.js.map +1 -0
- package/dist/chunk-VF3XSYTI.js +545 -0
- package/dist/chunk-VF3XSYTI.js.map +1 -0
- package/dist/{chunk-TLDB7WRY.js → chunk-YZPO4UHR.js} +28 -31
- package/dist/chunk-YZPO4UHR.js.map +1 -0
- package/dist/{chunk-KG4TD7EQ.js → chunk-ZUXV7UWZ.js} +1425 -697
- package/dist/chunk-ZUXV7UWZ.js.map +1 -0
- package/dist/cli.js +4 -2
- package/dist/cli.js.map +1 -1
- package/dist/{code-agent-session-D-g04tcy.d.ts → code-agent-session-CjZsVd19.d.ts} +1 -1
- package/dist/contract/index.d.ts +45 -31
- package/dist/contract/index.js +58 -19
- package/dist/contract/index.js.map +1 -1
- package/dist/{control-CcBiAEnn.d.ts → control-6vuGfmDH.d.ts} +5 -5
- package/dist/control.d.ts +6 -6
- package/dist/cost-ledger-DWy3XdJc.d.ts +183 -0
- package/dist/{default-registry-DltpYR5u.d.ts → default-registry-DaK8b3fv.d.ts} +2 -1
- package/dist/{emitter-BRchAAAx.d.ts → emitter-CjD7vUwv.d.ts} +2 -2
- package/dist/{failure-cluster-C48PiReX.d.ts → failure-cluster-DOAcSJ87.d.ts} +2 -2
- package/dist/{feedback-trajectory-pDcz1lQ1.d.ts → feedback-trajectory-BUnM58xL.d.ts} +3 -3
- package/dist/fuzz.d.ts +8 -16
- package/dist/fuzz.js +72 -42
- package/dist/fuzz.js.map +1 -1
- package/dist/{gepa-dne9JDPL.d.ts → gepa-eESocoDi.d.ts} +64 -12
- package/dist/hosted/index.d.ts +14 -7
- package/dist/{index-BTEpx9He.d.ts → index-PdX4VnPA.d.ts} +3 -3
- package/dist/index.d.ts +97 -55
- package/dist/index.js +343 -244
- package/dist/index.js.map +1 -1
- package/dist/{insight-report-IwwvqZZv.d.ts → insight-report-DY4nDW9Q.d.ts} +1 -1
- package/dist/{integrity-qemeBAyx.d.ts → integrity-DqlBiLyK.d.ts} +1 -1
- package/dist/kind-factory-ClZmO25A.d.ts +171 -0
- package/dist/{llm-client-DyqEH4jH.d.ts → llm-client-qoDd18Qz.d.ts} +27 -3
- package/dist/meta-eval/index.d.ts +8 -7
- package/dist/meta-eval/index.js +1 -1
- package/dist/multishot/index.d.ts +10 -3
- package/dist/openapi.json +1 -1
- package/dist/pipelines/index.d.ts +16 -6
- package/dist/pipelines/index.js +119 -23
- package/dist/pipelines/index.js.map +1 -1
- package/dist/{kind-factory-DcNg13sZ.d.ts → policy-edit-wG9uFEFm.d.ts} +114 -167
- package/dist/{pre-registration-D8h7ZxNL.d.ts → pre-registration-BWQhJ3vz.d.ts} +23 -4
- package/dist/{provenance-Bibyg1U9.d.ts → provenance-DpjwyseI.d.ts} +28 -16
- package/dist/{query-Ck190MOd.d.ts → query-CF7PG61p.d.ts} +5 -3
- package/dist/{release-report-CCtzajxP.d.ts → release-report-C8G2i5Xi.d.ts} +2 -2
- package/dist/reporting.d.ts +10 -9
- package/dist/{researcher-Dq-EtpbE.d.ts → researcher-C8XyxQsu.d.ts} +7 -7
- package/dist/rl.d.ts +17 -12
- package/dist/rl.js +2 -2
- package/dist/{rubric-predictive-validity-DYTLjGWu.d.ts → rubric-predictive-validity-p49lLVrE.d.ts} +1 -1
- package/dist/{run-campaign-UADIM77S.js → run-campaign-IM26A6PD.js} +4 -2
- package/dist/{run-record-B7RTi_ix.d.ts → run-record-BDH49H2E.d.ts} +2 -2
- package/dist/{runtime-trajectory-Dws7Kpgi.d.ts → runtime-trajectory-DGBIUt4B.d.ts} +1 -1
- package/dist/{schema-SGWcK9wa.d.ts → schema-B3Q3l9Z_.d.ts} +2 -0
- package/dist/{semantic-concept-judge-DxJmRkyJ.d.ts → semantic-concept-judge-CXnPEJbf.d.ts} +23 -5
- package/dist/{statistics-oUbOJe-S.d.ts → statistics-KUnG73jH.d.ts} +1 -1
- package/dist/{storage-Dw_f7WMt.d.ts → storage-DrX3v_5B.d.ts} +12 -1
- package/dist/{store-BsVi7ncX.d.ts → store-DGqD0Pyo.d.ts} +1 -1
- package/dist/storyboard/index.d.ts +1 -1
- package/dist/{summary-report-BJ5aNwZ1.d.ts → summary-report-C5bKFfm-.d.ts} +2 -2
- package/dist/{test-graded-scenario-mzYBKspu.d.ts → test-graded-scenario-B0ybnPY7.d.ts} +3 -3
- package/dist/traces.d.ts +19 -10
- package/dist/traces.js +16 -4
- package/dist/{types-C5gJrOVT.d.ts → types-BSw1rOUB.d.ts} +97 -38
- package/dist/{types-C7DGg5ex.d.ts → types-BkfcQnxV.d.ts} +15 -0
- package/dist/wire/index.d.ts +28 -19
- package/dist/wire/index.js +4 -2
- package/docs/design/loop-taxonomy.md +1 -2
- package/docs/distributed-driver.md +1 -1
- package/package.json +3 -3
- package/dist/chunk-4D5RVB3W.js.map +0 -1
- package/dist/chunk-5S5NJ63F.js.map +0 -1
- package/dist/chunk-ADYLPOSX.js.map +0 -1
- package/dist/chunk-FAOEFFRT.js.map +0 -1
- package/dist/chunk-GY4SYVPJ.js.map +0 -1
- package/dist/chunk-I6LVHOV3.js +0 -205
- package/dist/chunk-I6LVHOV3.js.map +0 -1
- package/dist/chunk-KG4TD7EQ.js.map +0 -1
- package/dist/chunk-LNQEP766.js.map +0 -1
- package/dist/chunk-MHNQWM4I.js.map +0 -1
- package/dist/chunk-NYFUT3B3.js.map +0 -1
- package/dist/chunk-QMXXSNC4.js +0 -761
- package/dist/chunk-QMXXSNC4.js.map +0 -1
- package/dist/chunk-TLDB7WRY.js.map +0 -1
- package/dist/chunk-WSBUZMBU.js.map +0 -1
- package/dist/cost-ledger-DuSqlw5B.d.ts +0 -113
- package/dist/policy-edit-RLn8GWof.d.ts +0 -103
- /package/dist/{chunk-RPDDVKI7.js.map → chunk-4JLWXDYA.js.map} +0 -0
- /package/dist/{chunk-J6P6PK2R.js.map → chunk-FQNLDL4D.js.map} +0 -0
- /package/dist/{chunk-ONM6PEAE.js.map → chunk-GQCZRZ7L.js.map} +0 -0
- /package/dist/{chunk-3LXTCTWL.js.map → chunk-JSDVRFAP.js.map} +0 -0
- /package/dist/{run-campaign-UADIM77S.js.map → run-campaign-IM26A6PD.js.map} +0 -0
|
@@ -8,6 +8,41 @@ import {
|
|
|
8
8
|
} from "./chunk-ONWEPEDO.js";
|
|
9
9
|
|
|
10
10
|
// src/llm-client.ts
|
|
11
|
+
function maximumChargeForLlmRequest(request, options = {}) {
|
|
12
|
+
if (request.maxTokens === void 0) return void 0;
|
|
13
|
+
if (!Number.isInteger(request.maxTokens) || request.maxTokens <= 0) {
|
|
14
|
+
throw new RangeError(`maximumChargeForLlmRequest: maxTokens must be a positive integer`);
|
|
15
|
+
}
|
|
16
|
+
if (request.messages.some(
|
|
17
|
+
(message) => Array.isArray(message.content) && message.content.some((part) => part.type === "image_url")
|
|
18
|
+
)) {
|
|
19
|
+
return void 0;
|
|
20
|
+
}
|
|
21
|
+
const attempts = resolveMaximumAttempts(options.maxRetries);
|
|
22
|
+
const requestBytes = new TextEncoder().encode(
|
|
23
|
+
JSON.stringify(buildBody(request, false))
|
|
24
|
+
).byteLength;
|
|
25
|
+
const batches = request.jsonSchema ? 2 : 1;
|
|
26
|
+
return {
|
|
27
|
+
model: request.model,
|
|
28
|
+
inputTokens: requestBytes * attempts * batches,
|
|
29
|
+
outputTokens: request.maxTokens * attempts * batches
|
|
30
|
+
};
|
|
31
|
+
}
|
|
32
|
+
function costReceiptFromLlm(result) {
|
|
33
|
+
const cachedTokens = result.usage.cachedPromptTokens ?? 0;
|
|
34
|
+
return {
|
|
35
|
+
model: result.model,
|
|
36
|
+
inputTokens: Math.max(0, result.usage.promptTokens - cachedTokens),
|
|
37
|
+
outputTokens: result.usage.completionTokens,
|
|
38
|
+
cachedTokens: cachedTokens > 0 ? cachedTokens : void 0,
|
|
39
|
+
actualCostUsd: result.costUsd ?? void 0,
|
|
40
|
+
usageUnknown: result.usage.captured === false
|
|
41
|
+
};
|
|
42
|
+
}
|
|
43
|
+
function costReceiptFromLlmError(error) {
|
|
44
|
+
return error instanceof LlmResponseError ? costReceiptFromLlm(error.result) : void 0;
|
|
45
|
+
}
|
|
11
46
|
var LlmCallError = class extends AgentEvalError {
|
|
12
47
|
constructor(message, status, body, model) {
|
|
13
48
|
super("judge", message);
|
|
@@ -19,9 +54,26 @@ var LlmCallError = class extends AgentEvalError {
|
|
|
19
54
|
body;
|
|
20
55
|
model;
|
|
21
56
|
};
|
|
57
|
+
var LlmResponseError = class extends AgentEvalError {
|
|
58
|
+
constructor(message, result, options) {
|
|
59
|
+
super("judge", message, options);
|
|
60
|
+
this.result = result;
|
|
61
|
+
}
|
|
62
|
+
result;
|
|
63
|
+
};
|
|
22
64
|
var DEFAULT_BASE_URL = "https://router.tangle.tools/v1";
|
|
23
65
|
var DEFAULT_TIMEOUT_MS = Number(process.env.TANGLE_LLM_TIMEOUT_MS) || 3e5;
|
|
24
66
|
var DEFAULT_MAX_RETRIES = Number(process.env.TANGLE_LLM_MAX_RETRIES) || 3;
|
|
67
|
+
function resolveMaximumAttempts(configured) {
|
|
68
|
+
const attempts = configured ?? DEFAULT_MAX_RETRIES;
|
|
69
|
+
if (!Number.isInteger(attempts) || attempts <= 0) {
|
|
70
|
+
throw new RangeError("LLM maximum attempts must be a positive integer");
|
|
71
|
+
}
|
|
72
|
+
return attempts;
|
|
73
|
+
}
|
|
74
|
+
function providerTokenCount(value) {
|
|
75
|
+
return typeof value === "number" && Number.isSafeInteger(value) && value >= 0 ? value : void 0;
|
|
76
|
+
}
|
|
25
77
|
var RETRYABLE_STATUS = /* @__PURE__ */ new Set([429, 502, 503, 504]);
|
|
26
78
|
var TRANSIENT_ERROR_PATTERNS = [
|
|
27
79
|
/AbortError/i,
|
|
@@ -79,6 +131,7 @@ function buildHeaders(opts) {
|
|
|
79
131
|
} else if (opts.bearer || opts.apiKey) {
|
|
80
132
|
headers.Authorization = `Bearer ${opts.bearer ?? opts.apiKey}`;
|
|
81
133
|
}
|
|
134
|
+
if (opts.idempotencyKey) headers["Idempotency-Key"] = opts.idempotencyKey;
|
|
82
135
|
return headers;
|
|
83
136
|
}
|
|
84
137
|
function isSchemaRejection(status, body) {
|
|
@@ -188,7 +241,7 @@ async function callLlm(req, opts = {}) {
|
|
|
188
241
|
const url = `${baseUrl}/chat/completions`;
|
|
189
242
|
const endpoint = "/chat/completions";
|
|
190
243
|
const timeoutMs = req.timeoutMs ?? opts.defaultTimeoutMs ?? DEFAULT_TIMEOUT_MS;
|
|
191
|
-
const
|
|
244
|
+
const maximumAttempts = resolveMaximumAttempts(opts.maxRetries);
|
|
192
245
|
const fetchFn = opts.fetch ?? globalThis.fetch;
|
|
193
246
|
const headers = buildHeaders(opts);
|
|
194
247
|
const provider = opts.provider ?? providerFromBaseUrl(baseUrl);
|
|
@@ -199,7 +252,7 @@ async function callLlm(req, opts = {}) {
|
|
|
199
252
|
const deadlineMs = opts.deadlineMs;
|
|
200
253
|
const deadlineStart = Date.now();
|
|
201
254
|
let lastErr;
|
|
202
|
-
for (let attempt = 0; attempt <
|
|
255
|
+
for (let attempt = 0; attempt < maximumAttempts; attempt++) {
|
|
203
256
|
if (callerSignal?.aborted) {
|
|
204
257
|
throw new DOMException("callLlm aborted by caller signal", "AbortError");
|
|
205
258
|
}
|
|
@@ -267,7 +320,7 @@ async function callLlm(req, opts = {}) {
|
|
|
267
320
|
body,
|
|
268
321
|
req.model
|
|
269
322
|
);
|
|
270
|
-
if (RETRYABLE_STATUS.has(res.status) && attempt <
|
|
323
|
+
if (RETRYABLE_STATUS.has(res.status) && attempt < maximumAttempts - 1 && !deadlineExceeded(deadlineStart, deadlineMs)) {
|
|
271
324
|
lastErr = err;
|
|
272
325
|
const retryAfter = parseRetryAfter(res.headers);
|
|
273
326
|
await sleep(retryAfter ?? backoffMs(attempt));
|
|
@@ -323,7 +376,13 @@ async function callLlm(req, opts = {}) {
|
|
|
323
376
|
});
|
|
324
377
|
}
|
|
325
378
|
const choice = json.choices?.[0];
|
|
326
|
-
const usageRaw = json.usage
|
|
379
|
+
const usageRaw = json.usage && typeof json.usage === "object" && !Array.isArray(json.usage) ? json.usage : void 0;
|
|
380
|
+
const promptTokens = providerTokenCount(usageRaw?.prompt_tokens);
|
|
381
|
+
const completionTokens = providerTokenCount(usageRaw?.completion_tokens);
|
|
382
|
+
const totalTokens = providerTokenCount(usageRaw?.total_tokens);
|
|
383
|
+
const cachedRaw = usageRaw?.prompt_tokens_details && typeof usageRaw.prompt_tokens_details === "object" && !Array.isArray(usageRaw.prompt_tokens_details) ? usageRaw.prompt_tokens_details.cached_tokens : void 0;
|
|
384
|
+
const cachedPromptTokens = cachedRaw === void 0 ? void 0 : providerTokenCount(cachedRaw);
|
|
385
|
+
const usageCaptured = promptTokens !== void 0 && completionTokens !== void 0 && (cachedRaw === void 0 || cachedPromptTokens !== void 0 && cachedPromptTokens <= promptTokens) && (totalTokens === void 0 || totalTokens === promptTokens + completionTokens);
|
|
327
386
|
const costFromProxy = json._response_cost ?? json.cost_usd;
|
|
328
387
|
const content = choice?.message?.content ?? "";
|
|
329
388
|
return {
|
|
@@ -331,12 +390,11 @@ async function callLlm(req, opts = {}) {
|
|
|
331
390
|
finishReason: choice?.finish_reason ?? null,
|
|
332
391
|
contentEmpty: content.trim().length === 0,
|
|
333
392
|
usage: {
|
|
334
|
-
promptTokens:
|
|
335
|
-
completionTokens:
|
|
336
|
-
totalTokens:
|
|
337
|
-
|
|
338
|
-
|
|
339
|
-
) : void 0
|
|
393
|
+
promptTokens: promptTokens ?? 0,
|
|
394
|
+
completionTokens: completionTokens ?? 0,
|
|
395
|
+
totalTokens: totalTokens ?? (promptTokens ?? 0) + (completionTokens ?? 0),
|
|
396
|
+
captured: usageCaptured,
|
|
397
|
+
cachedPromptTokens
|
|
340
398
|
},
|
|
341
399
|
costUsd: typeof costFromProxy === "number" ? costFromProxy : null,
|
|
342
400
|
model: json.model ?? req.model,
|
|
@@ -383,7 +441,7 @@ async function callLlm(req, opts = {}) {
|
|
|
383
441
|
redactedFields: []
|
|
384
442
|
});
|
|
385
443
|
}
|
|
386
|
-
if (attempt <
|
|
444
|
+
if (attempt < maximumAttempts - 1 && isTransientLlmError(err) && !deadlineExceeded(deadlineStart, deadlineMs)) {
|
|
387
445
|
await sleep(backoffMs(attempt));
|
|
388
446
|
continue;
|
|
389
447
|
}
|
|
@@ -410,27 +468,34 @@ function cryptoEventId() {
|
|
|
410
468
|
return `${Date.now().toString(36)}-${Math.random().toString(36).slice(2, 10)}`;
|
|
411
469
|
}
|
|
412
470
|
async function callLlmJson(req, opts = {}) {
|
|
471
|
+
const result = await callLlmStructured(req, opts);
|
|
472
|
+
const value = parseJsonResult(result);
|
|
473
|
+
return { value, result };
|
|
474
|
+
}
|
|
475
|
+
async function callLlmStructured(req, opts = {}) {
|
|
413
476
|
try {
|
|
414
|
-
|
|
415
|
-
const value = parseJsonResult(result);
|
|
416
|
-
return { value, result };
|
|
477
|
+
return await callLlm({ ...req, jsonMode: req.jsonMode ?? !req.jsonSchema }, opts);
|
|
417
478
|
} catch (err) {
|
|
418
479
|
if (err instanceof LlmCallError && isSchemaRejection(err.status, err.body) && req.jsonSchema) {
|
|
419
480
|
const degradedReq = { ...req, jsonMode: true, jsonSchema: void 0 };
|
|
420
|
-
|
|
421
|
-
const value = parseJsonResult(result);
|
|
422
|
-
return { value, result };
|
|
481
|
+
return await callLlm(degradedReq, opts);
|
|
423
482
|
}
|
|
424
483
|
throw err;
|
|
425
484
|
}
|
|
426
485
|
}
|
|
427
486
|
function parseJsonResult(result) {
|
|
428
|
-
|
|
429
|
-
|
|
430
|
-
|
|
431
|
-
|
|
487
|
+
try {
|
|
488
|
+
if (result.finishReason === "length") {
|
|
489
|
+
throw new Error(
|
|
490
|
+
`LLM returned truncated JSON content (model=${result.model}, finishReason=length)`
|
|
491
|
+
);
|
|
492
|
+
}
|
|
493
|
+
return parseJsonSafely(result.content, result.model);
|
|
494
|
+
} catch (error) {
|
|
495
|
+
if (error instanceof LlmResponseError) throw error;
|
|
496
|
+
const cause = error instanceof Error ? error : new Error(String(error));
|
|
497
|
+
throw new LlmResponseError(cause.message, result, { cause });
|
|
432
498
|
}
|
|
433
|
-
return parseJsonSafely(result.content, result.model);
|
|
434
499
|
}
|
|
435
500
|
function parseJsonSafely(content, model) {
|
|
436
501
|
const stripped = extractJsonPayload(content);
|
|
@@ -527,12 +592,15 @@ async function probeLlm(model, opts = {}) {
|
|
|
527
592
|
}
|
|
528
593
|
}
|
|
529
594
|
var LlmClient = class {
|
|
595
|
+
maximumAttempts;
|
|
596
|
+
opts;
|
|
530
597
|
constructor(opts = {}) {
|
|
531
598
|
this.opts = opts;
|
|
599
|
+
this.maximumAttempts = resolveMaximumAttempts(opts.maxRetries);
|
|
532
600
|
}
|
|
533
|
-
opts;
|
|
534
601
|
call(req, per) {
|
|
535
|
-
|
|
602
|
+
const options = { ...this.opts, ...per };
|
|
603
|
+
return req.jsonSchema ? callLlmStructured(req, options) : callLlm(req, options);
|
|
536
604
|
}
|
|
537
605
|
callJson(req, per) {
|
|
538
606
|
return callLlmJson(req, { ...this.opts, ...per });
|
|
@@ -540,7 +608,11 @@ var LlmClient = class {
|
|
|
540
608
|
};
|
|
541
609
|
|
|
542
610
|
export {
|
|
611
|
+
maximumChargeForLlmRequest,
|
|
612
|
+
costReceiptFromLlm,
|
|
613
|
+
costReceiptFromLlmError,
|
|
543
614
|
LlmCallError,
|
|
615
|
+
LlmResponseError,
|
|
544
616
|
isTransientLlmError,
|
|
545
617
|
backoffMs,
|
|
546
618
|
stripFencedJson,
|
|
@@ -551,4 +623,4 @@ export {
|
|
|
551
623
|
probeLlm,
|
|
552
624
|
LlmClient
|
|
553
625
|
};
|
|
554
|
-
//# sourceMappingURL=chunk-
|
|
626
|
+
//# sourceMappingURL=chunk-NJC7U437.js.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"sources":["../src/llm-client.ts"],"sourcesContent":["/**\n * LLM client with graceful degrade.\n *\n * OpenAI-compatible `/v1/chat/completions` client with:\n * - Exponential-backoff retry on 429 + 5xx gateway errors (502/503/504).\n * - Retry on transient network errors (fetch failed, AbortError, ECONNRESET).\n * - Graceful json_schema → json_object degrade on 400 with schema-reject body.\n * - Fenced-JSON stripping (```json ... ```) for models that wrap structured output.\n * - Configurable base URL + api key / bearer, works with LiteLLM proxies, OpenAI\n * directly, cli-bridge subscriptions, and any router that speaks the spec.\n *\n * Usage:\n * const { value, result } = await callLlmJson<MyType>(\n * { model: 'gpt-4o', messages: [...], jsonSchema: { name: 'x', schema: {...} } },\n * { baseUrl: 'https://router.tangle.tools/v1', apiKey: process.env.KEY },\n * )\n *\n * This is THE llm-calling seam for agent-eval primitives that need structured\n * output (semantic concept judge, reviewer directives, critic scores). Primitives\n * that need free-form text use `callLlm` and parse output themselves.\n */\n\nimport type { CostReceiptInput, MaximumCharge } from './cost-ledger'\nimport { AgentEvalError, CaptureIntegrityError } from './errors'\nimport {\n defaultProviderRedactor,\n type ProviderRedactor,\n providerFromBaseUrl,\n type RawProviderEvent,\n type RawProviderSink,\n} from './trace/raw-provider-sink'\n\n// ─── Types ──────────────────────────────────────────────────────────────\n\nexport interface LlmMessage {\n role: 'system' | 'user' | 'assistant'\n /**\n * Either a plain text content string OR a multimodal content array\n * (text + image_url parts) for vision-capable models.\n */\n content:\n | string\n | Array<\n | { type: 'text'; text: string }\n | { type: 'image_url'; image_url: { url: string; detail?: 'auto' | 'low' | 'high' } }\n >\n}\n\nexport interface LlmCallRequest {\n model: string\n messages: LlmMessage[]\n /** Optional JSON-mode response format (response_format: json_object). */\n jsonMode?: boolean\n /** Optional structured output via JSON Schema. Falls back to json_object on 400. */\n jsonSchema?: { name: string; schema: Record<string, unknown> }\n temperature?: number\n maxTokens?: number\n /** Per-call timeout, default 300s. */\n timeoutMs?: number\n}\n\n/** Conservative priced bound for the exact text request sent to a provider.\n * Returns undefined when output or multimodal input is not bounded, causing a\n * capped CostLedger to reject the call before execution. */\nexport function maximumChargeForLlmRequest(\n request: Pick<LlmCallRequest, 'model' | 'messages' | 'jsonSchema' | 'maxTokens'>,\n options: LlmClientOptions = {},\n): MaximumCharge | undefined {\n if (request.maxTokens === undefined) return undefined\n if (!Number.isInteger(request.maxTokens) || request.maxTokens <= 0) {\n throw new RangeError(`maximumChargeForLlmRequest: maxTokens must be a positive integer`)\n }\n if (\n request.messages.some(\n (message) =>\n Array.isArray(message.content) && message.content.some((part) => part.type === 'image_url'),\n )\n ) {\n return undefined\n }\n\n const attempts = resolveMaximumAttempts(options.maxRetries)\n // A byte-level tokenizer cannot emit more input tokens than request bytes.\n // Pricing the complete body also covers role/schema framing omitted from content-only estimates.\n const requestBytes = new TextEncoder().encode(\n JSON.stringify(buildBody(request, false)),\n ).byteLength\n // A rejected response schema can trigger one JSON-mode batch with the same output limit.\n const batches = request.jsonSchema ? 2 : 1\n return {\n model: request.model,\n inputTokens: requestBytes * attempts * batches,\n outputTokens: request.maxTokens * attempts * batches,\n }\n}\n\nexport interface LlmUsage {\n promptTokens: number\n completionTokens: number\n totalTokens: number\n /** False when the provider omitted or malformed prompt/completion usage. */\n captured?: boolean\n /** Proxies populate this when prompt caching is on. */\n cachedPromptTokens?: number\n}\n\nexport interface LlmCallResult {\n /** The text content of the first choice. Empty string if none. */\n content: string\n usage: LlmUsage\n /**\n * Cost in USD. Pulled from proxy's `_response_cost` field when present;\n * `null` when neither the proxy nor the caller can derive it.\n */\n costUsd: number | null\n /** Model name actually used (echoed from response). */\n model: string\n /** Wall-clock duration of the HTTP call (last attempt, if retried). */\n durationMs: number\n /**\n * `finish_reason` echoed from the first choice (`stop`, `length`,\n * `content_filter`, `tool_calls`, ...). `null` when the provider omits it.\n * Exposed so a free-form `callLlm` caller CAN detect a truncated answer\n * (`length`) instead of treating a cut-off completion as complete. Note:\n * `callLlm` does not itself reject on it — acting on this signal is the\n * caller's responsibility (in-repo free-form drivers do not yet enforce it).\n */\n finishReason?: string | null\n /**\n * True when `content.trim()` is empty. An empty completion is a silent zero\n * for free-form `callLlm` callers; this flag is the signal a caller can\n * inspect to fail loud rather than proceed on an empty string. `callLlm`\n * surfaces it but does not throw on it.\n */\n contentEmpty?: boolean\n /** Raw response body. */\n raw: Record<string, unknown>\n}\n\nexport type LlmCallMetadata = Pick<LlmCallResult, 'usage' | 'costUsd' | 'model' | 'durationMs'>\n\n/** Convert a provider result into the canonical paid-call receipt input. */\nexport function costReceiptFromLlm(result: LlmCallResult): CostReceiptInput {\n const cachedTokens = result.usage.cachedPromptTokens ?? 0\n return {\n model: result.model,\n inputTokens: Math.max(0, result.usage.promptTokens - cachedTokens),\n outputTokens: result.usage.completionTokens,\n cachedTokens: cachedTokens > 0 ? cachedTokens : undefined,\n actualCostUsd: result.costUsd ?? undefined,\n usageUnknown: result.usage.captured === false,\n }\n}\n\n/** Structured-response failures retain their completed provider receipt. */\nexport function costReceiptFromLlmError(error: Error): CostReceiptInput | undefined {\n return error instanceof LlmResponseError ? costReceiptFromLlm(error.result) : undefined\n}\n\nexport class LlmCallError extends AgentEvalError {\n constructor(\n message: string,\n public readonly status: number,\n public readonly body: string,\n public readonly model: string,\n ) {\n super('judge', message)\n }\n}\n\n/** A provider response completed and incurred measurable usage, but its content\n * could not satisfy the caller's response contract. The response envelope is\n * retained so accounting can commit the receipt before the error propagates. */\nexport class LlmResponseError extends AgentEvalError {\n constructor(\n message: string,\n public readonly result: LlmCallResult,\n options?: { cause?: unknown },\n ) {\n super('judge', message, options)\n }\n}\n\nexport interface LlmClientOptions {\n /** Base URL (without trailing slash). Must end at the `/v1` prefix. */\n baseUrl?: string\n /** Bearer token — either `apiKey` or `bearer` populates `Authorization: Bearer ...`. */\n apiKey?: string\n bearer?: string\n /** Override for the `Authorization` header (e.g. `X-Auth: ...`). Takes precedence over apiKey/bearer. */\n authHeader?: { name: string; value: string }\n /** Stable provider idempotency key, reused across retries of this logical call. */\n idempotencyKey?: string\n /** Default timeout in ms. Per-call can override. */\n defaultTimeoutMs?: number\n /**\n * Caller-supplied abort signal — e.g. a campaign-wide cancel. Linked to\n * each attempt's per-attempt timeout controller, so aborting it cancels\n * the in-flight fetch. A caller abort is FATAL: it is not retried even\n * though an AbortError otherwise matches the transient patterns.\n */\n signal?: AbortSignal\n /**\n * Cross-attempt wall-clock budget in ms, measured from the first attempt.\n * Before launching each attempt the loop checks the remaining budget and\n * stops retrying once it is exhausted, rather than waiting the full\n * per-attempt timeout on every retry. Bounds total time independent of\n * total attempts × `timeoutMs`.\n */\n deadlineMs?: number\n /** Total provider attempts. Legacy option name; default 3 (1 initial + 2 retries). */\n maxRetries?: number\n /** Fetch implementation — defaults to global `fetch`. Override for custom transport (e.g. tests). */\n fetch?: typeof fetch\n /**\n * Optional raw HTTP capture sink. When provided, every request, response,\n * and error (across all retry attempts) is recorded to the sink, with auth\n * headers and credential-shaped body fields redacted by default. This is\n * the layer-1 forensics primitive: structured `LlmSpan`s record intent,\n * raw events record what actually crossed the wire.\n */\n rawSink?: RawProviderSink\n /**\n * Logical provider id attached to raw events. When omitted, derived from\n * `baseUrl` via `providerFromBaseUrl`.\n */\n provider?: string\n /** Trace context attached to raw events; populated by emitter-aware callers. */\n traceContext?: { runId?: string; spanId?: string }\n /** Override the redaction strategy for this call. Defaults to `defaultProviderRedactor`. */\n redactor?: ProviderRedactor\n}\n\n// ─── Internals ──────────────────────────────────────────────────────────\n\nconst DEFAULT_BASE_URL = 'https://router.tangle.tools/v1'\n// Flagship / reasoning models routinely take several minutes on large prompts (a\n// reflection over many failures, a long tool transcript). A tight cap aborts a\n// legitimately-slow but healthy call — and because every retry attempt re-uses\n// the same window, such a model aborts on ALL attempts and the loop throws. The\n// default is generous enough to let those complete, bounded enough that a truly\n// hung call still fails over after retries, and tunable per deployment via\n// TANGLE_LLM_TIMEOUT_MS. Per-call `req.timeoutMs` / `opts.defaultTimeoutMs`\n// still win for callers that know their model's latency.\nconst DEFAULT_TIMEOUT_MS = Number(process.env.TANGLE_LLM_TIMEOUT_MS) || 300_000\nconst DEFAULT_MAX_RETRIES = Number(process.env.TANGLE_LLM_MAX_RETRIES) || 3\n\nfunction resolveMaximumAttempts(configured: number | undefined): number {\n const attempts = configured ?? DEFAULT_MAX_RETRIES\n if (!Number.isInteger(attempts) || attempts <= 0) {\n throw new RangeError('LLM maximum attempts must be a positive integer')\n }\n return attempts\n}\n\nfunction providerTokenCount(value: unknown): number | undefined {\n return typeof value === 'number' && Number.isSafeInteger(value) && value >= 0 ? value : undefined\n}\n\nconst RETRYABLE_STATUS = new Set([429, 502, 503, 504])\n\n/**\n * Transient transport/network error signatures, matched against an error's\n * name, message, and `code`. Covers fetch/undici network failures, aborts\n * and timeouts, and — critically — HTTP/2 transport faults a keep-alive\n * connection raises mid-response: `terminated`, `NGHTTP2_INTERNAL_ERROR`,\n * `UND_ERR_*`, `other side closed`. Those last ones carry no clean HTTP\n * status; unrecognised, they escape the retry loop and surface as an\n * uncaught rejection.\n */\nconst TRANSIENT_ERROR_PATTERNS: readonly RegExp[] = [\n /AbortError/i,\n /TimeoutError/i,\n /this operation was aborted/i,\n /fetch failed/i,\n /ECONNRESET/i,\n /ETIMEDOUT/i,\n /EAI_AGAIN/i,\n /socket hang up/i,\n /stream.*ended.*unexpectedly/i,\n /terminated/i,\n /other side closed/i,\n /NGHTTP2/i,\n /UND_ERR/i,\n]\n\n/**\n * True when an error is a transient transport/network fault worth retrying,\n * as opposed to a deterministic failure (4xx schema reject, JSON parse) that\n * a retry cannot fix. Inspects `LlmCallError.status`, then the error's\n * name/message/code, then recurses into `error.cause` — undici nests the\n * real socket fault one or more levels under `.cause`.\n *\n * This is THE retry classifier for the package: `callLlm` and\n * `withJudgeRetry` both route through it, so a connection-class error is\n * treated identically whether it surfaces in the HTTP client or a\n * TCloud-backed judge.\n */\nexport function isTransientLlmError(err: unknown): boolean {\n return classifyTransient(err, 0)\n}\n\nfunction classifyTransient(err: unknown, depth: number): boolean {\n if (err instanceof LlmCallError) return RETRYABLE_STATUS.has(err.status)\n if (!(err instanceof Error)) return false\n // Foreign errors (e.g. a TCloud judge SDK error) can carry a numeric HTTP\n // status without being an LlmCallError — a retryable status is decisive.\n const status = (err as { status?: unknown }).status\n if (typeof status === 'number' && RETRYABLE_STATUS.has(status)) return true\n const code = (err as { code?: unknown }).code\n const haystack = `${err.name}\\n${err.message}\\n${typeof code === 'string' ? code : ''}`\n if (TRANSIENT_ERROR_PATTERNS.some((p) => p.test(haystack))) return true\n const cause = (err as { cause?: unknown }).cause\n if (depth < 4 && cause instanceof Error && cause !== err) {\n return classifyTransient(cause, depth + 1)\n }\n return false\n}\n\nfunction parseRetryAfter(headers: Headers): number | null {\n const h = headers.get('retry-after')\n if (!h) return null\n const asNumber = Number(h)\n if (Number.isFinite(asNumber) && asNumber > 0) return asNumber * 1000\n const asDate = Date.parse(h)\n if (Number.isFinite(asDate)) return Math.max(0, asDate - Date.now())\n return null\n}\n\n/** Exponential backoff: 500ms, 1s, 2s, 4s, ... capped at 16s. Attempt is 0-indexed. */\nexport function backoffMs(attempt: number): number {\n return Math.min(500 * 2 ** attempt, 16_000)\n}\n\nfunction buildHeaders(opts: LlmClientOptions): Record<string, string> {\n const headers: Record<string, string> = {\n 'Content-Type': 'application/json',\n Accept: 'application/json',\n }\n if (opts.authHeader) {\n headers[opts.authHeader.name] = opts.authHeader.value\n } else if (opts.bearer || opts.apiKey) {\n headers.Authorization = `Bearer ${opts.bearer ?? opts.apiKey}`\n }\n if (opts.idempotencyKey) headers['Idempotency-Key'] = opts.idempotencyKey\n return headers\n}\n\nfunction isSchemaRejection(status: number, body: string): boolean {\n if (status !== 400) return false\n const lower = body.toLowerCase()\n return (\n lower.includes('response_format') ||\n lower.includes('json_schema') ||\n lower.includes('is unavailable') ||\n lower.includes('not supported')\n )\n}\n\nfunction buildBody(req: LlmCallRequest, forceJsonObject: boolean): Record<string, unknown> {\n const body: Record<string, unknown> = {\n model: req.model,\n messages: req.messages,\n temperature: req.temperature ?? 0,\n }\n if (req.maxTokens != null) {\n if (usesMaxCompletionTokens(req.model)) body.max_completion_tokens = req.maxTokens\n else body.max_tokens = req.maxTokens\n }\n\n if (req.jsonSchema && !forceJsonObject) {\n body.response_format = {\n type: 'json_schema',\n json_schema: { name: req.jsonSchema.name, schema: req.jsonSchema.schema, strict: true },\n }\n } else if (req.jsonMode || req.jsonSchema) {\n body.response_format = { type: 'json_object' }\n }\n\n return body\n}\n\nfunction usesMaxCompletionTokens(model: string): boolean {\n return /^gpt-5(?:[.-]|$)/i.test(model)\n}\n\nasync function sleep(ms: number): Promise<void> {\n return new Promise((resolve) => setTimeout(resolve, ms))\n}\n\n/**\n * Combine the per-attempt timeout signal with an optional caller signal into\n * one signal the fetch listens on. Prefers the native `AbortSignal.any`; falls\n * back to manual wiring on runtimes that predate it. The caller signal is also\n * propagated to the timeout controller so aborting it cancels the in-flight\n * fetch immediately.\n */\nfunction linkSignals(timeoutController: AbortController, caller?: AbortSignal): AbortSignal {\n if (!caller) return timeoutController.signal\n if (typeof (AbortSignal as { any?: unknown }).any === 'function') {\n return AbortSignal.any([timeoutController.signal, caller])\n }\n if (caller.aborted) {\n timeoutController.abort()\n } else {\n caller.addEventListener('abort', () => timeoutController.abort(), { once: true })\n }\n return timeoutController.signal\n}\n\n/** True once the cross-attempt wall-clock budget (if any) is exhausted. */\nfunction deadlineExceeded(start: number, deadlineMs: number | undefined): boolean {\n return deadlineMs != null && Date.now() - start >= deadlineMs\n}\n\n// ─── Public API ─────────────────────────────────────────────────────────\n\n/**\n * Strip a ```json / ``` code fence if the model emitted one.\n * Idempotent for naked JSON. Some models (claude-code via router, certain\n * deepseek models) wrap output even under json_object.\n */\nexport function stripFencedJson(raw: string): string {\n const trimmed = raw.trim()\n const m = trimmed.match(/^```(?:json)?\\s*\\n?([\\s\\S]*?)\\n?```\\s*$/)\n return m ? m[1]!.trim() : trimmed\n}\n\nexport function extractJsonPayload(raw: string): string {\n const stripped = stripFencedJson(raw)\n try {\n JSON.parse(stripped)\n return stripped\n } catch {\n // A response that declares a JSON root must parse as that complete root.\n // Scanning onward could turn a truncated object into one of its valid nested\n // arrays or objects and silently change the response schema.\n if (stripped.startsWith('{') || stripped.startsWith('[')) return stripped\n }\n\n // Only prose-leading responses may contain a recoverable JSON payload.\n const starts = [...stripped.matchAll(/[[{]/g)]\n .map((match) => match.index)\n .filter((index) => index != null)\n for (const start of starts) {\n const candidate = extractBalancedJson(stripped, start)\n if (!candidate) continue\n try {\n JSON.parse(candidate)\n return candidate\n } catch {\n // Keep scanning; earlier braces may belong to prose.\n }\n }\n\n return stripped\n}\n\nfunction extractBalancedJson(input: string, start: number): string | null {\n const opener = input[start]\n const closer = opener === '{' ? '}' : opener === '[' ? ']' : null\n if (!closer) return null\n\n const stack: string[] = [closer]\n let isInString = false\n let isEscaped = false\n\n for (let i = start + 1; i < input.length; i++) {\n const char = input[i]!\n if (isEscaped) {\n isEscaped = false\n continue\n }\n if (char === '\\\\') {\n isEscaped = isInString\n continue\n }\n if (char === '\"') {\n isInString = !isInString\n continue\n }\n if (isInString) continue\n\n if (char === '{') stack.push('}')\n else if (char === '[') stack.push(']')\n else if (char === stack[stack.length - 1]) {\n stack.pop()\n if (stack.length === 0) return input.slice(start, i + 1)\n }\n }\n\n return null\n}\n\n/**\n * Low-level call. Returns raw content + usage + cost. Retries on transient\n * failures; does NOT degrade schema here — callers that want graceful\n * degrade use `callLlmJson`.\n */\nexport async function callLlm(\n req: LlmCallRequest,\n opts: LlmClientOptions = {},\n): Promise<LlmCallResult> {\n const baseUrl = (opts.baseUrl ?? DEFAULT_BASE_URL).replace(/\\/+$/, '')\n const url = `${baseUrl}/chat/completions`\n const endpoint = '/chat/completions'\n const timeoutMs = req.timeoutMs ?? opts.defaultTimeoutMs ?? DEFAULT_TIMEOUT_MS\n const maximumAttempts = resolveMaximumAttempts(opts.maxRetries)\n const fetchFn = opts.fetch ?? globalThis.fetch\n const headers = buildHeaders(opts)\n const provider = opts.provider ?? providerFromBaseUrl(baseUrl)\n const sink = opts.rawSink\n const redactor = opts.redactor ?? defaultProviderRedactor\n const traceContext = opts.traceContext\n const callerSignal = opts.signal\n const deadlineMs = opts.deadlineMs\n const deadlineStart = Date.now()\n\n let lastErr: unknown\n for (let attempt = 0; attempt < maximumAttempts; attempt++) {\n // A caller cancel is fatal — never retried. Checking before each attempt\n // means an already-aborted signal short-circuits without firing fetch.\n if (callerSignal?.aborted) {\n throw new DOMException('callLlm aborted by caller signal', 'AbortError')\n }\n // Stop retrying once the cross-attempt budget is spent rather than burning\n // a full per-attempt timeout on each remaining retry.\n if (attempt > 0 && deadlineExceeded(deadlineStart, deadlineMs)) {\n throw lastErr instanceof Error ? lastErr : new Error(String(lastErr))\n }\n const controller = new AbortController()\n const attemptSignal = linkSignals(controller, callerSignal)\n const timeoutHandle = setTimeout(() => controller.abort(), timeoutMs)\n const started = Date.now()\n const requestBody = buildBody(req, false)\n let attemptErrorRecorded = false\n if (sink) {\n await recordRaw(sink, redactor, {\n eventId: cryptoEventId(),\n runId: traceContext?.runId,\n spanId: traceContext?.spanId,\n provider,\n model: req.model,\n endpoint,\n baseUrl,\n attemptIndex: attempt,\n direction: 'request',\n timestamp: started,\n requestHeaders: headers,\n requestBody,\n redactedFields: [],\n })\n }\n\n try {\n const res = await fetchFn(url, {\n method: 'POST',\n headers,\n body: JSON.stringify(requestBody),\n signal: attemptSignal,\n })\n clearTimeout(timeoutHandle)\n const responseHeaders = sink ? headersToObject(res.headers) : undefined\n\n if (!res.ok) {\n const body = await res.text()\n if (sink) {\n await recordRaw(sink, redactor, {\n eventId: cryptoEventId(),\n runId: traceContext?.runId,\n spanId: traceContext?.spanId,\n provider,\n model: req.model,\n endpoint,\n baseUrl,\n attemptIndex: attempt,\n direction: 'error',\n timestamp: Date.now(),\n durationMs: Date.now() - started,\n statusCode: res.status,\n responseHeaders,\n responseBody: body,\n errorMessage: `HTTP ${res.status}`,\n redactedFields: [],\n })\n attemptErrorRecorded = true\n }\n const err = new LlmCallError(\n `LLM call ${res.status}: ${body.slice(0, 300)}`,\n res.status,\n body,\n req.model,\n )\n if (\n RETRYABLE_STATUS.has(res.status) &&\n attempt < maximumAttempts - 1 &&\n !deadlineExceeded(deadlineStart, deadlineMs)\n ) {\n lastErr = err\n const retryAfter = parseRetryAfter(res.headers)\n await sleep(retryAfter ?? backoffMs(attempt))\n continue\n }\n throw err\n }\n\n const text = await res.text()\n let json: Record<string, unknown>\n try {\n json = JSON.parse(text) as Record<string, unknown>\n } catch (parseErr) {\n if (sink) {\n await recordRaw(sink, redactor, {\n eventId: cryptoEventId(),\n runId: traceContext?.runId,\n spanId: traceContext?.spanId,\n provider,\n model: req.model,\n endpoint,\n baseUrl,\n attemptIndex: attempt,\n direction: 'error',\n timestamp: Date.now(),\n durationMs: Date.now() - started,\n statusCode: res.status,\n responseHeaders,\n responseBody: text,\n errorMessage: `non-JSON response: ${parseErr instanceof Error ? parseErr.message : String(parseErr)}`,\n redactedFields: [],\n })\n attemptErrorRecorded = true\n }\n throw parseErr\n }\n if (sink) {\n await recordRaw(sink, redactor, {\n eventId: cryptoEventId(),\n runId: traceContext?.runId,\n spanId: traceContext?.spanId,\n provider,\n model: req.model,\n endpoint,\n baseUrl,\n attemptIndex: attempt,\n direction: 'response',\n timestamp: Date.now(),\n durationMs: Date.now() - started,\n statusCode: res.status,\n responseHeaders,\n responseBody: json,\n redactedFields: [],\n })\n }\n const choice = (\n json.choices as\n | Array<{ message?: { content?: string }; finish_reason?: string | null }>\n | undefined\n )?.[0]\n const usageRaw =\n json.usage && typeof json.usage === 'object' && !Array.isArray(json.usage)\n ? (json.usage as Record<string, unknown>)\n : undefined\n const promptTokens = providerTokenCount(usageRaw?.prompt_tokens)\n const completionTokens = providerTokenCount(usageRaw?.completion_tokens)\n const totalTokens = providerTokenCount(usageRaw?.total_tokens)\n const cachedRaw =\n usageRaw?.prompt_tokens_details &&\n typeof usageRaw.prompt_tokens_details === 'object' &&\n !Array.isArray(usageRaw.prompt_tokens_details)\n ? (usageRaw.prompt_tokens_details as Record<string, unknown>).cached_tokens\n : undefined\n const cachedPromptTokens = cachedRaw === undefined ? undefined : providerTokenCount(cachedRaw)\n const usageCaptured =\n promptTokens !== undefined &&\n completionTokens !== undefined &&\n (cachedRaw === undefined ||\n (cachedPromptTokens !== undefined && cachedPromptTokens <= promptTokens)) &&\n (totalTokens === undefined || totalTokens === promptTokens + completionTokens)\n const costFromProxy = (json._response_cost ?? json.cost_usd) as number | undefined\n const content = choice?.message?.content ?? ''\n\n return {\n content,\n finishReason: choice?.finish_reason ?? null,\n contentEmpty: content.trim().length === 0,\n usage: {\n promptTokens: promptTokens ?? 0,\n completionTokens: completionTokens ?? 0,\n totalTokens: totalTokens ?? (promptTokens ?? 0) + (completionTokens ?? 0),\n captured: usageCaptured,\n cachedPromptTokens,\n },\n costUsd: typeof costFromProxy === 'number' ? costFromProxy : null,\n model: (json.model as string) ?? req.model,\n durationMs: Date.now() - started,\n raw: json,\n }\n } catch (err) {\n clearTimeout(timeoutHandle)\n lastErr = err\n // A caller cancel is fatal even though an AbortError matches the\n // transient patterns — a cancelled call must surface immediately, not\n // be retried against the same dead intent.\n if (callerSignal?.aborted) {\n if (sink && !attemptErrorRecorded) {\n await recordRaw(sink, redactor, {\n eventId: cryptoEventId(),\n runId: traceContext?.runId,\n spanId: traceContext?.spanId,\n provider,\n model: req.model,\n endpoint,\n baseUrl,\n attemptIndex: attempt,\n direction: 'error',\n timestamp: Date.now(),\n durationMs: Date.now() - started,\n errorMessage: err instanceof Error ? err.message : String(err),\n redactedFields: [],\n })\n }\n throw err\n }\n if (sink && !attemptErrorRecorded) {\n // Record only if neither the !res.ok branch nor the JSON.parse catch\n // already produced an error event for this attempt. Covers network\n // failures, timeouts, and aborts.\n await recordRaw(sink, redactor, {\n eventId: cryptoEventId(),\n runId: traceContext?.runId,\n spanId: traceContext?.spanId,\n provider,\n model: req.model,\n endpoint,\n baseUrl,\n attemptIndex: attempt,\n direction: 'error',\n timestamp: Date.now(),\n durationMs: Date.now() - started,\n errorMessage: err instanceof Error ? err.message : String(err),\n redactedFields: [],\n })\n }\n if (\n attempt < maximumAttempts - 1 &&\n isTransientLlmError(err) &&\n !deadlineExceeded(deadlineStart, deadlineMs)\n ) {\n await sleep(backoffMs(attempt))\n continue\n }\n throw err\n }\n }\n throw lastErr instanceof Error ? lastErr : new Error(String(lastErr))\n}\n\nasync function recordRaw(\n sink: RawProviderSink,\n redactor: ProviderRedactor,\n event: RawProviderEvent,\n): Promise<void> {\n // Errors from sinks must not crash the LLM call. Forensic capture is\n // best-effort; the structured trace is the system of record.\n try {\n await sink.record(redactor(event))\n } catch {\n // Intentionally swallowed.\n }\n}\n\nfunction headersToObject(h: Headers): Record<string, string> {\n const out: Record<string, string> = {}\n h.forEach((value, key) => {\n out[key] = value\n })\n return out\n}\n\nfunction cryptoEventId(): string {\n if (typeof globalThis.crypto?.randomUUID === 'function') return globalThis.crypto.randomUUID()\n return `${Date.now().toString(36)}-${Math.random().toString(36).slice(2, 10)}`\n}\n\n/**\n * Structured-output call. Returns parsed JSON plus the raw result envelope.\n * Degrades `jsonSchema` → `jsonMode` on a 400 that names the schema param —\n * critical for deepseek-v3/v4, kimi-k2.6, and other models that don't accept\n * the `response_format.json_schema` shape but DO accept `json_object`.\n */\nexport async function callLlmJson<T = unknown>(\n req: LlmCallRequest,\n opts: LlmClientOptions = {},\n): Promise<{ value: T; result: LlmCallResult }> {\n const result = await callLlmStructured(req, opts)\n const value = parseJsonResult<T>(result)\n return { value, result }\n}\n\n/** Shared schema-to-JSON-mode fallback that preserves the raw result. */\nasync function callLlmStructured(\n req: LlmCallRequest,\n opts: LlmClientOptions = {},\n): Promise<LlmCallResult> {\n try {\n return await callLlm({ ...req, jsonMode: req.jsonMode ?? !req.jsonSchema }, opts)\n } catch (err) {\n if (err instanceof LlmCallError && isSchemaRejection(err.status, err.body) && req.jsonSchema) {\n const degradedReq: LlmCallRequest = { ...req, jsonMode: true, jsonSchema: undefined }\n return await callLlm(degradedReq, opts)\n }\n throw err\n }\n}\n\nfunction parseJsonResult<T>(result: LlmCallResult): T {\n try {\n if (result.finishReason === 'length') {\n throw new Error(\n `LLM returned truncated JSON content (model=${result.model}, finishReason=length)`,\n )\n }\n return parseJsonSafely<T>(result.content, result.model)\n } catch (error) {\n if (error instanceof LlmResponseError) throw error\n const cause = error instanceof Error ? error : new Error(String(error))\n throw new LlmResponseError(cause.message, result, { cause })\n }\n}\n\nfunction parseJsonSafely<T>(content: string, model: string): T {\n const stripped = extractJsonPayload(content)\n try {\n return JSON.parse(stripped) as T\n } catch (err) {\n throw new Error(\n `LLM returned non-JSON content (model=${model}): ${\n err instanceof Error ? err.message : String(err)\n }\\n--- raw content ---\\n${content.slice(0, 800)}`,\n )\n }\n}\n\n// ─── Route assertion ────────────────────────────────────────────────────\n\nexport type LlmRouteAssertionReason =\n | 'no_explicit_base_url'\n | 'base_url_blocked'\n | 'base_url_not_allowed'\n | 'no_auth'\n | 'wrong_provider'\n\nexport class LlmRouteAssertionError extends CaptureIntegrityError {\n constructor(\n message: string,\n public readonly reason: LlmRouteAssertionReason,\n public readonly baseUrl: string,\n ) {\n super(message)\n }\n}\n\nexport interface LlmRouteRequirements {\n /**\n * Throw if `opts.baseUrl` is undefined, i.e. the call would fall back to\n * `DEFAULT_BASE_URL`. Set this for evaluation runs where silently using\n * the public/free-tier router is a defect — the launch reviewer needs to\n * know exactly which provider answered.\n */\n requireExplicitBaseUrl?: boolean\n /**\n * Allowlist of acceptable base URLs. Strings match by prefix\n * (case-insensitive); RegExps test against the full base URL.\n */\n allowedBaseUrls?: Array<string | RegExp>\n /** Blocklist that takes precedence over `allowedBaseUrls`. */\n blockedBaseUrls?: Array<string | RegExp>\n /** Throw if no auth header / api key is configured. */\n requireAuth?: boolean\n /**\n * Logical provider id the configured `baseUrl` is expected to match (via\n * `providerFromBaseUrl`). Mainly useful when paired with `requireExplicitBaseUrl`.\n */\n expectedProvider?: string\n}\n\n/**\n * Fail-loud assertion that the configured LLM client points at the route\n * the caller intends. Designed for the matrix-runner preflight: invoke\n * once before any LLM call to catch misconfiguration before a sweep burns\n * dollars on the wrong provider.\n *\n * Throws `LlmRouteAssertionError`. Pure — no I/O — so it's safe to call\n * from constructors and CI gates.\n */\nexport function assertLlmRoute(opts: LlmClientOptions, req: LlmRouteRequirements = {}): void {\n const baseUrlExplicit = opts.baseUrl !== undefined\n const baseUrl = (opts.baseUrl ?? DEFAULT_BASE_URL).replace(/\\/+$/, '')\n\n if (req.requireExplicitBaseUrl && !baseUrlExplicit) {\n throw new LlmRouteAssertionError(\n `assertLlmRoute: requireExplicitBaseUrl set but opts.baseUrl is undefined; would fall back to ${DEFAULT_BASE_URL}.`,\n 'no_explicit_base_url',\n baseUrl,\n )\n }\n\n if (req.blockedBaseUrls?.some((p) => matchUrl(baseUrl, p))) {\n throw new LlmRouteAssertionError(\n `assertLlmRoute: baseUrl ${baseUrl} matches a blocked pattern.`,\n 'base_url_blocked',\n baseUrl,\n )\n }\n\n if (req.allowedBaseUrls && req.allowedBaseUrls.length > 0) {\n const ok = req.allowedBaseUrls.some((p) => matchUrl(baseUrl, p))\n if (!ok) {\n throw new LlmRouteAssertionError(\n `assertLlmRoute: baseUrl ${baseUrl} is not in the allowed list (${req.allowedBaseUrls.map(describePattern).join(', ')}).`,\n 'base_url_not_allowed',\n baseUrl,\n )\n }\n }\n\n if (req.requireAuth && !opts.apiKey && !opts.bearer && !opts.authHeader) {\n throw new LlmRouteAssertionError(\n `assertLlmRoute: requireAuth set but no apiKey, bearer, or authHeader was supplied.`,\n 'no_auth',\n baseUrl,\n )\n }\n\n if (req.expectedProvider) {\n const actual = opts.provider ?? providerFromBaseUrl(baseUrl)\n if (actual !== req.expectedProvider) {\n throw new LlmRouteAssertionError(\n `assertLlmRoute: expected provider ${req.expectedProvider} but baseUrl ${baseUrl} resolves to ${actual}.`,\n 'wrong_provider',\n baseUrl,\n )\n }\n }\n}\n\nfunction matchUrl(url: string, pattern: string | RegExp): boolean {\n if (pattern instanceof RegExp) return pattern.test(url)\n return url.toLowerCase().startsWith(pattern.toLowerCase())\n}\n\nfunction describePattern(p: string | RegExp): string {\n return p instanceof RegExp ? p.source : p\n}\n\n/**\n * Probe whether a model is reachable. Returns latency + null error on\n * success; `ok=false` + error message on any failure (HTTP, timeout,\n * network, parse). Designed for sweep preflights — fail loud at the\n * boundary before burning a 30-leaf run on a misconfigured router.\n *\n * Sends a tiny `ping` message with `maxTokens=64`. Reasoning models\n * (glm-5.1, deepseek-v4) can burn the entire budget on internal reasoning\n * for short prompts, so don't tighten this further. We don't validate\n * content; HTTP 200 means reachable.\n */\nexport async function probeLlm(\n model: string,\n opts: LlmClientOptions & { timeoutMs?: number } = {},\n): Promise<{ ok: boolean; latencyMs: number; error: string | null }> {\n const start = Date.now()\n try {\n await callLlm(\n {\n model,\n messages: [{ role: 'user', content: 'ping' }],\n maxTokens: 64,\n timeoutMs: opts.timeoutMs ?? 30_000,\n },\n opts,\n )\n return { ok: true, latencyMs: Date.now() - start, error: null }\n } catch (err) {\n return {\n ok: false,\n latencyMs: Date.now() - start,\n error: err instanceof Error ? err.message : String(err),\n }\n }\n}\n\n/**\n * Stateful client — construct once with defaults, call many times.\n * Thin wrapper around the free functions; exists for callers that want\n * to inject a single configured instance into multiple primitives.\n */\nexport class LlmClient {\n readonly maximumAttempts: number\n private readonly opts: LlmClientOptions\n\n constructor(opts: LlmClientOptions = {}) {\n this.opts = opts\n this.maximumAttempts = resolveMaximumAttempts(opts.maxRetries)\n }\n\n call(req: LlmCallRequest, per?: LlmClientOptions): Promise<LlmCallResult> {\n const options = { ...this.opts, ...per }\n return req.jsonSchema ? callLlmStructured(req, options) : callLlm(req, options)\n }\n\n callJson<T = unknown>(\n req: LlmCallRequest,\n per?: LlmClientOptions,\n ): Promise<{ value: T; result: LlmCallResult }> {\n return callLlmJson<T>(req, { ...this.opts, ...per })\n }\n}\n"],"mappings":";;;;;;;;;;AAgEO,SAAS,2BACd,SACA,UAA4B,CAAC,GACF;AAC3B,MAAI,QAAQ,cAAc,OAAW,QAAO;AAC5C,MAAI,CAAC,OAAO,UAAU,QAAQ,SAAS,KAAK,QAAQ,aAAa,GAAG;AAClE,UAAM,IAAI,WAAW,kEAAkE;AAAA,EACzF;AACA,MACE,QAAQ,SAAS;AAAA,IACf,CAAC,YACC,MAAM,QAAQ,QAAQ,OAAO,KAAK,QAAQ,QAAQ,KAAK,CAAC,SAAS,KAAK,SAAS,WAAW;AAAA,EAC9F,GACA;AACA,WAAO;AAAA,EACT;AAEA,QAAM,WAAW,uBAAuB,QAAQ,UAAU;AAG1D,QAAM,eAAe,IAAI,YAAY,EAAE;AAAA,IACrC,KAAK,UAAU,UAAU,SAAS,KAAK,CAAC;AAAA,EAC1C,EAAE;AAEF,QAAM,UAAU,QAAQ,aAAa,IAAI;AACzC,SAAO;AAAA,IACL,OAAO,QAAQ;AAAA,IACf,aAAa,eAAe,WAAW;AAAA,IACvC,cAAc,QAAQ,YAAY,WAAW;AAAA,EAC/C;AACF;AAgDO,SAAS,mBAAmB,QAAyC;AAC1E,QAAM,eAAe,OAAO,MAAM,sBAAsB;AACxD,SAAO;AAAA,IACL,OAAO,OAAO;AAAA,IACd,aAAa,KAAK,IAAI,GAAG,OAAO,MAAM,eAAe,YAAY;AAAA,IACjE,cAAc,OAAO,MAAM;AAAA,IAC3B,cAAc,eAAe,IAAI,eAAe;AAAA,IAChD,eAAe,OAAO,WAAW;AAAA,IACjC,cAAc,OAAO,MAAM,aAAa;AAAA,EAC1C;AACF;AAGO,SAAS,wBAAwB,OAA4C;AAClF,SAAO,iBAAiB,mBAAmB,mBAAmB,MAAM,MAAM,IAAI;AAChF;AAEO,IAAM,eAAN,cAA2B,eAAe;AAAA,EAC/C,YACE,SACgB,QACA,MACA,OAChB;AACA,UAAM,SAAS,OAAO;AAJN;AACA;AACA;AAAA,EAGlB;AAAA,EALkB;AAAA,EACA;AAAA,EACA;AAIpB;AAKO,IAAM,mBAAN,cAA+B,eAAe;AAAA,EACnD,YACE,SACgB,QAChB,SACA;AACA,UAAM,SAAS,SAAS,OAAO;AAHf;AAAA,EAIlB;AAAA,EAJkB;AAKpB;AAsDA,IAAM,mBAAmB;AASzB,IAAM,qBAAqB,OAAO,QAAQ,IAAI,qBAAqB,KAAK;AACxE,IAAM,sBAAsB,OAAO,QAAQ,IAAI,sBAAsB,KAAK;AAE1E,SAAS,uBAAuB,YAAwC;AACtE,QAAM,WAAW,cAAc;AAC/B,MAAI,CAAC,OAAO,UAAU,QAAQ,KAAK,YAAY,GAAG;AAChD,UAAM,IAAI,WAAW,iDAAiD;AAAA,EACxE;AACA,SAAO;AACT;AAEA,SAAS,mBAAmB,OAAoC;AAC9D,SAAO,OAAO,UAAU,YAAY,OAAO,cAAc,KAAK,KAAK,SAAS,IAAI,QAAQ;AAC1F;AAEA,IAAM,mBAAmB,oBAAI,IAAI,CAAC,KAAK,KAAK,KAAK,GAAG,CAAC;AAWrD,IAAM,2BAA8C;AAAA,EAClD;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AAAA,EACA;AACF;AAcO,SAAS,oBAAoB,KAAuB;AACzD,SAAO,kBAAkB,KAAK,CAAC;AACjC;AAEA,SAAS,kBAAkB,KAAc,OAAwB;AAC/D,MAAI,eAAe,aAAc,QAAO,iBAAiB,IAAI,IAAI,MAAM;AACvE,MAAI,EAAE,eAAe,OAAQ,QAAO;AAGpC,QAAM,SAAU,IAA6B;AAC7C,MAAI,OAAO,WAAW,YAAY,iBAAiB,IAAI,MAAM,EAAG,QAAO;AACvE,QAAM,OAAQ,IAA2B;AACzC,QAAM,WAAW,GAAG,IAAI,IAAI;AAAA,EAAK,IAAI,OAAO;AAAA,EAAK,OAAO,SAAS,WAAW,OAAO,EAAE;AACrF,MAAI,yBAAyB,KAAK,CAAC,MAAM,EAAE,KAAK,QAAQ,CAAC,EAAG,QAAO;AACnE,QAAM,QAAS,IAA4B;AAC3C,MAAI,QAAQ,KAAK,iBAAiB,SAAS,UAAU,KAAK;AACxD,WAAO,kBAAkB,OAAO,QAAQ,CAAC;AAAA,EAC3C;AACA,SAAO;AACT;AAEA,SAAS,gBAAgB,SAAiC;AACxD,QAAM,IAAI,QAAQ,IAAI,aAAa;AACnC,MAAI,CAAC,EAAG,QAAO;AACf,QAAM,WAAW,OAAO,CAAC;AACzB,MAAI,OAAO,SAAS,QAAQ,KAAK,WAAW,EAAG,QAAO,WAAW;AACjE,QAAM,SAAS,KAAK,MAAM,CAAC;AAC3B,MAAI,OAAO,SAAS,MAAM,EAAG,QAAO,KAAK,IAAI,GAAG,SAAS,KAAK,IAAI,CAAC;AACnE,SAAO;AACT;AAGO,SAAS,UAAU,SAAyB;AACjD,SAAO,KAAK,IAAI,MAAM,KAAK,SAAS,IAAM;AAC5C;AAEA,SAAS,aAAa,MAAgD;AACpE,QAAM,UAAkC;AAAA,IACtC,gBAAgB;AAAA,IAChB,QAAQ;AAAA,EACV;AACA,MAAI,KAAK,YAAY;AACnB,YAAQ,KAAK,WAAW,IAAI,IAAI,KAAK,WAAW;AAAA,EAClD,WAAW,KAAK,UAAU,KAAK,QAAQ;AACrC,YAAQ,gBAAgB,UAAU,KAAK,UAAU,KAAK,MAAM;AAAA,EAC9D;AACA,MAAI,KAAK,eAAgB,SAAQ,iBAAiB,IAAI,KAAK;AAC3D,SAAO;AACT;AAEA,SAAS,kBAAkB,QAAgB,MAAuB;AAChE,MAAI,WAAW,IAAK,QAAO;AAC3B,QAAM,QAAQ,KAAK,YAAY;AAC/B,SACE,MAAM,SAAS,iBAAiB,KAChC,MAAM,SAAS,aAAa,KAC5B,MAAM,SAAS,gBAAgB,KAC/B,MAAM,SAAS,eAAe;AAElC;AAEA,SAAS,UAAU,KAAqB,iBAAmD;AACzF,QAAM,OAAgC;AAAA,IACpC,OAAO,IAAI;AAAA,IACX,UAAU,IAAI;AAAA,IACd,aAAa,IAAI,eAAe;AAAA,EAClC;AACA,MAAI,IAAI,aAAa,MAAM;AACzB,QAAI,wBAAwB,IAAI,KAAK,EAAG,MAAK,wBAAwB,IAAI;AAAA,QACpE,MAAK,aAAa,IAAI;AAAA,EAC7B;AAEA,MAAI,IAAI,cAAc,CAAC,iBAAiB;AACtC,SAAK,kBAAkB;AAAA,MACrB,MAAM;AAAA,MACN,aAAa,EAAE,MAAM,IAAI,WAAW,MAAM,QAAQ,IAAI,WAAW,QAAQ,QAAQ,KAAK;AAAA,IACxF;AAAA,EACF,WAAW,IAAI,YAAY,IAAI,YAAY;AACzC,SAAK,kBAAkB,EAAE,MAAM,cAAc;AAAA,EAC/C;AAEA,SAAO;AACT;AAEA,SAAS,wBAAwB,OAAwB;AACvD,SAAO,oBAAoB,KAAK,KAAK;AACvC;AAEA,eAAe,MAAM,IAA2B;AAC9C,SAAO,IAAI,QAAQ,CAAC,YAAY,WAAW,SAAS,EAAE,CAAC;AACzD;AASA,SAAS,YAAY,mBAAoC,QAAmC;AAC1F,MAAI,CAAC,OAAQ,QAAO,kBAAkB;AACtC,MAAI,OAAQ,YAAkC,QAAQ,YAAY;AAChE,WAAO,YAAY,IAAI,CAAC,kBAAkB,QAAQ,MAAM,CAAC;AAAA,EAC3D;AACA,MAAI,OAAO,SAAS;AAClB,sBAAkB,MAAM;AAAA,EAC1B,OAAO;AACL,WAAO,iBAAiB,SAAS,MAAM,kBAAkB,MAAM,GAAG,EAAE,MAAM,KAAK,CAAC;AAAA,EAClF;AACA,SAAO,kBAAkB;AAC3B;AAGA,SAAS,iBAAiB,OAAe,YAAyC;AAChF,SAAO,cAAc,QAAQ,KAAK,IAAI,IAAI,SAAS;AACrD;AASO,SAAS,gBAAgB,KAAqB;AACnD,QAAM,UAAU,IAAI,KAAK;AACzB,QAAM,IAAI,QAAQ,MAAM,yCAAyC;AACjE,SAAO,IAAI,EAAE,CAAC,EAAG,KAAK,IAAI;AAC5B;AAEO,SAAS,mBAAmB,KAAqB;AACtD,QAAM,WAAW,gBAAgB,GAAG;AACpC,MAAI;AACF,SAAK,MAAM,QAAQ;AACnB,WAAO;AAAA,EACT,QAAQ;AAIN,QAAI,SAAS,WAAW,GAAG,KAAK,SAAS,WAAW,GAAG,EAAG,QAAO;AAAA,EACnE;AAGA,QAAM,SAAS,CAAC,GAAG,SAAS,SAAS,OAAO,CAAC,EAC1C,IAAI,CAAC,UAAU,MAAM,KAAK,EAC1B,OAAO,CAAC,UAAU,SAAS,IAAI;AAClC,aAAW,SAAS,QAAQ;AAC1B,UAAM,YAAY,oBAAoB,UAAU,KAAK;AACrD,QAAI,CAAC,UAAW;AAChB,QAAI;AACF,WAAK,MAAM,SAAS;AACpB,aAAO;AAAA,IACT,QAAQ;AAAA,IAER;AAAA,EACF;AAEA,SAAO;AACT;AAEA,SAAS,oBAAoB,OAAe,OAA8B;AACxE,QAAM,SAAS,MAAM,KAAK;AAC1B,QAAM,SAAS,WAAW,MAAM,MAAM,WAAW,MAAM,MAAM;AAC7D,MAAI,CAAC,OAAQ,QAAO;AAEpB,QAAM,QAAkB,CAAC,MAAM;AAC/B,MAAI,aAAa;AACjB,MAAI,YAAY;AAEhB,WAAS,IAAI,QAAQ,GAAG,IAAI,MAAM,QAAQ,KAAK;AAC7C,UAAM,OAAO,MAAM,CAAC;AACpB,QAAI,WAAW;AACb,kBAAY;AACZ;AAAA,IACF;AACA,QAAI,SAAS,MAAM;AACjB,kBAAY;AACZ;AAAA,IACF;AACA,QAAI,SAAS,KAAK;AAChB,mBAAa,CAAC;AACd;AAAA,IACF;AACA,QAAI,WAAY;AAEhB,QAAI,SAAS,IAAK,OAAM,KAAK,GAAG;AAAA,aACvB,SAAS,IAAK,OAAM,KAAK,GAAG;AAAA,aAC5B,SAAS,MAAM,MAAM,SAAS,CAAC,GAAG;AACzC,YAAM,IAAI;AACV,UAAI,MAAM,WAAW,EAAG,QAAO,MAAM,MAAM,OAAO,IAAI,CAAC;AAAA,IACzD;AAAA,EACF;AAEA,SAAO;AACT;AAOA,eAAsB,QACpB,KACA,OAAyB,CAAC,GACF;AACxB,QAAM,WAAW,KAAK,WAAW,kBAAkB,QAAQ,QAAQ,EAAE;AACrE,QAAM,MAAM,GAAG,OAAO;AACtB,QAAM,WAAW;AACjB,QAAM,YAAY,IAAI,aAAa,KAAK,oBAAoB;AAC5D,QAAM,kBAAkB,uBAAuB,KAAK,UAAU;AAC9D,QAAM,UAAU,KAAK,SAAS,WAAW;AACzC,QAAM,UAAU,aAAa,IAAI;AACjC,QAAM,WAAW,KAAK,YAAY,oBAAoB,OAAO;AAC7D,QAAM,OAAO,KAAK;AAClB,QAAM,WAAW,KAAK,YAAY;AAClC,QAAM,eAAe,KAAK;AAC1B,QAAM,eAAe,KAAK;AAC1B,QAAM,aAAa,KAAK;AACxB,QAAM,gBAAgB,KAAK,IAAI;AAE/B,MAAI;AACJ,WAAS,UAAU,GAAG,UAAU,iBAAiB,WAAW;AAG1D,QAAI,cAAc,SAAS;AACzB,YAAM,IAAI,aAAa,oCAAoC,YAAY;AAAA,IACzE;AAGA,QAAI,UAAU,KAAK,iBAAiB,eAAe,UAAU,GAAG;AAC9D,YAAM,mBAAmB,QAAQ,UAAU,IAAI,MAAM,OAAO,OAAO,CAAC;AAAA,IACtE;AACA,UAAM,aAAa,IAAI,gBAAgB;AACvC,UAAM,gBAAgB,YAAY,YAAY,YAAY;AAC1D,UAAM,gBAAgB,WAAW,MAAM,WAAW,MAAM,GAAG,SAAS;AACpE,UAAM,UAAU,KAAK,IAAI;AACzB,UAAM,cAAc,UAAU,KAAK,KAAK;AACxC,QAAI,uBAAuB;AAC3B,QAAI,MAAM;AACR,YAAM,UAAU,MAAM,UAAU;AAAA,QAC9B,SAAS,cAAc;AAAA,QACvB,OAAO,cAAc;AAAA,QACrB,QAAQ,cAAc;AAAA,QACtB;AAAA,QACA,OAAO,IAAI;AAAA,QACX;AAAA,QACA;AAAA,QACA,cAAc;AAAA,QACd,WAAW;AAAA,QACX,WAAW;AAAA,QACX,gBAAgB;AAAA,QAChB;AAAA,QACA,gBAAgB,CAAC;AAAA,MACnB,CAAC;AAAA,IACH;AAEA,QAAI;AACF,YAAM,MAAM,MAAM,QAAQ,KAAK;AAAA,QAC7B,QAAQ;AAAA,QACR;AAAA,QACA,MAAM,KAAK,UAAU,WAAW;AAAA,QAChC,QAAQ;AAAA,MACV,CAAC;AACD,mBAAa,aAAa;AAC1B,YAAM,kBAAkB,OAAO,gBAAgB,IAAI,OAAO,IAAI;AAE9D,UAAI,CAAC,IAAI,IAAI;AACX,cAAM,OAAO,MAAM,IAAI,KAAK;AAC5B,YAAI,MAAM;AACR,gBAAM,UAAU,MAAM,UAAU;AAAA,YAC9B,SAAS,cAAc;AAAA,YACvB,OAAO,cAAc;AAAA,YACrB,QAAQ,cAAc;AAAA,YACtB;AAAA,YACA,OAAO,IAAI;AAAA,YACX;AAAA,YACA;AAAA,YACA,cAAc;AAAA,YACd,WAAW;AAAA,YACX,WAAW,KAAK,IAAI;AAAA,YACpB,YAAY,KAAK,IAAI,IAAI;AAAA,YACzB,YAAY,IAAI;AAAA,YAChB;AAAA,YACA,cAAc;AAAA,YACd,cAAc,QAAQ,IAAI,MAAM;AAAA,YAChC,gBAAgB,CAAC;AAAA,UACnB,CAAC;AACD,iCAAuB;AAAA,QACzB;AACA,cAAM,MAAM,IAAI;AAAA,UACd,YAAY,IAAI,MAAM,KAAK,KAAK,MAAM,GAAG,GAAG,CAAC;AAAA,UAC7C,IAAI;AAAA,UACJ;AAAA,UACA,IAAI;AAAA,QACN;AACA,YACE,iBAAiB,IAAI,IAAI,MAAM,KAC/B,UAAU,kBAAkB,KAC5B,CAAC,iBAAiB,eAAe,UAAU,GAC3C;AACA,oBAAU;AACV,gBAAM,aAAa,gBAAgB,IAAI,OAAO;AAC9C,gBAAM,MAAM,cAAc,UAAU,OAAO,CAAC;AAC5C;AAAA,QACF;AACA,cAAM;AAAA,MACR;AAEA,YAAM,OAAO,MAAM,IAAI,KAAK;AAC5B,UAAI;AACJ,UAAI;AACF,eAAO,KAAK,MAAM,IAAI;AAAA,MACxB,SAAS,UAAU;AACjB,YAAI,MAAM;AACR,gBAAM,UAAU,MAAM,UAAU;AAAA,YAC9B,SAAS,cAAc;AAAA,YACvB,OAAO,cAAc;AAAA,YACrB,QAAQ,cAAc;AAAA,YACtB;AAAA,YACA,OAAO,IAAI;AAAA,YACX;AAAA,YACA;AAAA,YACA,cAAc;AAAA,YACd,WAAW;AAAA,YACX,WAAW,KAAK,IAAI;AAAA,YACpB,YAAY,KAAK,IAAI,IAAI;AAAA,YACzB,YAAY,IAAI;AAAA,YAChB;AAAA,YACA,cAAc;AAAA,YACd,cAAc,sBAAsB,oBAAoB,QAAQ,SAAS,UAAU,OAAO,QAAQ,CAAC;AAAA,YACnG,gBAAgB,CAAC;AAAA,UACnB,CAAC;AACD,iCAAuB;AAAA,QACzB;AACA,cAAM;AAAA,MACR;AACA,UAAI,MAAM;AACR,cAAM,UAAU,MAAM,UAAU;AAAA,UAC9B,SAAS,cAAc;AAAA,UACvB,OAAO,cAAc;AAAA,UACrB,QAAQ,cAAc;AAAA,UACtB;AAAA,UACA,OAAO,IAAI;AAAA,UACX;AAAA,UACA;AAAA,UACA,cAAc;AAAA,UACd,WAAW;AAAA,UACX,WAAW,KAAK,IAAI;AAAA,UACpB,YAAY,KAAK,IAAI,IAAI;AAAA,UACzB,YAAY,IAAI;AAAA,UAChB;AAAA,UACA,cAAc;AAAA,UACd,gBAAgB,CAAC;AAAA,QACnB,CAAC;AAAA,MACH;AACA,YAAM,SACJ,KAAK,UAGH,CAAC;AACL,YAAM,WACJ,KAAK,SAAS,OAAO,KAAK,UAAU,YAAY,CAAC,MAAM,QAAQ,KAAK,KAAK,IACpE,KAAK,QACN;AACN,YAAM,eAAe,mBAAmB,UAAU,aAAa;AAC/D,YAAM,mBAAmB,mBAAmB,UAAU,iBAAiB;AACvE,YAAM,cAAc,mBAAmB,UAAU,YAAY;AAC7D,YAAM,YACJ,UAAU,yBACV,OAAO,SAAS,0BAA0B,YAC1C,CAAC,MAAM,QAAQ,SAAS,qBAAqB,IACxC,SAAS,sBAAkD,gBAC5D;AACN,YAAM,qBAAqB,cAAc,SAAY,SAAY,mBAAmB,SAAS;AAC7F,YAAM,gBACJ,iBAAiB,UACjB,qBAAqB,WACpB,cAAc,UACZ,uBAAuB,UAAa,sBAAsB,kBAC5D,gBAAgB,UAAa,gBAAgB,eAAe;AAC/D,YAAM,gBAAiB,KAAK,kBAAkB,KAAK;AACnD,YAAM,UAAU,QAAQ,SAAS,WAAW;AAE5C,aAAO;AAAA,QACL;AAAA,QACA,cAAc,QAAQ,iBAAiB;AAAA,QACvC,cAAc,QAAQ,KAAK,EAAE,WAAW;AAAA,QACxC,OAAO;AAAA,UACL,cAAc,gBAAgB;AAAA,UAC9B,kBAAkB,oBAAoB;AAAA,UACtC,aAAa,gBAAgB,gBAAgB,MAAM,oBAAoB;AAAA,UACvE,UAAU;AAAA,UACV;AAAA,QACF;AAAA,QACA,SAAS,OAAO,kBAAkB,WAAW,gBAAgB;AAAA,QAC7D,OAAQ,KAAK,SAAoB,IAAI;AAAA,QACrC,YAAY,KAAK,IAAI,IAAI;AAAA,QACzB,KAAK;AAAA,MACP;AAAA,IACF,SAAS,KAAK;AACZ,mBAAa,aAAa;AAC1B,gBAAU;AAIV,UAAI,cAAc,SAAS;AACzB,YAAI,QAAQ,CAAC,sBAAsB;AACjC,gBAAM,UAAU,MAAM,UAAU;AAAA,YAC9B,SAAS,cAAc;AAAA,YACvB,OAAO,cAAc;AAAA,YACrB,QAAQ,cAAc;AAAA,YACtB;AAAA,YACA,OAAO,IAAI;AAAA,YACX;AAAA,YACA;AAAA,YACA,cAAc;AAAA,YACd,WAAW;AAAA,YACX,WAAW,KAAK,IAAI;AAAA,YACpB,YAAY,KAAK,IAAI,IAAI;AAAA,YACzB,cAAc,eAAe,QAAQ,IAAI,UAAU,OAAO,GAAG;AAAA,YAC7D,gBAAgB,CAAC;AAAA,UACnB,CAAC;AAAA,QACH;AACA,cAAM;AAAA,MACR;AACA,UAAI,QAAQ,CAAC,sBAAsB;AAIjC,cAAM,UAAU,MAAM,UAAU;AAAA,UAC9B,SAAS,cAAc;AAAA,UACvB,OAAO,cAAc;AAAA,UACrB,QAAQ,cAAc;AAAA,UACtB;AAAA,UACA,OAAO,IAAI;AAAA,UACX;AAAA,UACA;AAAA,UACA,cAAc;AAAA,UACd,WAAW;AAAA,UACX,WAAW,KAAK,IAAI;AAAA,UACpB,YAAY,KAAK,IAAI,IAAI;AAAA,UACzB,cAAc,eAAe,QAAQ,IAAI,UAAU,OAAO,GAAG;AAAA,UAC7D,gBAAgB,CAAC;AAAA,QACnB,CAAC;AAAA,MACH;AACA,UACE,UAAU,kBAAkB,KAC5B,oBAAoB,GAAG,KACvB,CAAC,iBAAiB,eAAe,UAAU,GAC3C;AACA,cAAM,MAAM,UAAU,OAAO,CAAC;AAC9B;AAAA,MACF;AACA,YAAM;AAAA,IACR;AAAA,EACF;AACA,QAAM,mBAAmB,QAAQ,UAAU,IAAI,MAAM,OAAO,OAAO,CAAC;AACtE;AAEA,eAAe,UACb,MACA,UACA,OACe;AAGf,MAAI;AACF,UAAM,KAAK,OAAO,SAAS,KAAK,CAAC;AAAA,EACnC,QAAQ;AAAA,EAER;AACF;AAEA,SAAS,gBAAgB,GAAoC;AAC3D,QAAM,MAA8B,CAAC;AACrC,IAAE,QAAQ,CAAC,OAAO,QAAQ;AACxB,QAAI,GAAG,IAAI;AAAA,EACb,CAAC;AACD,SAAO;AACT;AAEA,SAAS,gBAAwB;AAC/B,MAAI,OAAO,WAAW,QAAQ,eAAe,WAAY,QAAO,WAAW,OAAO,WAAW;AAC7F,SAAO,GAAG,KAAK,IAAI,EAAE,SAAS,EAAE,CAAC,IAAI,KAAK,OAAO,EAAE,SAAS,EAAE,EAAE,MAAM,GAAG,EAAE,CAAC;AAC9E;AAQA,eAAsB,YACpB,KACA,OAAyB,CAAC,GACoB;AAC9C,QAAM,SAAS,MAAM,kBAAkB,KAAK,IAAI;AAChD,QAAM,QAAQ,gBAAmB,MAAM;AACvC,SAAO,EAAE,OAAO,OAAO;AACzB;AAGA,eAAe,kBACb,KACA,OAAyB,CAAC,GACF;AACxB,MAAI;AACF,WAAO,MAAM,QAAQ,EAAE,GAAG,KAAK,UAAU,IAAI,YAAY,CAAC,IAAI,WAAW,GAAG,IAAI;AAAA,EAClF,SAAS,KAAK;AACZ,QAAI,eAAe,gBAAgB,kBAAkB,IAAI,QAAQ,IAAI,IAAI,KAAK,IAAI,YAAY;AAC5F,YAAM,cAA8B,EAAE,GAAG,KAAK,UAAU,MAAM,YAAY,OAAU;AACpF,aAAO,MAAM,QAAQ,aAAa,IAAI;AAAA,IACxC;AACA,UAAM;AAAA,EACR;AACF;AAEA,SAAS,gBAAmB,QAA0B;AACpD,MAAI;AACF,QAAI,OAAO,iBAAiB,UAAU;AACpC,YAAM,IAAI;AAAA,QACR,8CAA8C,OAAO,KAAK;AAAA,MAC5D;AAAA,IACF;AACA,WAAO,gBAAmB,OAAO,SAAS,OAAO,KAAK;AAAA,EACxD,SAAS,OAAO;AACd,QAAI,iBAAiB,iBAAkB,OAAM;AAC7C,UAAM,QAAQ,iBAAiB,QAAQ,QAAQ,IAAI,MAAM,OAAO,KAAK,CAAC;AACtE,UAAM,IAAI,iBAAiB,MAAM,SAAS,QAAQ,EAAE,MAAM,CAAC;AAAA,EAC7D;AACF;AAEA,SAAS,gBAAmB,SAAiB,OAAkB;AAC7D,QAAM,WAAW,mBAAmB,OAAO;AAC3C,MAAI;AACF,WAAO,KAAK,MAAM,QAAQ;AAAA,EAC5B,SAAS,KAAK;AACZ,UAAM,IAAI;AAAA,MACR,wCAAwC,KAAK,MAC3C,eAAe,QAAQ,IAAI,UAAU,OAAO,GAAG,CACjD;AAAA;AAAA,EAA0B,QAAQ,MAAM,GAAG,GAAG,CAAC;AAAA,IACjD;AAAA,EACF;AACF;AAWO,IAAM,yBAAN,cAAqC,sBAAsB;AAAA,EAChE,YACE,SACgB,QACA,SAChB;AACA,UAAM,OAAO;AAHG;AACA;AAAA,EAGlB;AAAA,EAJkB;AAAA,EACA;AAIpB;AAmCO,SAAS,eAAe,MAAwB,MAA4B,CAAC,GAAS;AAC3F,QAAM,kBAAkB,KAAK,YAAY;AACzC,QAAM,WAAW,KAAK,WAAW,kBAAkB,QAAQ,QAAQ,EAAE;AAErE,MAAI,IAAI,0BAA0B,CAAC,iBAAiB;AAClD,UAAM,IAAI;AAAA,MACR,gGAAgG,gBAAgB;AAAA,MAChH;AAAA,MACA;AAAA,IACF;AAAA,EACF;AAEA,MAAI,IAAI,iBAAiB,KAAK,CAAC,MAAM,SAAS,SAAS,CAAC,CAAC,GAAG;AAC1D,UAAM,IAAI;AAAA,MACR,2BAA2B,OAAO;AAAA,MAClC;AAAA,MACA;AAAA,IACF;AAAA,EACF;AAEA,MAAI,IAAI,mBAAmB,IAAI,gBAAgB,SAAS,GAAG;AACzD,UAAM,KAAK,IAAI,gBAAgB,KAAK,CAAC,MAAM,SAAS,SAAS,CAAC,CAAC;AAC/D,QAAI,CAAC,IAAI;AACP,YAAM,IAAI;AAAA,QACR,2BAA2B,OAAO,gCAAgC,IAAI,gBAAgB,IAAI,eAAe,EAAE,KAAK,IAAI,CAAC;AAAA,QACrH;AAAA,QACA;AAAA,MACF;AAAA,IACF;AAAA,EACF;AAEA,MAAI,IAAI,eAAe,CAAC,KAAK,UAAU,CAAC,KAAK,UAAU,CAAC,KAAK,YAAY;AACvE,UAAM,IAAI;AAAA,MACR;AAAA,MACA;AAAA,MACA;AAAA,IACF;AAAA,EACF;AAEA,MAAI,IAAI,kBAAkB;AACxB,UAAM,SAAS,KAAK,YAAY,oBAAoB,OAAO;AAC3D,QAAI,WAAW,IAAI,kBAAkB;AACnC,YAAM,IAAI;AAAA,QACR,qCAAqC,IAAI,gBAAgB,gBAAgB,OAAO,gBAAgB,MAAM;AAAA,QACtG;AAAA,QACA;AAAA,MACF;AAAA,IACF;AAAA,EACF;AACF;AAEA,SAAS,SAAS,KAAa,SAAmC;AAChE,MAAI,mBAAmB,OAAQ,QAAO,QAAQ,KAAK,GAAG;AACtD,SAAO,IAAI,YAAY,EAAE,WAAW,QAAQ,YAAY,CAAC;AAC3D;AAEA,SAAS,gBAAgB,GAA4B;AACnD,SAAO,aAAa,SAAS,EAAE,SAAS;AAC1C;AAaA,eAAsB,SACpB,OACA,OAAkD,CAAC,GACgB;AACnE,QAAM,QAAQ,KAAK,IAAI;AACvB,MAAI;AACF,UAAM;AAAA,MACJ;AAAA,QACE;AAAA,QACA,UAAU,CAAC,EAAE,MAAM,QAAQ,SAAS,OAAO,CAAC;AAAA,QAC5C,WAAW;AAAA,QACX,WAAW,KAAK,aAAa;AAAA,MAC/B;AAAA,MACA;AAAA,IACF;AACA,WAAO,EAAE,IAAI,MAAM,WAAW,KAAK,IAAI,IAAI,OAAO,OAAO,KAAK;AAAA,EAChE,SAAS,KAAK;AACZ,WAAO;AAAA,MACL,IAAI;AAAA,MACJ,WAAW,KAAK,IAAI,IAAI;AAAA,MACxB,OAAO,eAAe,QAAQ,IAAI,UAAU,OAAO,GAAG;AAAA,IACxD;AAAA,EACF;AACF;AAOO,IAAM,YAAN,MAAgB;AAAA,EACZ;AAAA,EACQ;AAAA,EAEjB,YAAY,OAAyB,CAAC,GAAG;AACvC,SAAK,OAAO;AACZ,SAAK,kBAAkB,uBAAuB,KAAK,UAAU;AAAA,EAC/D;AAAA,EAEA,KAAK,KAAqB,KAAgD;AACxE,UAAM,UAAU,EAAE,GAAG,KAAK,MAAM,GAAG,IAAI;AACvC,WAAO,IAAI,aAAa,kBAAkB,KAAK,OAAO,IAAI,QAAQ,KAAK,OAAO;AAAA,EAChF;AAAA,EAEA,SACE,KACA,KAC8C;AAC9C,WAAO,YAAe,KAAK,EAAE,GAAG,KAAK,MAAM,GAAG,IAAI,CAAC;AAAA,EACrD;AACF;","names":[]}
|
|
@@ -4,8 +4,9 @@ import {
|
|
|
4
4
|
import {
|
|
5
5
|
argHash,
|
|
6
6
|
groupBy,
|
|
7
|
+
hasCapturedToolArgs,
|
|
7
8
|
toolSpans
|
|
8
|
-
} from "./chunk-
|
|
9
|
+
} from "./chunk-LQUTGLOZ.js";
|
|
9
10
|
|
|
10
11
|
// src/failure-taxonomy.ts
|
|
11
12
|
var DEFAULT_RULES = [
|
|
@@ -340,15 +341,30 @@ function classifyFailure(ctx, rules = DEFAULT_RULES) {
|
|
|
340
341
|
async function computeToolUseMetrics(store, runId, options = {}) {
|
|
341
342
|
const tools = await toolSpans(store, runId);
|
|
342
343
|
if (tools.length === 0) {
|
|
343
|
-
return {
|
|
344
|
+
return {
|
|
345
|
+
runId,
|
|
346
|
+
totalCalls: 0,
|
|
347
|
+
callsWithCapturedArgs: 0,
|
|
348
|
+
byTool: {},
|
|
349
|
+
errorRate: 0,
|
|
350
|
+
duplicateRate: 0,
|
|
351
|
+
retryRate: 0
|
|
352
|
+
};
|
|
344
353
|
}
|
|
345
354
|
const byTool = {};
|
|
346
355
|
let totalErrors = 0;
|
|
347
356
|
let totalDuplicates = 0;
|
|
357
|
+
let callsWithCapturedArgs = 0;
|
|
348
358
|
const sortedTools = [...tools].sort((a, b) => a.startedAt - b.startedAt);
|
|
349
359
|
const seenSignatures = /* @__PURE__ */ new Set();
|
|
350
360
|
for (const t of sortedTools) {
|
|
351
|
-
byTool[t.toolName] ??= {
|
|
361
|
+
byTool[t.toolName] ??= {
|
|
362
|
+
calls: 0,
|
|
363
|
+
callsWithCapturedArgs: 0,
|
|
364
|
+
errors: 0,
|
|
365
|
+
avgLatencyMs: 0,
|
|
366
|
+
duplicates: 0
|
|
367
|
+
};
|
|
352
368
|
const stat = byTool[t.toolName];
|
|
353
369
|
stat.calls += 1;
|
|
354
370
|
if (t.status === "error") {
|
|
@@ -356,12 +372,16 @@ async function computeToolUseMetrics(store, runId, options = {}) {
|
|
|
356
372
|
totalErrors += 1;
|
|
357
373
|
}
|
|
358
374
|
if (typeof t.latencyMs === "number") stat.avgLatencyMs += t.latencyMs;
|
|
359
|
-
|
|
360
|
-
|
|
361
|
-
stat.
|
|
362
|
-
|
|
375
|
+
if (hasCapturedToolArgs(t)) {
|
|
376
|
+
callsWithCapturedArgs += 1;
|
|
377
|
+
stat.callsWithCapturedArgs += 1;
|
|
378
|
+
const sig = `${t.toolName}|${argHash(t.args)}`;
|
|
379
|
+
if (seenSignatures.has(sig)) {
|
|
380
|
+
stat.duplicates += 1;
|
|
381
|
+
totalDuplicates += 1;
|
|
382
|
+
}
|
|
383
|
+
seenSignatures.add(sig);
|
|
363
384
|
}
|
|
364
|
-
seenSignatures.add(sig);
|
|
365
385
|
}
|
|
366
386
|
for (const stat of Object.values(byTool)) {
|
|
367
387
|
stat.avgLatencyMs = stat.calls > 0 ? stat.avgLatencyMs / stat.calls : 0;
|
|
@@ -386,9 +406,10 @@ async function computeToolUseMetrics(store, runId, options = {}) {
|
|
|
386
406
|
return {
|
|
387
407
|
runId,
|
|
388
408
|
totalCalls: sortedTools.length,
|
|
409
|
+
callsWithCapturedArgs,
|
|
389
410
|
byTool,
|
|
390
411
|
errorRate: totalErrors / sortedTools.length,
|
|
391
|
-
duplicateRate: totalDuplicates /
|
|
412
|
+
duplicateRate: callsWithCapturedArgs > 0 ? totalDuplicates / callsWithCapturedArgs : 0,
|
|
392
413
|
retryRate,
|
|
393
414
|
selectionAccuracy
|
|
394
415
|
};
|
|
@@ -598,4 +619,4 @@ export {
|
|
|
598
619
|
iqr,
|
|
599
620
|
welchsTTest
|
|
600
621
|
};
|
|
601
|
-
//# sourceMappingURL=chunk-
|
|
622
|
+
//# sourceMappingURL=chunk-ODVOOEWQ.js.map
|
|
@@ -0,0 +1 @@
|
|
|
1
|
+
{"version":3,"sources":["../src/failure-taxonomy.ts","../src/tool-use-metrics.ts","../src/trajectory.ts","../src/baseline.ts"],"sourcesContent":["/**\n * Failure taxonomy — canonical classes + a default classifier.\n *\n * Every failed run should end up in a named class. The classifier here\n * is rule-based (fast, deterministic); an LLM fallback can be added by\n * the consumer for novel cases and trained into the rule base over time.\n *\n * Consumers call `classifyFailure(run, spans, events)` and persist the\n * returned class as `Run.outcome.failureClass`.\n */\n\nimport type { FailureClass, Run, Span, TraceEvent } from './trace/schema'\nimport { FAILURE_CLASSES } from './trace/schema'\n\nexport { FAILURE_CLASSES, type FailureClass }\n\nexport interface FailureContext {\n run: Run\n spans: Span[]\n events: TraceEvent[]\n}\n\nexport interface FailureClassification {\n failureClass: FailureClass\n reason: string\n triggerSpanId?: string\n triggerEventId?: string\n}\n\n/** Ordered rules — first match wins. */\nexport interface FailureRule {\n id: string\n match: (ctx: FailureContext) => {\n failureClass: FailureClass\n reason: string\n triggerSpanId?: string\n triggerEventId?: string\n } | null\n}\n\nexport const DEFAULT_RULES: FailureRule[] = [\n // Outcome already named? Respect it.\n {\n id: 'explicit-outcome',\n match: ({ run }) => {\n const fc = run.outcome?.failureClass\n if (fc && fc !== 'unknown')\n return { failureClass: fc, reason: 'outcome.failureClass set explicitly' }\n return null\n },\n },\n {\n id: 'knowledge-readiness-blocked',\n match: ({ events }) => {\n const event = events.find(\n (e) =>\n e.kind === 'custom' &&\n e.payload.kind === 'readiness_scored' &&\n e.payload.passed === false,\n )\n return event\n ? {\n failureClass: 'knowledge_readiness_blocked',\n reason: 'knowledge readiness report blocked execution',\n triggerEventId: event.eventId,\n }\n : null\n },\n },\n {\n id: 'bad-integration-manifest',\n match: ({ events }) => {\n const event = events.find(\n (e) =>\n e.kind === 'custom' &&\n ((e.payload.kind === 'integration_manifest_validated' && e.payload.valid === false) ||\n (e.payload.kind === 'integration_invoke_failed' &&\n e.payload.code === 'manifest_invalid')),\n )\n return event\n ? {\n failureClass: 'bad_integration_manifest',\n reason: 'integration manifest validation failed before launch',\n triggerEventId: event.eventId,\n }\n : null\n },\n },\n {\n id: 'missing-integration-connection',\n match: ({ events }) => {\n const event = events.find(\n (e) =>\n e.kind === 'custom' &&\n e.payload.kind === 'integration_manifest_resolved' &&\n hasResolutionStatus(e.payload, 'missing_connection'),\n )\n return event\n ? {\n failureClass: 'missing_integration_connection',\n reason: 'required integration connection was missing',\n triggerEventId: event.eventId,\n }\n : null\n },\n },\n {\n id: 'missing-integration-scope',\n match: ({ events }) => {\n const event = events.find(\n (e) =>\n e.kind === 'custom' &&\n ((e.payload.kind === 'integration_manifest_resolved' && hasMissingScopes(e.payload)) ||\n (e.payload.kind === 'integration_invoke_failed' && e.payload.code === 'scope_denied')),\n )\n return event\n ? {\n failureClass: 'missing_integration_scope',\n reason: 'integration grant or connection lacks required scopes',\n triggerEventId: event.eventId,\n }\n : null\n },\n },\n {\n id: 'integration-approval-required',\n match: ({ events }) => {\n const event = events.find(\n (e) =>\n e.kind === 'custom' &&\n ((e.payload.kind === 'integration_invoke' && e.payload.status === 'approval_required') ||\n (e.payload.kind === 'integration_invoke_failed' &&\n e.payload.code === 'approval_required') ||\n e.payload.kind === 'integration_approval_required'),\n )\n return event\n ? {\n failureClass: 'integration_approval_required',\n reason: 'integration write paused for user approval',\n triggerEventId: event.eventId,\n }\n : null\n },\n },\n {\n id: 'integration-auth-expired',\n match: ({ events }) => {\n const event = events.find(\n (e) =>\n e.kind === 'custom' &&\n e.payload.kind === 'integration_invoke_failed' &&\n (e.payload.code === 'auth_expired' ||\n e.payload.code === 'connection_not_active' ||\n e.payload.code === 'capability_expired' ||\n e.payload.status === 'expired'),\n )\n return event\n ? {\n failureClass: 'integration_auth_expired',\n reason: 'integration connection or capability expired',\n triggerEventId: event.eventId,\n }\n : null\n },\n },\n {\n id: 'unsafe-integration-write-denied',\n match: ({ events }) => {\n const event = events.find(\n (e) =>\n e.kind === 'custom' &&\n e.payload.kind === 'integration_invoke_failed' &&\n (e.payload.code === 'unsafe_write_denied' ||\n e.payload.code === 'policy_denied' ||\n e.payload.code === 'action_denied'),\n )\n return event\n ? {\n failureClass: 'unsafe_integration_write_denied',\n reason: 'integration write was denied by policy or capability scope',\n triggerEventId: event.eventId,\n }\n : null\n },\n },\n {\n id: 'integration-provider-failure',\n match: ({ events }) => {\n const event = events.find(\n (e) =>\n e.kind === 'custom' &&\n e.payload.kind === 'integration_invoke_failed' &&\n ![\n 'scope_denied',\n 'approval_required',\n 'auth_expired',\n 'connection_not_active',\n 'capability_expired',\n 'unsafe_write_denied',\n 'policy_denied',\n 'action_denied',\n 'manifest_invalid',\n ].includes(String(e.payload.code)),\n )\n return event\n ? {\n failureClass: 'integration_provider_failure',\n reason: 'integration provider invocation failed',\n triggerEventId: event.eventId,\n }\n : null\n },\n },\n {\n id: 'missing-credentials',\n match: ({ events }) => {\n const event = events.find(\n (e) =>\n e.kind === 'custom' &&\n e.payload.kind === 'knowledge_gap' &&\n e.payload.category === 'credential_or_secret',\n )\n return event\n ? {\n failureClass: 'missing_credentials',\n reason: 'required credential or secret was missing',\n triggerEventId: event.eventId,\n }\n : null\n },\n },\n {\n id: 'bad-retrieval',\n match: ({ run, spans }) => {\n if (run.outcome?.pass !== false) return null\n const retrieval = spans.find(\n (s) =>\n s.kind === 'retrieval' && (s.hits.length === 0 || s.hits.every((hit) => hit.score <= 0)),\n )\n return retrieval\n ? {\n failureClass: 'bad_retrieval',\n reason: 'retrieval returned no useful hits for a failed run',\n triggerSpanId: retrieval.spanId,\n }\n : null\n },\n },\n {\n id: 'insufficient-evidence',\n match: ({ events }) => {\n const event = events.find(\n (e) =>\n e.kind === 'custom' &&\n e.payload.kind === 'knowledge_gap' &&\n e.payload.reason === 'insufficient_evidence',\n )\n return event\n ? {\n failureClass: 'insufficient_evidence',\n reason: 'task proceeded with insufficient supporting evidence',\n triggerEventId: event.eventId,\n }\n : null\n },\n },\n {\n id: 'contradictory-evidence',\n match: ({ events }) => {\n const event = events.find(\n (e) =>\n e.kind === 'custom' &&\n e.payload.kind === 'knowledge_gap' &&\n e.payload.reason === 'contradictory_evidence',\n )\n return event\n ? {\n failureClass: 'contradictory_evidence',\n reason: 'supporting evidence contradicted itself',\n triggerEventId: event.eventId,\n }\n : null\n },\n },\n // Budget breach events\n {\n id: 'budget-breach',\n match: ({ events }) => {\n const breach = events.find((e) => e.kind === 'budget_breach')\n return breach\n ? {\n failureClass: 'budget_exceeded',\n reason: `budget breached on ${breach.payload.dimension ?? 'unknown dimension'}`,\n triggerEventId: breach.eventId,\n }\n : null\n },\n },\n // Policy violations\n {\n id: 'policy-violation',\n match: ({ events }) => {\n const e = events.find((x) => x.kind === 'policy_violation')\n return e\n ? {\n failureClass: 'policy_violation',\n reason: 'policy_violation event emitted',\n triggerEventId: e.eventId,\n }\n : null\n },\n },\n // Sandbox non-zero exit code\n {\n id: 'sandbox-failure',\n match: ({ spans }) => {\n const s = spans.find(\n (x) => x.kind === 'sandbox' && typeof x.exitCode === 'number' && x.exitCode !== 0,\n )\n if (!s) return null\n return {\n failureClass: 'sandbox_failure',\n reason: `sandbox exited ${(s as Extract<Span, { kind: 'sandbox' }>).exitCode}`,\n triggerSpanId: s.spanId,\n }\n },\n },\n // Timeout: run aborted by external signal\n {\n id: 'timeout',\n match: ({ run, events }) => {\n if (run.status !== 'aborted') return null\n const hasTimeout = events.some(\n (e) =>\n e.kind === 'error' &&\n String(e.payload.reason ?? '')\n .toLowerCase()\n .includes('timeout'),\n )\n const note = (run.outcome?.notes ?? '').toLowerCase()\n if (hasTimeout || note.includes('timeout') || note.includes('deadline')) {\n return { failureClass: 'timeout', reason: 'timeout signal observed' }\n }\n return null\n },\n },\n // Tool recovery failure: many consecutive tool errors on the same tool\n {\n id: 'tool-recovery-failure',\n match: ({ spans }) => {\n const tools = spans.filter((s) => s.kind === 'tool')\n const byTool = new Map<string, Span[]>()\n for (const t of tools) {\n const name = (t as Extract<Span, { kind: 'tool' }>).toolName\n const arr = byTool.get(name) ?? []\n arr.push(t)\n byTool.set(name, arr)\n }\n for (const [name, arr] of byTool) {\n const errs = arr.filter((s) => s.status === 'error')\n if (errs.length >= 3 && errs.length === arr.length) {\n return {\n failureClass: 'tool_recovery_failure',\n reason: `${errs.length} consecutive errors on tool \"${name}\"`,\n triggerSpanId: errs[errs.length - 1]!.spanId,\n }\n }\n }\n return null\n },\n },\n // Tool selection error: the run failed and agent called zero tools despite having them\n {\n id: 'tool-selection-error',\n match: ({ run, spans }) => {\n if (run.outcome?.pass !== false) return null\n const hasToolsAvailable = spans.some(\n (s) =>\n s.kind === 'agent' &&\n (s.attributes?.toolsAvailable as number | undefined) !== undefined &&\n (s.attributes?.toolsAvailable as number) > 0,\n )\n const tools = spans.filter((s) => s.kind === 'tool')\n if (hasToolsAvailable && tools.length === 0) {\n return {\n failureClass: 'tool_selection_error',\n reason: 'tools were available but none were called',\n }\n }\n return null\n },\n },\n // Format drift: scored by a judge with dimension='format' below threshold\n {\n id: 'format-drift',\n match: ({ spans }) => {\n const judge = spans.find(\n (s) =>\n s.kind === 'judge' &&\n (s as Extract<Span, { kind: 'judge' }>).dimension === 'format' &&\n (s as Extract<Span, { kind: 'judge' }>).score < 0.5,\n )\n return judge\n ? {\n failureClass: 'format_drift',\n reason: 'format judge scored below 0.5',\n triggerSpanId: judge.spanId,\n }\n : null\n },\n },\n]\n\nfunction hasResolutionStatus(payload: Record<string, unknown>, status: string): boolean {\n if (status === 'missing_connection' && stringArray(payload.missingConnections).length > 0)\n return true\n return resolutionItems(payload).some((item) => item.status === status)\n}\n\nfunction hasMissingScopes(payload: Record<string, unknown>): boolean {\n if (stringArray(payload.missingScopes).length > 0) return true\n return resolutionItems(payload).some(\n (item) => Array.isArray(item.missingScopes) && item.missingScopes.length > 0,\n )\n}\n\nfunction resolutionItems(payload: Record<string, unknown>): Array<Record<string, unknown>> {\n return [\n ...records(payload.missing),\n ...records(payload.optionalMissing),\n ...records(payload.ready),\n ]\n}\n\nfunction records(value: unknown): Array<Record<string, unknown>> {\n if (!Array.isArray(value)) return []\n return value.filter(\n (item): item is Record<string, unknown> =>\n Boolean(item) && typeof item === 'object' && !Array.isArray(item),\n )\n}\n\nfunction stringArray(value: unknown): string[] {\n return Array.isArray(value)\n ? value.filter((item): item is string => typeof item === 'string')\n : []\n}\n\n/** Classify the failure mode of a run using an ordered rule list. */\nexport function classifyFailure(\n ctx: FailureContext,\n rules: FailureRule[] = DEFAULT_RULES,\n): FailureClassification {\n if (ctx.run.outcome?.pass !== false && ctx.run.status === 'completed') {\n return { failureClass: 'success', reason: 'run completed with pass=true (or no explicit fail)' }\n }\n for (const rule of rules) {\n const hit = rule.match(ctx)\n if (hit) return hit\n }\n return { failureClass: 'unknown', reason: 'no rule matched; run failed for unclassified reason' }\n}\n","/**\n * Tool-use metrics — derived purely from trace data.\n *\n * No scoring assumptions: consumers supply optional ground-truth tool\n * selections per turn + optional \"information used downstream\" signals.\n * Without those, we still compute descriptive metrics (error rate,\n * retry rate, duplicate-call rate) that are useful on their own.\n */\n\nimport { argHash, groupBy, hasCapturedToolArgs, toolSpans } from './trace/query'\nimport type { Span } from './trace/schema'\nimport type { TraceStore } from './trace/store'\n\nexport interface ToolUseMetrics {\n runId: string\n totalCalls: number\n /** Calls whose arguments were captured and can be compared for duplication. */\n callsWithCapturedArgs: number\n byTool: Record<string, ToolStats>\n errorRate: number\n /** Ratio of captured-argument calls already seen with the same tool name and arguments. */\n duplicateRate: number\n /** Ratio of error calls followed by ≥1 retry on same tool. */\n retryRate: number\n /** Optional: of the calls agent made, fraction the evaluator marked as \"correct selection\". */\n selectionAccuracy?: number\n}\n\nexport interface ToolStats {\n calls: number\n callsWithCapturedArgs: number\n errors: number\n avgLatencyMs: number\n duplicates: number\n}\n\nexport interface ToolUseOptions {\n /** Map of spanId → whether the evaluator judged the tool selection correct. Optional. */\n selectionLabels?: Record<string, boolean>\n}\n\nexport async function computeToolUseMetrics(\n store: TraceStore,\n runId: string,\n options: ToolUseOptions = {},\n): Promise<ToolUseMetrics> {\n const tools = await toolSpans(store, runId)\n if (tools.length === 0) {\n return {\n runId,\n totalCalls: 0,\n callsWithCapturedArgs: 0,\n byTool: {},\n errorRate: 0,\n duplicateRate: 0,\n retryRate: 0,\n }\n }\n\n const byTool: Record<string, ToolStats> = {}\n let totalErrors = 0\n let totalDuplicates = 0\n let callsWithCapturedArgs = 0\n const sortedTools = [...tools].sort((a, b) => a.startedAt - b.startedAt)\n const seenSignatures = new Set<string>()\n\n // duplicate detection + per-tool aggregation\n for (const t of sortedTools) {\n byTool[t.toolName] ??= {\n calls: 0,\n callsWithCapturedArgs: 0,\n errors: 0,\n avgLatencyMs: 0,\n duplicates: 0,\n }\n const stat = byTool[t.toolName]!\n stat.calls += 1\n if (t.status === 'error') {\n stat.errors += 1\n totalErrors += 1\n }\n if (typeof t.latencyMs === 'number') stat.avgLatencyMs += t.latencyMs\n if (hasCapturedToolArgs(t)) {\n callsWithCapturedArgs += 1\n stat.callsWithCapturedArgs += 1\n const sig = `${t.toolName}|${argHash(t.args)}`\n if (seenSignatures.has(sig)) {\n stat.duplicates += 1\n totalDuplicates += 1\n }\n seenSignatures.add(sig)\n }\n }\n\n for (const stat of Object.values(byTool)) {\n stat.avgLatencyMs = stat.calls > 0 ? stat.avgLatencyMs / stat.calls : 0\n }\n\n // retry detection: per-tool chronological adjacency where error → next same-tool call\n let retryOpportunities = 0\n let retriesFollowed = 0\n for (const [, arr] of groupBy(sortedTools, (t) => t.toolName)) {\n for (let i = 0; i < arr.length; i++) {\n if (arr[i]!.status !== 'error') continue\n retryOpportunities += 1\n if (arr[i + 1]) retriesFollowed += 1\n }\n }\n const retryRate = retryOpportunities > 0 ? retriesFollowed / retryOpportunities : 0\n\n let selectionAccuracy: number | undefined\n if (options.selectionLabels) {\n const labeled = sortedTools.filter((t) => t.spanId in options.selectionLabels!)\n if (labeled.length > 0) {\n selectionAccuracy =\n labeled.filter((t) => options.selectionLabels![t.spanId]).length / labeled.length\n }\n }\n\n return {\n runId,\n totalCalls: sortedTools.length,\n callsWithCapturedArgs,\n byTool,\n errorRate: totalErrors / sortedTools.length,\n duplicateRate: callsWithCapturedArgs > 0 ? totalDuplicates / callsWithCapturedArgs : 0,\n retryRate,\n selectionAccuracy,\n }\n}\n\nexport type { Span }\n","/**\n * Trajectory — ordered, structured view over a run's spans.\n *\n * A pure function `buildTrajectory(store, runId) → Trajectory` returns\n * a topologically ordered list of `TrajectoryStep` with parent-child\n * grouping collapsed into a single line-of-agent-work. Separate\n * analyzers (stuck-loop detection, waste ratio) live in\n * `pipelines/` and consume the trajectory.\n */\n\nimport type { Span, TraceEvent } from './trace/schema'\nimport type { TraceStore } from './trace/store'\n\nexport interface TrajectoryStep {\n index: number\n span: Span\n /** Depth in the span tree from the root. 0 = top-level. */\n depth: number\n /** Events attached to this span. */\n events: TraceEvent[]\n}\n\nexport interface Trajectory {\n runId: string\n steps: TrajectoryStep[]\n llmTurns: number\n toolCalls: number\n judgeVerdicts: number\n retrievals: number\n totalDurationMs: number\n}\n\nexport async function buildTrajectory(store: TraceStore, runId: string): Promise<Trajectory> {\n const spans = await store.spans({ runId })\n const events = await store.events({ runId })\n const childrenOf = new Map<string | undefined, Span[]>()\n for (const s of spans) {\n const arr = childrenOf.get(s.parentSpanId) ?? []\n arr.push(s)\n childrenOf.set(s.parentSpanId, arr)\n }\n // Sort children by startedAt so DFS yields chronological order within siblings.\n for (const arr of childrenOf.values()) arr.sort((a, b) => a.startedAt - b.startedAt)\n\n const eventsBySpan = new Map<string, TraceEvent[]>()\n for (const e of events) {\n if (!e.spanId) continue\n const arr = eventsBySpan.get(e.spanId) ?? []\n arr.push(e)\n eventsBySpan.set(e.spanId, arr)\n }\n\n const steps: TrajectoryStep[] = []\n const walk = (spanId: string | undefined, depth: number): void => {\n const kids = childrenOf.get(spanId) ?? []\n for (const child of kids) {\n steps.push({\n index: steps.length,\n span: child,\n depth,\n events: eventsBySpan.get(child.spanId) ?? [],\n })\n walk(child.spanId, depth + 1)\n }\n }\n walk(undefined, 0)\n\n const llmTurns = steps.filter((s) => s.span.kind === 'llm').length\n const toolCalls = steps.filter((s) => s.span.kind === 'tool').length\n const judgeVerdicts = steps.filter((s) => s.span.kind === 'judge').length\n const retrievals = steps.filter((s) => s.span.kind === 'retrieval').length\n\n let totalDurationMs = 0\n if (steps.length > 0) {\n const starts = spans.map((s) => s.startedAt)\n const ends = spans.map((s) => s.endedAt ?? s.startedAt)\n totalDurationMs = Math.max(...ends) - Math.min(...starts)\n }\n\n return { runId, steps, llmTurns, toolCalls, judgeVerdicts, retrievals, totalDurationMs }\n}\n\n// Re-export core types for convenience so consumers don't import from two paths.\nexport type { Span, TraceEvent } from './trace/schema'\nexport type { TraceStore } from './trace/store'\n","/**\n * Baseline regression detection.\n *\n * Lifted from ADC baseline.ts. Every promotion-blocking signal boils down\n * to: \"is this run measurably worse than baseline?\" — with enough\n * statistical rigor to distinguish noise from drift.\n *\n * Uses:\n * - Welch's t-test (unequal variance) for per-metric mean comparison\n * - Cohen's d for effect size magnitude\n * - IQR for stability flag (unstable samples can't be trusted for comparisons)\n *\n * Returns a structured verdict: improved | regressed | stable | unstable.\n */\n\nimport { cohensD } from './statistics'\n\nexport interface MetricSamples {\n /** Stable metric key (e.g. \"overallScore\", \"firstTokenMs\"). */\n metric: string\n /** Whether higher values are better. */\n higherIsBetter: boolean\n baseline: number[]\n candidate: number[]\n}\n\nexport interface MetricVerdict {\n metric: string\n baselineMean: number\n candidateMean: number\n delta: number\n cohensD: number\n welchT: number\n welchDf: number\n welchP: number\n stable: boolean\n /** IQR of the combined samples — used as a rough stability indicator. */\n iqr: number\n verdict: 'improved' | 'regressed' | 'stable' | 'unstable'\n}\n\nexport interface BaselineReport {\n metrics: MetricVerdict[]\n /** True if any critical metric regressed. */\n hasRegression: boolean\n /** True if any metric is unstable (too noisy to judge). */\n hasUnstable: boolean\n}\n\nexport interface BaselineOptions {\n /** Effect size threshold for meaningful delta (default 0.5 — medium effect). */\n effectThreshold?: number\n /** p-value threshold for statistical significance (default 0.05). */\n alpha?: number\n /** IQR/mean ratio above which samples are flagged unstable (default 0.30). */\n unstableCvThreshold?: number\n}\n\n/**\n * Compare candidate samples against baseline per metric. Verdict logic:\n * - unstable: IQR/|mean| > threshold on either set — not enough signal\n * - improved: meaningful effect in the \"better\" direction AND p < alpha\n * - regressed: meaningful effect in the \"worse\" direction AND p < alpha\n * - stable: otherwise (no significant change)\n */\nexport function compareToBaseline(\n samples: MetricSamples[],\n options: BaselineOptions = {},\n): BaselineReport {\n const effectThreshold = options.effectThreshold ?? 0.5\n const alpha = options.alpha ?? 0.05\n const cvThreshold = options.unstableCvThreshold ?? 0.3\n\n const metrics: MetricVerdict[] = samples.map((s) => {\n if (s.baseline.length < 2 || s.candidate.length < 2) {\n throw new Error(`compareToBaseline: need ≥2 samples per side for \"${s.metric}\"`)\n }\n const bMean = mean(s.baseline)\n const cMean = mean(s.candidate)\n const delta = cMean - bMean\n const d = cohensD(s.baseline, s.candidate) // positive = candidate higher\n const { t, df, p } = welchsTTest(s.baseline, s.candidate)\n // Stability is per-side: a comparison is trustworthy only when BOTH\n // samples are internally consistent. Combining the sides would flag\n // large-but-real deltas as \"unstable\" which is exactly what we want\n // to detect.\n const baselineIqr = iqr(s.baseline)\n const candidateIqr = iqr(s.candidate)\n const baselineStable = baselineIqr / Math.max(Math.abs(bMean), 1e-9) <= cvThreshold\n const candidateStable = candidateIqr / Math.max(Math.abs(cMean), 1e-9) <= cvThreshold\n const stable = baselineStable && candidateStable\n const reportedIqr = Math.max(baselineIqr, candidateIqr)\n\n let verdict: MetricVerdict['verdict']\n if (!stable) {\n verdict = 'unstable'\n } else if (p < alpha && Math.abs(d) >= effectThreshold) {\n const candidateIsBetter = s.higherIsBetter ? delta > 0 : delta < 0\n verdict = candidateIsBetter ? 'improved' : 'regressed'\n } else {\n verdict = 'stable'\n }\n\n return {\n metric: s.metric,\n baselineMean: bMean,\n candidateMean: cMean,\n delta,\n cohensD: d,\n welchT: t,\n welchDf: df,\n welchP: p,\n stable,\n iqr: reportedIqr,\n verdict,\n }\n })\n\n return {\n metrics,\n hasRegression: metrics.some((m) => m.verdict === 'regressed'),\n hasUnstable: metrics.some((m) => m.verdict === 'unstable'),\n }\n}\n\nfunction mean(xs: number[]): number {\n return xs.reduce((a, b) => a + b, 0) / xs.length\n}\n\n/** Inter-quartile range; 0 when the sample has no spread. */\nexport function iqr(xs: number[]): number {\n if (xs.length === 0) return 0\n const sorted = [...xs].sort((a, b) => a - b)\n const q = (p: number) => {\n const idx = p * (sorted.length - 1)\n const lo = Math.floor(idx)\n const hi = Math.ceil(idx)\n return sorted[lo]! + (sorted[hi]! - sorted[lo]!) * (idx - lo)\n }\n return q(0.75) - q(0.25)\n}\n\n/**\n * Welch's t-test — unequal-variance two-sample t. Uses the same Student-t\n * CDF as `pairedTTest` (via incomplete beta); falls back to normal tail\n * when df is large.\n */\nexport function welchsTTest(a: number[], b: number[]): { t: number; df: number; p: number } {\n if (a.length < 2 || b.length < 2) return { t: 0, df: 0, p: 1 }\n const mA = mean(a)\n const mB = mean(b)\n const vA = variance(a, mA)\n const vB = variance(b, mB)\n const seSquared = vA / a.length + vB / b.length\n if (seSquared === 0) return { t: mA === mB ? 0 : Infinity, df: 0, p: mA === mB ? 1 : 0 }\n const t = (mB - mA) / Math.sqrt(seSquared)\n const df =\n (seSquared * seSquared) /\n ((vA / a.length) ** 2 / (a.length - 1) + (vB / b.length) ** 2 / (b.length - 1))\n const p = 2 * (1 - studentTCdf(Math.abs(t), df))\n return { t, df, p }\n}\n\nfunction variance(xs: number[], m: number): number {\n return xs.reduce((acc, x) => acc + (x - m) ** 2, 0) / (xs.length - 1)\n}\n\n// Re-used from statistics.ts via small local copy to avoid exporting internals.\nfunction studentTCdf(t: number, df: number): number {\n if (df <= 0) return 0.5\n if (df > 100) return normalCdf(t)\n const x = df / (df + t * t)\n const ib = incompleteBeta(x, df / 2, 0.5)\n return t >= 0 ? 1 - 0.5 * ib : 0.5 * ib\n}\n\nfunction incompleteBeta(x: number, a: number, b: number): number {\n if (x <= 0) return 0\n if (x >= 1) return 1\n const lnBeta = lnGamma(a) + lnGamma(b) - lnGamma(a + b)\n const front = Math.exp(Math.log(x) * a + Math.log(1 - x) * b - lnBeta) / a\n let c = 1\n let d = 1 - ((a + b) * x) / (a + 1)\n if (Math.abs(d) < 1e-30) d = 1e-30\n d = 1 / d\n let f = d\n for (let m = 1; m <= 200; m++) {\n const m2 = 2 * m\n let num = (m * (b - m) * x) / ((a + m2 - 1) * (a + m2))\n d = 1 + num * d\n if (Math.abs(d) < 1e-30) d = 1e-30\n c = 1 + num / c\n if (Math.abs(c) < 1e-30) c = 1e-30\n d = 1 / d\n f *= d * c\n num = -((a + m) * (a + b + m) * x) / ((a + m2) * (a + m2 + 1))\n d = 1 + num * d\n if (Math.abs(d) < 1e-30) d = 1e-30\n c = 1 + num / c\n if (Math.abs(c) < 1e-30) c = 1e-30\n d = 1 / d\n const delta = d * c\n f *= delta\n if (Math.abs(delta - 1) < 3e-7) break\n }\n return front * f\n}\n\nfunction lnGamma(z: number): number {\n const coefs = [\n 0.99999999999980993, 676.5203681218851, -1259.1392167224028, 771.32342877765313,\n -176.61502916214059, 12.507343278686905, -0.13857109526572012, 9.9843695780195716e-6,\n 1.5056327351493116e-7,\n ]\n if (z < 0.5) return Math.log(Math.PI / Math.sin(Math.PI * z)) - lnGamma(1 - z)\n z -= 1\n let x = coefs[0]!\n for (let i = 1; i < 9; i++) x += coefs[i]! / (z + i)\n const t = z + 7.5\n return 0.5 * Math.log(2 * Math.PI) + (z + 0.5) * Math.log(t) - t + Math.log(x)\n}\n\nfunction normalCdf(x: number): number {\n const a1 = 0.254829592\n const a2 = -0.284496736\n const a3 = 1.421413741\n const a4 = -1.453152027\n const a5 = 1.061405429\n const p = 0.3275911\n const sign = x < 0 ? -1 : 1\n const absX = Math.abs(x)\n const t = 1 / (1 + p * absX)\n const y = 1 - ((((a5 * t + a4) * t + a3) * t + a2) * t + a1) * t * Math.exp((-absX * absX) / 2)\n return 0.5 * (1 + sign * y)\n}\n"],"mappings":";;;;;;;;;;;AAwCO,IAAM,gBAA+B;AAAA;AAAA,EAE1C;AAAA,IACE,IAAI;AAAA,IACJ,OAAO,CAAC,EAAE,IAAI,MAAM;AAClB,YAAM,KAAK,IAAI,SAAS;AACxB,UAAI,MAAM,OAAO;AACf,eAAO,EAAE,cAAc,IAAI,QAAQ,sCAAsC;AAC3E,aAAO;AAAA,IACT;AAAA,EACF;AAAA,EACA;AAAA,IACE,IAAI;AAAA,IACJ,OAAO,CAAC,EAAE,OAAO,MAAM;AACrB,YAAM,QAAQ,OAAO;AAAA,QACnB,CAAC,MACC,EAAE,SAAS,YACX,EAAE,QAAQ,SAAS,sBACnB,EAAE,QAAQ,WAAW;AAAA,MACzB;AACA,aAAO,QACH;AAAA,QACE,cAAc;AAAA,QACd,QAAQ;AAAA,QACR,gBAAgB,MAAM;AAAA,MACxB,IACA;AAAA,IACN;AAAA,EACF;AAAA,EACA;AAAA,IACE,IAAI;AAAA,IACJ,OAAO,CAAC,EAAE,OAAO,MAAM;AACrB,YAAM,QAAQ,OAAO;AAAA,QACnB,CAAC,MACC,EAAE,SAAS,aACT,EAAE,QAAQ,SAAS,oCAAoC,EAAE,QAAQ,UAAU,SAC1E,EAAE,QAAQ,SAAS,+BAClB,EAAE,QAAQ,SAAS;AAAA,MAC3B;AACA,aAAO,QACH;AAAA,QACE,cAAc;AAAA,QACd,QAAQ;AAAA,QACR,gBAAgB,MAAM;AAAA,MACxB,IACA;AAAA,IACN;AAAA,EACF;AAAA,EACA;AAAA,IACE,IAAI;AAAA,IACJ,OAAO,CAAC,EAAE,OAAO,MAAM;AACrB,YAAM,QAAQ,OAAO;AAAA,QACnB,CAAC,MACC,EAAE,SAAS,YACX,EAAE,QAAQ,SAAS,mCACnB,oBAAoB,EAAE,SAAS,oBAAoB;AAAA,MACvD;AACA,aAAO,QACH;AAAA,QACE,cAAc;AAAA,QACd,QAAQ;AAAA,QACR,gBAAgB,MAAM;AAAA,MACxB,IACA;AAAA,IACN;AAAA,EACF;AAAA,EACA;AAAA,IACE,IAAI;AAAA,IACJ,OAAO,CAAC,EAAE,OAAO,MAAM;AACrB,YAAM,QAAQ,OAAO;AAAA,QACnB,CAAC,MACC,EAAE,SAAS,aACT,EAAE,QAAQ,SAAS,mCAAmC,iBAAiB,EAAE,OAAO,KAC/E,EAAE,QAAQ,SAAS,+BAA+B,EAAE,QAAQ,SAAS;AAAA,MAC5E;AACA,aAAO,QACH;AAAA,QACE,cAAc;AAAA,QACd,QAAQ;AAAA,QACR,gBAAgB,MAAM;AAAA,MACxB,IACA;AAAA,IACN;AAAA,EACF;AAAA,EACA;AAAA,IACE,IAAI;AAAA,IACJ,OAAO,CAAC,EAAE,OAAO,MAAM;AACrB,YAAM,QAAQ,OAAO;AAAA,QACnB,CAAC,MACC,EAAE,SAAS,aACT,EAAE,QAAQ,SAAS,wBAAwB,EAAE,QAAQ,WAAW,uBAC/D,EAAE,QAAQ,SAAS,+BAClB,EAAE,QAAQ,SAAS,uBACrB,EAAE,QAAQ,SAAS;AAAA,MACzB;AACA,aAAO,QACH;AAAA,QACE,cAAc;AAAA,QACd,QAAQ;AAAA,QACR,gBAAgB,MAAM;AAAA,MACxB,IACA;AAAA,IACN;AAAA,EACF;AAAA,EACA;AAAA,IACE,IAAI;AAAA,IACJ,OAAO,CAAC,EAAE,OAAO,MAAM;AACrB,YAAM,QAAQ,OAAO;AAAA,QACnB,CAAC,MACC,EAAE,SAAS,YACX,EAAE,QAAQ,SAAS,gCAClB,EAAE,QAAQ,SAAS,kBAClB,EAAE,QAAQ,SAAS,2BACnB,EAAE,QAAQ,SAAS,wBACnB,EAAE,QAAQ,WAAW;AAAA,MAC3B;AACA,aAAO,QACH;AAAA,QACE,cAAc;AAAA,QACd,QAAQ;AAAA,QACR,gBAAgB,MAAM;AAAA,MACxB,IACA;AAAA,IACN;AAAA,EACF;AAAA,EACA;AAAA,IACE,IAAI;AAAA,IACJ,OAAO,CAAC,EAAE,OAAO,MAAM;AACrB,YAAM,QAAQ,OAAO;AAAA,QACnB,CAAC,MACC,EAAE,SAAS,YACX,EAAE,QAAQ,SAAS,gCAClB,EAAE,QAAQ,SAAS,yBAClB,EAAE,QAAQ,SAAS,mBACnB,EAAE,QAAQ,SAAS;AAAA,MACzB;AACA,aAAO,QACH;AAAA,QACE,cAAc;AAAA,QACd,QAAQ;AAAA,QACR,gBAAgB,MAAM;AAAA,MACxB,IACA;AAAA,IACN;AAAA,EACF;AAAA,EACA;AAAA,IACE,IAAI;AAAA,IACJ,OAAO,CAAC,EAAE,OAAO,MAAM;AACrB,YAAM,QAAQ,OAAO;AAAA,QACnB,CAAC,MACC,EAAE,SAAS,YACX,EAAE,QAAQ,SAAS,+BACnB,CAAC;AAAA,UACC;AAAA,UACA;AAAA,UACA;AAAA,UACA;AAAA,UACA;AAAA,UACA;AAAA,UACA;AAAA,UACA;AAAA,UACA;AAAA,QACF,EAAE,SAAS,OAAO,EAAE,QAAQ,IAAI,CAAC;AAAA,MACrC;AACA,aAAO,QACH;AAAA,QACE,cAAc;AAAA,QACd,QAAQ;AAAA,QACR,gBAAgB,MAAM;AAAA,MACxB,IACA;AAAA,IACN;AAAA,EACF;AAAA,EACA;AAAA,IACE,IAAI;AAAA,IACJ,OAAO,CAAC,EAAE,OAAO,MAAM;AACrB,YAAM,QAAQ,OAAO;AAAA,QACnB,CAAC,MACC,EAAE,SAAS,YACX,EAAE,QAAQ,SAAS,mBACnB,EAAE,QAAQ,aAAa;AAAA,MAC3B;AACA,aAAO,QACH;AAAA,QACE,cAAc;AAAA,QACd,QAAQ;AAAA,QACR,gBAAgB,MAAM;AAAA,MACxB,IACA;AAAA,IACN;AAAA,EACF;AAAA,EACA;AAAA,IACE,IAAI;AAAA,IACJ,OAAO,CAAC,EAAE,KAAK,MAAM,MAAM;AACzB,UAAI,IAAI,SAAS,SAAS,MAAO,QAAO;AACxC,YAAM,YAAY,MAAM;AAAA,QACtB,CAAC,MACC,EAAE,SAAS,gBAAgB,EAAE,KAAK,WAAW,KAAK,EAAE,KAAK,MAAM,CAAC,QAAQ,IAAI,SAAS,CAAC;AAAA,MAC1F;AACA,aAAO,YACH;AAAA,QACE,cAAc;AAAA,QACd,QAAQ;AAAA,QACR,eAAe,UAAU;AAAA,MAC3B,IACA;AAAA,IACN;AAAA,EACF;AAAA,EACA;AAAA,IACE,IAAI;AAAA,IACJ,OAAO,CAAC,EAAE,OAAO,MAAM;AACrB,YAAM,QAAQ,OAAO;AAAA,QACnB,CAAC,MACC,EAAE,SAAS,YACX,EAAE,QAAQ,SAAS,mBACnB,EAAE,QAAQ,WAAW;AAAA,MACzB;AACA,aAAO,QACH;AAAA,QACE,cAAc;AAAA,QACd,QAAQ;AAAA,QACR,gBAAgB,MAAM;AAAA,MACxB,IACA;AAAA,IACN;AAAA,EACF;AAAA,EACA;AAAA,IACE,IAAI;AAAA,IACJ,OAAO,CAAC,EAAE,OAAO,MAAM;AACrB,YAAM,QAAQ,OAAO;AAAA,QACnB,CAAC,MACC,EAAE,SAAS,YACX,EAAE,QAAQ,SAAS,mBACnB,EAAE,QAAQ,WAAW;AAAA,MACzB;AACA,aAAO,QACH;AAAA,QACE,cAAc;AAAA,QACd,QAAQ;AAAA,QACR,gBAAgB,MAAM;AAAA,MACxB,IACA;AAAA,IACN;AAAA,EACF;AAAA;AAAA,EAEA;AAAA,IACE,IAAI;AAAA,IACJ,OAAO,CAAC,EAAE,OAAO,MAAM;AACrB,YAAM,SAAS,OAAO,KAAK,CAAC,MAAM,EAAE,SAAS,eAAe;AAC5D,aAAO,SACH;AAAA,QACE,cAAc;AAAA,QACd,QAAQ,sBAAsB,OAAO,QAAQ,aAAa,mBAAmB;AAAA,QAC7E,gBAAgB,OAAO;AAAA,MACzB,IACA;AAAA,IACN;AAAA,EACF;AAAA;AAAA,EAEA;AAAA,IACE,IAAI;AAAA,IACJ,OAAO,CAAC,EAAE,OAAO,MAAM;AACrB,YAAM,IAAI,OAAO,KAAK,CAAC,MAAM,EAAE,SAAS,kBAAkB;AAC1D,aAAO,IACH;AAAA,QACE,cAAc;AAAA,QACd,QAAQ;AAAA,QACR,gBAAgB,EAAE;AAAA,MACpB,IACA;AAAA,IACN;AAAA,EACF;AAAA;AAAA,EAEA;AAAA,IACE,IAAI;AAAA,IACJ,OAAO,CAAC,EAAE,MAAM,MAAM;AACpB,YAAM,IAAI,MAAM;AAAA,QACd,CAAC,MAAM,EAAE,SAAS,aAAa,OAAO,EAAE,aAAa,YAAY,EAAE,aAAa;AAAA,MAClF;AACA,UAAI,CAAC,EAAG,QAAO;AACf,aAAO;AAAA,QACL,cAAc;AAAA,QACd,QAAQ,kBAAmB,EAAyC,QAAQ;AAAA,QAC5E,eAAe,EAAE;AAAA,MACnB;AAAA,IACF;AAAA,EACF;AAAA;AAAA,EAEA;AAAA,IACE,IAAI;AAAA,IACJ,OAAO,CAAC,EAAE,KAAK,OAAO,MAAM;AAC1B,UAAI,IAAI,WAAW,UAAW,QAAO;AACrC,YAAM,aAAa,OAAO;AAAA,QACxB,CAAC,MACC,EAAE,SAAS,WACX,OAAO,EAAE,QAAQ,UAAU,EAAE,EAC1B,YAAY,EACZ,SAAS,SAAS;AAAA,MACzB;AACA,YAAM,QAAQ,IAAI,SAAS,SAAS,IAAI,YAAY;AACpD,UAAI,cAAc,KAAK,SAAS,SAAS,KAAK,KAAK,SAAS,UAAU,GAAG;AACvE,eAAO,EAAE,cAAc,WAAW,QAAQ,0BAA0B;AAAA,MACtE;AACA,aAAO;AAAA,IACT;AAAA,EACF;AAAA;AAAA,EAEA;AAAA,IACE,IAAI;AAAA,IACJ,OAAO,CAAC,EAAE,MAAM,MAAM;AACpB,YAAM,QAAQ,MAAM,OAAO,CAAC,MAAM,EAAE,SAAS,MAAM;AACnD,YAAM,SAAS,oBAAI,IAAoB;AACvC,iBAAW,KAAK,OAAO;AACrB,cAAM,OAAQ,EAAsC;AACpD,cAAM,MAAM,OAAO,IAAI,IAAI,KAAK,CAAC;AACjC,YAAI,KAAK,CAAC;AACV,eAAO,IAAI,MAAM,GAAG;AAAA,MACtB;AACA,iBAAW,CAAC,MAAM,GAAG,KAAK,QAAQ;AAChC,cAAM,OAAO,IAAI,OAAO,CAAC,MAAM,EAAE,WAAW,OAAO;AACnD,YAAI,KAAK,UAAU,KAAK,KAAK,WAAW,IAAI,QAAQ;AAClD,iBAAO;AAAA,YACL,cAAc;AAAA,YACd,QAAQ,GAAG,KAAK,MAAM,gCAAgC,IAAI;AAAA,YAC1D,eAAe,KAAK,KAAK,SAAS,CAAC,EAAG;AAAA,UACxC;AAAA,QACF;AAAA,MACF;AACA,aAAO;AAAA,IACT;AAAA,EACF;AAAA;AAAA,EAEA;AAAA,IACE,IAAI;AAAA,IACJ,OAAO,CAAC,EAAE,KAAK,MAAM,MAAM;AACzB,UAAI,IAAI,SAAS,SAAS,MAAO,QAAO;AACxC,YAAM,oBAAoB,MAAM;AAAA,QAC9B,CAAC,MACC,EAAE,SAAS,WACV,EAAE,YAAY,mBAA0C,UACxD,EAAE,YAAY,iBAA4B;AAAA,MAC/C;AACA,YAAM,QAAQ,MAAM,OAAO,CAAC,MAAM,EAAE,SAAS,MAAM;AACnD,UAAI,qBAAqB,MAAM,WAAW,GAAG;AAC3C,eAAO;AAAA,UACL,cAAc;AAAA,UACd,QAAQ;AAAA,QACV;AAAA,MACF;AACA,aAAO;AAAA,IACT;AAAA,EACF;AAAA;AAAA,EAEA;AAAA,IACE,IAAI;AAAA,IACJ,OAAO,CAAC,EAAE,MAAM,MAAM;AACpB,YAAM,QAAQ,MAAM;AAAA,QAClB,CAAC,MACC,EAAE,SAAS,WACV,EAAuC,cAAc,YACrD,EAAuC,QAAQ;AAAA,MACpD;AACA,aAAO,QACH;AAAA,QACE,cAAc;AAAA,QACd,QAAQ;AAAA,QACR,eAAe,MAAM;AAAA,MACvB,IACA;AAAA,IACN;AAAA,EACF;AACF;AAEA,SAAS,oBAAoB,SAAkC,QAAyB;AACtF,MAAI,WAAW,wBAAwB,YAAY,QAAQ,kBAAkB,EAAE,SAAS;AACtF,WAAO;AACT,SAAO,gBAAgB,OAAO,EAAE,KAAK,CAAC,SAAS,KAAK,WAAW,MAAM;AACvE;AAEA,SAAS,iBAAiB,SAA2C;AACnE,MAAI,YAAY,QAAQ,aAAa,EAAE,SAAS,EAAG,QAAO;AAC1D,SAAO,gBAAgB,OAAO,EAAE;AAAA,IAC9B,CAAC,SAAS,MAAM,QAAQ,KAAK,aAAa,KAAK,KAAK,cAAc,SAAS;AAAA,EAC7E;AACF;AAEA,SAAS,gBAAgB,SAAkE;AACzF,SAAO;AAAA,IACL,GAAG,QAAQ,QAAQ,OAAO;AAAA,IAC1B,GAAG,QAAQ,QAAQ,eAAe;AAAA,IAClC,GAAG,QAAQ,QAAQ,KAAK;AAAA,EAC1B;AACF;AAEA,SAAS,QAAQ,OAAgD;AAC/D,MAAI,CAAC,MAAM,QAAQ,KAAK,EAAG,QAAO,CAAC;AACnC,SAAO,MAAM;AAAA,IACX,CAAC,SACC,QAAQ,IAAI,KAAK,OAAO,SAAS,YAAY,CAAC,MAAM,QAAQ,IAAI;AAAA,EACpE;AACF;AAEA,SAAS,YAAY,OAA0B;AAC7C,SAAO,MAAM,QAAQ,KAAK,IACtB,MAAM,OAAO,CAAC,SAAyB,OAAO,SAAS,QAAQ,IAC/D,CAAC;AACP;AAGO,SAAS,gBACd,KACA,QAAuB,eACA;AACvB,MAAI,IAAI,IAAI,SAAS,SAAS,SAAS,IAAI,IAAI,WAAW,aAAa;AACrE,WAAO,EAAE,cAAc,WAAW,QAAQ,qDAAqD;AAAA,EACjG;AACA,aAAW,QAAQ,OAAO;AACxB,UAAM,MAAM,KAAK,MAAM,GAAG;AAC1B,QAAI,IAAK,QAAO;AAAA,EAClB;AACA,SAAO,EAAE,cAAc,WAAW,QAAQ,sDAAsD;AAClG;;;ACpaA,eAAsB,sBACpB,OACA,OACA,UAA0B,CAAC,GACF;AACzB,QAAM,QAAQ,MAAM,UAAU,OAAO,KAAK;AAC1C,MAAI,MAAM,WAAW,GAAG;AACtB,WAAO;AAAA,MACL;AAAA,MACA,YAAY;AAAA,MACZ,uBAAuB;AAAA,MACvB,QAAQ,CAAC;AAAA,MACT,WAAW;AAAA,MACX,eAAe;AAAA,MACf,WAAW;AAAA,IACb;AAAA,EACF;AAEA,QAAM,SAAoC,CAAC;AAC3C,MAAI,cAAc;AAClB,MAAI,kBAAkB;AACtB,MAAI,wBAAwB;AAC5B,QAAM,cAAc,CAAC,GAAG,KAAK,EAAE,KAAK,CAAC,GAAG,MAAM,EAAE,YAAY,EAAE,SAAS;AACvE,QAAM,iBAAiB,oBAAI,IAAY;AAGvC,aAAW,KAAK,aAAa;AAC3B,WAAO,EAAE,QAAQ,MAAM;AAAA,MACrB,OAAO;AAAA,MACP,uBAAuB;AAAA,MACvB,QAAQ;AAAA,MACR,cAAc;AAAA,MACd,YAAY;AAAA,IACd;AACA,UAAM,OAAO,OAAO,EAAE,QAAQ;AAC9B,SAAK,SAAS;AACd,QAAI,EAAE,WAAW,SAAS;AACxB,WAAK,UAAU;AACf,qBAAe;AAAA,IACjB;AACA,QAAI,OAAO,EAAE,cAAc,SAAU,MAAK,gBAAgB,EAAE;AAC5D,QAAI,oBAAoB,CAAC,GAAG;AAC1B,+BAAyB;AACzB,WAAK,yBAAyB;AAC9B,YAAM,MAAM,GAAG,EAAE,QAAQ,IAAI,QAAQ,EAAE,IAAI,CAAC;AAC5C,UAAI,eAAe,IAAI,GAAG,GAAG;AAC3B,aAAK,cAAc;AACnB,2BAAmB;AAAA,MACrB;AACA,qBAAe,IAAI,GAAG;AAAA,IACxB;AAAA,EACF;AAEA,aAAW,QAAQ,OAAO,OAAO,MAAM,GAAG;AACxC,SAAK,eAAe,KAAK,QAAQ,IAAI,KAAK,eAAe,KAAK,QAAQ;AAAA,EACxE;AAGA,MAAI,qBAAqB;AACzB,MAAI,kBAAkB;AACtB,aAAW,CAAC,EAAE,GAAG,KAAK,QAAQ,aAAa,CAAC,MAAM,EAAE,QAAQ,GAAG;AAC7D,aAAS,IAAI,GAAG,IAAI,IAAI,QAAQ,KAAK;AACnC,UAAI,IAAI,CAAC,EAAG,WAAW,QAAS;AAChC,4BAAsB;AACtB,UAAI,IAAI,IAAI,CAAC,EAAG,oBAAmB;AAAA,IACrC;AAAA,EACF;AACA,QAAM,YAAY,qBAAqB,IAAI,kBAAkB,qBAAqB;AAElF,MAAI;AACJ,MAAI,QAAQ,iBAAiB;AAC3B,UAAM,UAAU,YAAY,OAAO,CAAC,MAAM,EAAE,UAAU,QAAQ,eAAgB;AAC9E,QAAI,QAAQ,SAAS,GAAG;AACtB,0BACE,QAAQ,OAAO,CAAC,MAAM,QAAQ,gBAAiB,EAAE,MAAM,CAAC,EAAE,SAAS,QAAQ;AAAA,IAC/E;AAAA,EACF;AAEA,SAAO;AAAA,IACL;AAAA,IACA,YAAY,YAAY;AAAA,IACxB;AAAA,IACA;AAAA,IACA,WAAW,cAAc,YAAY;AAAA,IACrC,eAAe,wBAAwB,IAAI,kBAAkB,wBAAwB;AAAA,IACrF;AAAA,IACA;AAAA,EACF;AACF;;;ACjGA,eAAsB,gBAAgB,OAAmB,OAAoC;AAC3F,QAAM,QAAQ,MAAM,MAAM,MAAM,EAAE,MAAM,CAAC;AACzC,QAAM,SAAS,MAAM,MAAM,OAAO,EAAE,MAAM,CAAC;AAC3C,QAAM,aAAa,oBAAI,IAAgC;AACvD,aAAW,KAAK,OAAO;AACrB,UAAM,MAAM,WAAW,IAAI,EAAE,YAAY,KAAK,CAAC;AAC/C,QAAI,KAAK,CAAC;AACV,eAAW,IAAI,EAAE,cAAc,GAAG;AAAA,EACpC;AAEA,aAAW,OAAO,WAAW,OAAO,EAAG,KAAI,KAAK,CAAC,GAAG,MAAM,EAAE,YAAY,EAAE,SAAS;AAEnF,QAAM,eAAe,oBAAI,IAA0B;AACnD,aAAW,KAAK,QAAQ;AACtB,QAAI,CAAC,EAAE,OAAQ;AACf,UAAM,MAAM,aAAa,IAAI,EAAE,MAAM,KAAK,CAAC;AAC3C,QAAI,KAAK,CAAC;AACV,iBAAa,IAAI,EAAE,QAAQ,GAAG;AAAA,EAChC;AAEA,QAAM,QAA0B,CAAC;AACjC,QAAM,OAAO,CAAC,QAA4B,UAAwB;AAChE,UAAM,OAAO,WAAW,IAAI,MAAM,KAAK,CAAC;AACxC,eAAW,SAAS,MAAM;AACxB,YAAM,KAAK;AAAA,QACT,OAAO,MAAM;AAAA,QACb,MAAM;AAAA,QACN;AAAA,QACA,QAAQ,aAAa,IAAI,MAAM,MAAM,KAAK,CAAC;AAAA,MAC7C,CAAC;AACD,WAAK,MAAM,QAAQ,QAAQ,CAAC;AAAA,IAC9B;AAAA,EACF;AACA,OAAK,QAAW,CAAC;AAEjB,QAAM,WAAW,MAAM,OAAO,CAAC,MAAM,EAAE,KAAK,SAAS,KAAK,EAAE;AAC5D,QAAM,YAAY,MAAM,OAAO,CAAC,MAAM,EAAE,KAAK,SAAS,MAAM,EAAE;AAC9D,QAAM,gBAAgB,MAAM,OAAO,CAAC,MAAM,EAAE,KAAK,SAAS,OAAO,EAAE;AACnE,QAAM,aAAa,MAAM,OAAO,CAAC,MAAM,EAAE,KAAK,SAAS,WAAW,EAAE;AAEpE,MAAI,kBAAkB;AACtB,MAAI,MAAM,SAAS,GAAG;AACpB,UAAM,SAAS,MAAM,IAAI,CAAC,MAAM,EAAE,SAAS;AAC3C,UAAM,OAAO,MAAM,IAAI,CAAC,MAAM,EAAE,WAAW,EAAE,SAAS;AACtD,sBAAkB,KAAK,IAAI,GAAG,IAAI,IAAI,KAAK,IAAI,GAAG,MAAM;AAAA,EAC1D;AAEA,SAAO,EAAE,OAAO,OAAO,UAAU,WAAW,eAAe,YAAY,gBAAgB;AACzF;;;ACfO,SAAS,kBACd,SACA,UAA2B,CAAC,GACZ;AAChB,QAAM,kBAAkB,QAAQ,mBAAmB;AACnD,QAAM,QAAQ,QAAQ,SAAS;AAC/B,QAAM,cAAc,QAAQ,uBAAuB;AAEnD,QAAM,UAA2B,QAAQ,IAAI,CAAC,MAAM;AAClD,QAAI,EAAE,SAAS,SAAS,KAAK,EAAE,UAAU,SAAS,GAAG;AACnD,YAAM,IAAI,MAAM,yDAAoD,EAAE,MAAM,GAAG;AAAA,IACjF;AACA,UAAM,QAAQ,KAAK,EAAE,QAAQ;AAC7B,UAAM,QAAQ,KAAK,EAAE,SAAS;AAC9B,UAAM,QAAQ,QAAQ;AACtB,UAAM,IAAI,QAAQ,EAAE,UAAU,EAAE,SAAS;AACzC,UAAM,EAAE,GAAG,IAAI,EAAE,IAAI,YAAY,EAAE,UAAU,EAAE,SAAS;AAKxD,UAAM,cAAc,IAAI,EAAE,QAAQ;AAClC,UAAM,eAAe,IAAI,EAAE,SAAS;AACpC,UAAM,iBAAiB,cAAc,KAAK,IAAI,KAAK,IAAI,KAAK,GAAG,IAAI,KAAK;AACxE,UAAM,kBAAkB,eAAe,KAAK,IAAI,KAAK,IAAI,KAAK,GAAG,IAAI,KAAK;AAC1E,UAAM,SAAS,kBAAkB;AACjC,UAAM,cAAc,KAAK,IAAI,aAAa,YAAY;AAEtD,QAAI;AACJ,QAAI,CAAC,QAAQ;AACX,gBAAU;AAAA,IACZ,WAAW,IAAI,SAAS,KAAK,IAAI,CAAC,KAAK,iBAAiB;AACtD,YAAM,oBAAoB,EAAE,iBAAiB,QAAQ,IAAI,QAAQ;AACjE,gBAAU,oBAAoB,aAAa;AAAA,IAC7C,OAAO;AACL,gBAAU;AAAA,IACZ;AAEA,WAAO;AAAA,MACL,QAAQ,EAAE;AAAA,MACV,cAAc;AAAA,MACd,eAAe;AAAA,MACf;AAAA,MACA,SAAS;AAAA,MACT,QAAQ;AAAA,MACR,SAAS;AAAA,MACT,QAAQ;AAAA,MACR;AAAA,MACA,KAAK;AAAA,MACL;AAAA,IACF;AAAA,EACF,CAAC;AAED,SAAO;AAAA,IACL;AAAA,IACA,eAAe,QAAQ,KAAK,CAAC,MAAM,EAAE,YAAY,WAAW;AAAA,IAC5D,aAAa,QAAQ,KAAK,CAAC,MAAM,EAAE,YAAY,UAAU;AAAA,EAC3D;AACF;AAEA,SAAS,KAAK,IAAsB;AAClC,SAAO,GAAG,OAAO,CAAC,GAAG,MAAM,IAAI,GAAG,CAAC,IAAI,GAAG;AAC5C;AAGO,SAAS,IAAI,IAAsB;AACxC,MAAI,GAAG,WAAW,EAAG,QAAO;AAC5B,QAAM,SAAS,CAAC,GAAG,EAAE,EAAE,KAAK,CAAC,GAAG,MAAM,IAAI,CAAC;AAC3C,QAAM,IAAI,CAAC,MAAc;AACvB,UAAM,MAAM,KAAK,OAAO,SAAS;AACjC,UAAM,KAAK,KAAK,MAAM,GAAG;AACzB,UAAM,KAAK,KAAK,KAAK,GAAG;AACxB,WAAO,OAAO,EAAE,KAAM,OAAO,EAAE,IAAK,OAAO,EAAE,MAAO,MAAM;AAAA,EAC5D;AACA,SAAO,EAAE,IAAI,IAAI,EAAE,IAAI;AACzB;AAOO,SAAS,YAAY,GAAa,GAAmD;AAC1F,MAAI,EAAE,SAAS,KAAK,EAAE,SAAS,EAAG,QAAO,EAAE,GAAG,GAAG,IAAI,GAAG,GAAG,EAAE;AAC7D,QAAM,KAAK,KAAK,CAAC;AACjB,QAAM,KAAK,KAAK,CAAC;AACjB,QAAM,KAAK,SAAS,GAAG,EAAE;AACzB,QAAM,KAAK,SAAS,GAAG,EAAE;AACzB,QAAM,YAAY,KAAK,EAAE,SAAS,KAAK,EAAE;AACzC,MAAI,cAAc,EAAG,QAAO,EAAE,GAAG,OAAO,KAAK,IAAI,UAAU,IAAI,GAAG,GAAG,OAAO,KAAK,IAAI,EAAE;AACvF,QAAM,KAAK,KAAK,MAAM,KAAK,KAAK,SAAS;AACzC,QAAM,KACH,YAAY,cACX,KAAK,EAAE,WAAW,KAAK,EAAE,SAAS,MAAM,KAAK,EAAE,WAAW,KAAK,EAAE,SAAS;AAC9E,QAAM,IAAI,KAAK,IAAI,YAAY,KAAK,IAAI,CAAC,GAAG,EAAE;AAC9C,SAAO,EAAE,GAAG,IAAI,EAAE;AACpB;AAEA,SAAS,SAAS,IAAc,GAAmB;AACjD,SAAO,GAAG,OAAO,CAAC,KAAK,MAAM,OAAO,IAAI,MAAM,GAAG,CAAC,KAAK,GAAG,SAAS;AACrE;AAGA,SAAS,YAAY,GAAW,IAAoB;AAClD,MAAI,MAAM,EAAG,QAAO;AACpB,MAAI,KAAK,IAAK,QAAO,UAAU,CAAC;AAChC,QAAM,IAAI,MAAM,KAAK,IAAI;AACzB,QAAM,KAAK,eAAe,GAAG,KAAK,GAAG,GAAG;AACxC,SAAO,KAAK,IAAI,IAAI,MAAM,KAAK,MAAM;AACvC;AAEA,SAAS,eAAe,GAAW,GAAW,GAAmB;AAC/D,MAAI,KAAK,EAAG,QAAO;AACnB,MAAI,KAAK,EAAG,QAAO;AACnB,QAAM,SAAS,QAAQ,CAAC,IAAI,QAAQ,CAAC,IAAI,QAAQ,IAAI,CAAC;AACtD,QAAM,QAAQ,KAAK,IAAI,KAAK,IAAI,CAAC,IAAI,IAAI,KAAK,IAAI,IAAI,CAAC,IAAI,IAAI,MAAM,IAAI;AACzE,MAAI,IAAI;AACR,MAAI,IAAI,KAAM,IAAI,KAAK,KAAM,IAAI;AACjC,MAAI,KAAK,IAAI,CAAC,IAAI,MAAO,KAAI;AAC7B,MAAI,IAAI;AACR,MAAI,IAAI;AACR,WAAS,IAAI,GAAG,KAAK,KAAK,KAAK;AAC7B,UAAM,KAAK,IAAI;AACf,QAAI,MAAO,KAAK,IAAI,KAAK,MAAO,IAAI,KAAK,MAAM,IAAI;AACnD,QAAI,IAAI,MAAM;AACd,QAAI,KAAK,IAAI,CAAC,IAAI,MAAO,KAAI;AAC7B,QAAI,IAAI,MAAM;AACd,QAAI,KAAK,IAAI,CAAC,IAAI,MAAO,KAAI;AAC7B,QAAI,IAAI;AACR,SAAK,IAAI;AACT,UAAM,GAAG,IAAI,MAAM,IAAI,IAAI,KAAK,OAAO,IAAI,OAAO,IAAI,KAAK;AAC3D,QAAI,IAAI,MAAM;AACd,QAAI,KAAK,IAAI,CAAC,IAAI,MAAO,KAAI;AAC7B,QAAI,IAAI,MAAM;AACd,QAAI,KAAK,IAAI,CAAC,IAAI,MAAO,KAAI;AAC7B,QAAI,IAAI;AACR,UAAM,QAAQ,IAAI;AAClB,SAAK;AACL,QAAI,KAAK,IAAI,QAAQ,CAAC,IAAI,KAAM;AAAA,EAClC;AACA,SAAO,QAAQ;AACjB;AAEA,SAAS,QAAQ,GAAmB;AAClC,QAAM,QAAQ;AAAA,IACZ;AAAA,IAAqB;AAAA,IAAmB;AAAA,IAAqB;AAAA,IAC7D;AAAA,IAAqB;AAAA,IAAoB;AAAA,IAAsB;AAAA,IAC/D;AAAA,EACF;AACA,MAAI,IAAI,IAAK,QAAO,KAAK,IAAI,KAAK,KAAK,KAAK,IAAI,KAAK,KAAK,CAAC,CAAC,IAAI,QAAQ,IAAI,CAAC;AAC7E,OAAK;AACL,MAAI,IAAI,MAAM,CAAC;AACf,WAAS,IAAI,GAAG,IAAI,GAAG,IAAK,MAAK,MAAM,CAAC,KAAM,IAAI;AAClD,QAAM,IAAI,IAAI;AACd,SAAO,MAAM,KAAK,IAAI,IAAI,KAAK,EAAE,KAAK,IAAI,OAAO,KAAK,IAAI,CAAC,IAAI,IAAI,KAAK,IAAI,CAAC;AAC/E;AAEA,SAAS,UAAU,GAAmB;AACpC,QAAM,KAAK;AACX,QAAM,KAAK;AACX,QAAM,KAAK;AACX,QAAM,KAAK;AACX,QAAM,KAAK;AACX,QAAM,IAAI;AACV,QAAM,OAAO,IAAI,IAAI,KAAK;AAC1B,QAAM,OAAO,KAAK,IAAI,CAAC;AACvB,QAAM,IAAI,KAAK,IAAI,IAAI;AACvB,QAAM,IAAI,QAAQ,KAAK,IAAI,MAAM,IAAI,MAAM,IAAI,MAAM,IAAI,MAAM,IAAI,KAAK,IAAK,CAAC,OAAO,OAAQ,CAAC;AAC9F,SAAO,OAAO,IAAI,OAAO;AAC3B;","names":[]}
|
|
@@ -10,6 +10,17 @@ var LLM_OUTPUT_TOKENS = "llm.token_count.completion";
|
|
|
10
10
|
var LLM_CACHED_TOKENS = "llm.token_count.prompt_cache_hit";
|
|
11
11
|
var LLM_COST_USD = "llm.cost_usd";
|
|
12
12
|
var TOOL_NAME = "tool.name";
|
|
13
|
+
var TOOL_ARGS_CAPTURED = "tool.args_captured";
|
|
14
|
+
var TOOL_LATENCY_MS = "tool.latency_ms";
|
|
15
|
+
var INPUT_VALUE = "input.value";
|
|
16
|
+
var OUTPUT_VALUE = "output.value";
|
|
17
|
+
var TOOL_SPAN_ATTRIBUTE_KEYS = [
|
|
18
|
+
TOOL_NAME,
|
|
19
|
+
TOOL_ARGS_CAPTURED,
|
|
20
|
+
TOOL_LATENCY_MS,
|
|
21
|
+
INPUT_VALUE,
|
|
22
|
+
OUTPUT_VALUE
|
|
23
|
+
];
|
|
13
24
|
var SPAN_KIND_ATTR_KEYS = [OPENINFERENCE_SPAN_KIND, "inference.observation_kind"];
|
|
14
25
|
var LLM_MODEL_ATTR_KEYS = [
|
|
15
26
|
LLM_MODEL_NAME,
|
|
@@ -45,6 +56,21 @@ var LLM_COST_ATTR_KEYS = [
|
|
|
45
56
|
"gen_ai.usage.cost"
|
|
46
57
|
];
|
|
47
58
|
var TOOL_NAME_ATTR_KEYS = [TOOL_NAME, "inference.tool.name"];
|
|
59
|
+
function toolSpanOtlpAttributes(span) {
|
|
60
|
+
const argsCaptured = span.argsCaptured !== false;
|
|
61
|
+
const attributes = {
|
|
62
|
+
[TOOL_NAME]: span.toolName,
|
|
63
|
+
[TOOL_ARGS_CAPTURED]: argsCaptured
|
|
64
|
+
};
|
|
65
|
+
if (span.latencyMs !== void 0) attributes[TOOL_LATENCY_MS] = span.latencyMs;
|
|
66
|
+
if (argsCaptured) attributes[INPUT_VALUE] = stringifyTraceValue(span.args);
|
|
67
|
+
if (span.result !== void 0) attributes[OUTPUT_VALUE] = stringifyTraceValue(span.result);
|
|
68
|
+
return attributes;
|
|
69
|
+
}
|
|
70
|
+
function applyToolSpanOtlpAttributes(attributes, span) {
|
|
71
|
+
for (const key of TOOL_SPAN_ATTRIBUTE_KEYS) delete attributes[key];
|
|
72
|
+
Object.assign(attributes, toolSpanOtlpAttributes(span));
|
|
73
|
+
}
|
|
48
74
|
function traceSpanKindToOpenInferenceKind(kind) {
|
|
49
75
|
switch (kind) {
|
|
50
76
|
case "llm":
|
|
@@ -63,6 +89,15 @@ function traceSpanKindToOpenInferenceKind(kind) {
|
|
|
63
89
|
return "SPAN";
|
|
64
90
|
}
|
|
65
91
|
}
|
|
92
|
+
function stringifyTraceValue(value) {
|
|
93
|
+
if (value === void 0) return "null";
|
|
94
|
+
if (typeof value === "string") return value;
|
|
95
|
+
try {
|
|
96
|
+
return JSON.stringify(value) ?? String(value);
|
|
97
|
+
} catch {
|
|
98
|
+
return String(value);
|
|
99
|
+
}
|
|
100
|
+
}
|
|
66
101
|
|
|
67
102
|
// src/trace-analyst/otlp-span.ts
|
|
68
103
|
function projectOtlpFlatLine(raw) {
|
|
@@ -82,9 +117,9 @@ function projectOtlpFlatLine(raw) {
|
|
|
82
117
|
const kind = inferOtlpKind(attributes);
|
|
83
118
|
let duration_ms = 0;
|
|
84
119
|
if (start_time && end_time) {
|
|
85
|
-
const a =
|
|
86
|
-
const b =
|
|
87
|
-
if (
|
|
120
|
+
const a = spanEpochMillis(start_time);
|
|
121
|
+
const b = spanEpochMillis(end_time);
|
|
122
|
+
if (a !== null && b !== null) duration_ms = Math.max(0, b - a);
|
|
88
123
|
}
|
|
89
124
|
return {
|
|
90
125
|
trace_id,
|
|
@@ -1003,6 +1038,10 @@ export {
|
|
|
1003
1038
|
LLM_CACHED_TOKENS,
|
|
1004
1039
|
LLM_COST_USD,
|
|
1005
1040
|
TOOL_NAME,
|
|
1041
|
+
TOOL_ARGS_CAPTURED,
|
|
1042
|
+
TOOL_LATENCY_MS,
|
|
1043
|
+
INPUT_VALUE,
|
|
1044
|
+
OUTPUT_VALUE,
|
|
1006
1045
|
SPAN_KIND_ATTR_KEYS,
|
|
1007
1046
|
LLM_MODEL_ATTR_KEYS,
|
|
1008
1047
|
LLM_INPUT_TOKEN_ATTR_KEYS,
|
|
@@ -1010,6 +1049,7 @@ export {
|
|
|
1010
1049
|
LLM_CACHED_TOKEN_ATTR_KEYS,
|
|
1011
1050
|
LLM_COST_ATTR_KEYS,
|
|
1012
1051
|
TOOL_NAME_ATTR_KEYS,
|
|
1052
|
+
applyToolSpanOtlpAttributes,
|
|
1013
1053
|
traceSpanKindToOpenInferenceKind,
|
|
1014
1054
|
projectOtlpFlatLine,
|
|
1015
1055
|
readOtlpStatus,
|
|
@@ -1031,4 +1071,4 @@ export {
|
|
|
1031
1071
|
buildTraceAnalystTools,
|
|
1032
1072
|
traceAnalystFunctionGroup
|
|
1033
1073
|
};
|
|
1034
|
-
//# sourceMappingURL=chunk-
|
|
1074
|
+
//# sourceMappingURL=chunk-S2F4J57L.js.map
|