ccqa 1.24.0 → 1.25.0
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/dist/bin/ccqa.mjs +511 -227
- package/dist/hub-client/index.d.mts +16 -2
- package/dist/package.json +1 -1
- package/package.json +1 -1
package/dist/bin/ccqa.mjs
CHANGED
|
@@ -3115,12 +3115,29 @@ const NO_DRIFT_CAUSE = "NO_DRIFT";
|
|
|
3115
3115
|
const DRIFT_ACTUAL_CAUSES = [...DRIFT_FAILURE_CAUSES, NO_DRIFT_CAUSE];
|
|
3116
3116
|
const ACTUAL_CAUSES = [...FAILURE_CAUSES, NO_DRIFT_CAUSE];
|
|
3117
3117
|
const ActualCauseSchema = z.enum(ACTUAL_CAUSES);
|
|
3118
|
+
/**
|
|
3119
|
+
* Report rows come in kinds, and each answers fewer questions than the last.
|
|
3120
|
+
* A "record" row answers none: it says a recording happened and what it cost,
|
|
3121
|
+
* and nothing was judged, so both vocabularies below are empty for it.
|
|
3122
|
+
*
|
|
3123
|
+
* The one source for the vocabulary: the report envelope, the hub's
|
|
3124
|
+
* `Run.kind` and the `?kind=` query params all derive from this enum, so a
|
|
3125
|
+
* fourth kind cannot be half-added. What a kind that judges nothing may and
|
|
3126
|
+
* may not advance is ADR-0017.
|
|
3127
|
+
*/
|
|
3128
|
+
const ReportKindSchema = z.enum([
|
|
3129
|
+
"run",
|
|
3130
|
+
"drift",
|
|
3131
|
+
"record"
|
|
3132
|
+
]);
|
|
3118
3133
|
/** What a person may record on a row of this kind. */
|
|
3119
3134
|
function causesForKind(kind) {
|
|
3135
|
+
if (kind === "record") return [];
|
|
3120
3136
|
return kind === "drift" ? DRIFT_ACTUAL_CAUSES : FAILURE_CAUSES;
|
|
3121
3137
|
}
|
|
3122
3138
|
/** What the model may answer on a row of this kind. */
|
|
3123
3139
|
function predictedForKind(kind) {
|
|
3140
|
+
if (kind === "record") return [];
|
|
3124
3141
|
return kind === "drift" ? [...DRIFT_FAILURE_CAUSES, "UNKNOWN"] : PREDICTED_LABELS;
|
|
3125
3142
|
}
|
|
3126
3143
|
const SUB_DIAGNOSES = [...FIXABLE_DIAGNOSIS_TYPES, "NONE"];
|
|
@@ -3146,6 +3163,20 @@ const FailureEvidenceSchema = z.object({
|
|
|
3146
3163
|
*/
|
|
3147
3164
|
const DriftSurfaceSchema = z.enum(["spec", "generated"]);
|
|
3148
3165
|
/**
|
|
3166
|
+
* Which repair a `SPEC_CHANGE` needs: `FEATURE_REMOVED` means the behaviour
|
|
3167
|
+
* the spec checks is gone from the code, so the spec goes with it;
|
|
3168
|
+
* `BEHAVIOUR_CHANGED` means it still exists but works differently, so the spec
|
|
3169
|
+
* is rewritten and re-recorded.
|
|
3170
|
+
*
|
|
3171
|
+
* A fourth axis rather than a `subDiagnosis` value, because that vocabulary is
|
|
3172
|
+
* `[...FIXABLE_DIAGNOSIS_TYPES, "NONE"]` — the shapes a machine knows how to
|
|
3173
|
+
* repair — so a spec change always lands on `NONE` there.
|
|
3174
|
+
*
|
|
3175
|
+
* Deliberately without an "unknown" member: absence carries it, and every
|
|
3176
|
+
* reader must treat an absent value as a call to leave to a human.
|
|
3177
|
+
*/
|
|
3178
|
+
const SpecChangeKindSchema = z.enum(["FEATURE_REMOVED", "BEHAVIOUR_CHANGED"]);
|
|
3179
|
+
/**
|
|
3149
3180
|
* LLM output shape. Deliberately NOT .strict(): the model occasionally adds
|
|
3150
3181
|
* keys, and rejecting the whole analysis over an extra field would collapse
|
|
3151
3182
|
* a usable prediction into UNKNOWN. Zod's default strips unknown keys.
|
|
@@ -3162,7 +3193,8 @@ const FailureAnalysisSchema = z.object({
|
|
|
3162
3193
|
recommendation: z.string().default(""),
|
|
3163
3194
|
evidence: z.array(FailureEvidenceSchema),
|
|
3164
3195
|
reasoning: z.string(),
|
|
3165
|
-
surface: DriftSurfaceSchema.optional()
|
|
3196
|
+
surface: DriftSurfaceSchema.optional(),
|
|
3197
|
+
specChangeKind: SpecChangeKindSchema.optional()
|
|
3166
3198
|
});
|
|
3167
3199
|
/**
|
|
3168
3200
|
* What a drift audit may conclude, in the same vocabulary `ccqa run
|
|
@@ -3201,11 +3233,33 @@ const DriftDiagnosisSchema = z.object({
|
|
|
3201
3233
|
confidence: z.number().min(0).max(1),
|
|
3202
3234
|
surface: DriftSurfaceSchema.default("spec"),
|
|
3203
3235
|
subDiagnosis: DriftSubDiagnosisSchema.default("NONE"),
|
|
3236
|
+
specChangeKind: SpecChangeKindSchema.optional(),
|
|
3204
3237
|
headline: z.string(),
|
|
3205
3238
|
recommendation: z.string().default(""),
|
|
3206
3239
|
evidence: z.array(FailureEvidenceSchema),
|
|
3207
3240
|
reasoning: z.string().default("")
|
|
3208
3241
|
});
|
|
3242
|
+
/**
|
|
3243
|
+
* `specChangeKind` names which repair a `SPEC_CHANGE` needs, so under any other
|
|
3244
|
+
* label it is dropped. Dropped rather than rejected — a stray value must not
|
|
3245
|
+
* cost an otherwise usable verdict.
|
|
3246
|
+
*
|
|
3247
|
+
* Two producers apply it. `ccqa audit` normalizes at the parse boundary
|
|
3248
|
+
* (`src/drift/analyze.ts`), so every consumer of a verdict it produced — the
|
|
3249
|
+
* JSON output, the report rows, the hub push — is clean by construction. The
|
|
3250
|
+
* hub normalizes rows on the way into the drift ledger, which is what a
|
|
3251
|
+
* foreign client's push passes through.
|
|
3252
|
+
*
|
|
3253
|
+
* What does NOT hold: a pushed report's stored archive is kept verbatim and
|
|
3254
|
+
* served back unchanged, so a row a foreign client wrote can still carry a
|
|
3255
|
+
* stray field. That is why the UI re-checks the label before rendering the
|
|
3256
|
+
* chip rather than trusting the stored row.
|
|
3257
|
+
*/
|
|
3258
|
+
function normalizeDiagnosis(diagnosis) {
|
|
3259
|
+
if (diagnosis.label === "SPEC_CHANGE" || diagnosis.specChangeKind === void 0) return diagnosis;
|
|
3260
|
+
const { specChangeKind: _dropped, ...rest } = diagnosis;
|
|
3261
|
+
return rest;
|
|
3262
|
+
}
|
|
3209
3263
|
const ReportAssertionSchema = z.object({
|
|
3210
3264
|
name: z.string(),
|
|
3211
3265
|
status: z.enum([
|
|
@@ -3359,7 +3413,7 @@ const GitEnvelopeSchema = z.object({
|
|
|
3359
3413
|
});
|
|
3360
3414
|
const RunReportDataSchema = z.object({
|
|
3361
3415
|
schemaVersion: z.literal(1),
|
|
3362
|
-
kind:
|
|
3416
|
+
kind: ReportKindSchema.default("run"),
|
|
3363
3417
|
createdAt: z.string(),
|
|
3364
3418
|
runId: z.string().nullable(),
|
|
3365
3419
|
runUrl: z.string().nullable().optional(),
|
|
@@ -6738,7 +6792,7 @@ z.object({
|
|
|
6738
6792
|
profile: z.string().nullable(),
|
|
6739
6793
|
branch: z.string().nullable(),
|
|
6740
6794
|
status: RunStatusSchema,
|
|
6741
|
-
kind:
|
|
6795
|
+
kind: ReportKindSchema.default("run"),
|
|
6742
6796
|
drift: z.object({
|
|
6743
6797
|
specs: z.number(),
|
|
6744
6798
|
testDrift: z.number(),
|
|
@@ -7108,6 +7162,7 @@ z.object({
|
|
|
7108
7162
|
const SpecDriftEntrySchema = z.object({
|
|
7109
7163
|
label: DriftLabelSchema.nullable(),
|
|
7110
7164
|
surface: DriftSurfaceSchema.optional(),
|
|
7165
|
+
specChangeKind: SpecChangeKindSchema.optional(),
|
|
7111
7166
|
confidence: z.number().optional(),
|
|
7112
7167
|
headline: z.string().optional(),
|
|
7113
7168
|
gitHead: z.string(),
|
|
@@ -7158,6 +7213,24 @@ const CreateLearningJobRequestSchema = z.object({
|
|
|
7158
7213
|
profile: z.string(),
|
|
7159
7214
|
runLimit: z.number().int().positive().max(1e3).optional()
|
|
7160
7215
|
});
|
|
7216
|
+
const MAX_ACK_KEYS = 5e3;
|
|
7217
|
+
const MAX_ACK_KEY_LENGTH = 256;
|
|
7218
|
+
/**
|
|
7219
|
+
* A named set of opaque keys a consumer has already acted on, as stored (see
|
|
7220
|
+
* `AckStore`). `at` is null only for a set that was never written — an ack
|
|
7221
|
+
* nobody has recorded yet reads as empty rather than missing.
|
|
7222
|
+
*/
|
|
7223
|
+
const AckSchema = z.object({
|
|
7224
|
+
keys: z.array(z.string()),
|
|
7225
|
+
at: z.string().nullable()
|
|
7226
|
+
});
|
|
7227
|
+
/** Body of `PUT /projects/:project/acks/:name?profile=` — the whole set, not a delta. */
|
|
7228
|
+
const PutAckRequestSchema = z.object({ keys: z.array(z.string().min(1).max(MAX_ACK_KEY_LENGTH)).max(MAX_ACK_KEYS) });
|
|
7229
|
+
AckSchema.extend({
|
|
7230
|
+
project: z.string(),
|
|
7231
|
+
profile: z.string(),
|
|
7232
|
+
name: z.string()
|
|
7233
|
+
});
|
|
7161
7234
|
//#endregion
|
|
7162
7235
|
//#region src/run/hub-selection.ts
|
|
7163
7236
|
/**
|
|
@@ -10993,14 +11066,9 @@ async function generateAgentBrowserTest(ctx) {
|
|
|
10993
11066
|
blank();
|
|
10994
11067
|
const agentBrowserSession = fix.useSnapshot ? `ccqa-generate-${Date.now()}` : void 0;
|
|
10995
11068
|
const runVitestForSession = (path) => runVitest(path, agentBrowserSession);
|
|
10996
|
-
let signalHandler = null;
|
|
10997
11069
|
if (agentBrowserSession) {
|
|
10998
11070
|
await closeSession(agentBrowserSession);
|
|
10999
|
-
|
|
11000
|
-
closeSession(agentBrowserSession).finally(() => process.exit(130));
|
|
11001
|
-
};
|
|
11002
|
-
process.once("SIGINT", signalHandler);
|
|
11003
|
-
process.once("SIGTERM", signalHandler);
|
|
11071
|
+
ctx.teardown?.trackSession(agentBrowserSession);
|
|
11004
11072
|
}
|
|
11005
11073
|
try {
|
|
11006
11074
|
const initialRun = await timedPhase("vitest run #1", () => runVitestForSession(scriptPath), "run");
|
|
@@ -11027,11 +11095,10 @@ async function generateAgentBrowserTest(ctx) {
|
|
|
11027
11095
|
passed
|
|
11028
11096
|
};
|
|
11029
11097
|
} finally {
|
|
11030
|
-
if (
|
|
11031
|
-
|
|
11032
|
-
|
|
11098
|
+
if (agentBrowserSession) {
|
|
11099
|
+
ctx.teardown?.untrackSession(agentBrowserSession);
|
|
11100
|
+
await closeSession(agentBrowserSession);
|
|
11033
11101
|
}
|
|
11034
|
-
if (agentBrowserSession) await closeSession(agentBrowserSession);
|
|
11035
11102
|
}
|
|
11036
11103
|
}
|
|
11037
11104
|
/**
|
|
@@ -11836,6 +11903,75 @@ function createIncrementalReport(reportDir, envelope, sink, costNow) {
|
|
|
11836
11903
|
};
|
|
11837
11904
|
}
|
|
11838
11905
|
//#endregion
|
|
11906
|
+
//#region src/cli/open-hub-run.ts
|
|
11907
|
+
/** The one wording for "this flag needs a hub, and none is configured". */
|
|
11908
|
+
function needsHubConnection(flag) {
|
|
11909
|
+
return `${flag} requires a hub connection (--hub-url/--hub-token or CCQA_HUB_URL/CCQA_HUB_TOKEN)`;
|
|
11910
|
+
}
|
|
11911
|
+
const REPORT_TO_HUB_NEEDS_CONNECTION = needsHubConnection("--report-to-hub");
|
|
11912
|
+
/**
|
|
11913
|
+
* The connection a `--report-to-hub` command publishes through. Both CLI
|
|
11914
|
+
* callers check this before the expensive part — the audit's sweep, the
|
|
11915
|
+
* recording's spec lock and browser — so a job that cannot publish spends
|
|
11916
|
+
* nothing finding out.
|
|
11917
|
+
*/
|
|
11918
|
+
function requireReportToHubConnection(conn) {
|
|
11919
|
+
if (conn) return conn;
|
|
11920
|
+
error(REPORT_TO_HUB_NEEDS_CONNECTION);
|
|
11921
|
+
process.exit(2);
|
|
11922
|
+
}
|
|
11923
|
+
/**
|
|
11924
|
+
* Open the run a `--report-to-hub` command patches into. Failure is fatal: a
|
|
11925
|
+
* job that asked to publish and cannot reach the hub has not done what it was
|
|
11926
|
+
* told. Thrown rather than exited, so a caller's `finally` still runs (the
|
|
11927
|
+
* audit releases its spec claims there). Not retried: a dropped response after
|
|
11928
|
+
* the hub committed would leave a second orphan running run.
|
|
11929
|
+
*/
|
|
11930
|
+
async function openHubRun(kind, conn, cwd, profile) {
|
|
11931
|
+
const [branch, gitHead] = await Promise.all([detectBranch(cwd), getGitHead(cwd)]);
|
|
11932
|
+
const ciRunId = githubRunId();
|
|
11933
|
+
const runUrl = githubRunUrl();
|
|
11934
|
+
try {
|
|
11935
|
+
const run = await conn.hub.openRun({
|
|
11936
|
+
project: conn.project,
|
|
11937
|
+
kind,
|
|
11938
|
+
...branch ? { branch } : {},
|
|
11939
|
+
...profile ? { profile } : {},
|
|
11940
|
+
...gitHead ? { gitHead } : {},
|
|
11941
|
+
...ciRunId ? { ciRunId } : {},
|
|
11942
|
+
...runUrl ? { runUrl } : {}
|
|
11943
|
+
});
|
|
11944
|
+
return {
|
|
11945
|
+
hub: conn.hub,
|
|
11946
|
+
kind,
|
|
11947
|
+
runId: run.id,
|
|
11948
|
+
gitHead
|
|
11949
|
+
};
|
|
11950
|
+
} catch (err) {
|
|
11951
|
+
throw new RunUsageError(`--report-to-hub: could not open a run on the hub (${errMessage(err)})`);
|
|
11952
|
+
}
|
|
11953
|
+
}
|
|
11954
|
+
/**
|
|
11955
|
+
* Close an open run with its final rows and envelope, answering whether it
|
|
11956
|
+
* closed. A failed seal leaves the run `running` with whatever rows landed — a
|
|
11957
|
+
* wrong record, not merely a missing one — so a CLI caller must not exit clean.
|
|
11958
|
+
* The exit is the caller's to make rather than taken here: `ccqa record` seals
|
|
11959
|
+
* from inside a teardown finalizer, and exiting there would skip the
|
|
11960
|
+
* browser-session reap queued behind it.
|
|
11961
|
+
*/
|
|
11962
|
+
async function sealHubRun(push, body) {
|
|
11963
|
+
try {
|
|
11964
|
+
await push.hub.patchRun(push.runId, {
|
|
11965
|
+
...body,
|
|
11966
|
+
done: true
|
|
11967
|
+
});
|
|
11968
|
+
return true;
|
|
11969
|
+
} catch (err) {
|
|
11970
|
+
error(`hub: could not close the ${push.kind} run ${push.runId}: ${errMessage(err)}`);
|
|
11971
|
+
return false;
|
|
11972
|
+
}
|
|
11973
|
+
}
|
|
11974
|
+
//#endregion
|
|
11839
11975
|
//#region src/prompts/agent-update.ts
|
|
11840
11976
|
/**
|
|
11841
11977
|
* Build the prompts used by the `--learn-*-prompt` flags to refresh
|
|
@@ -12373,9 +12509,9 @@ async function executeRun(targets, opts) {
|
|
|
12373
12509
|
});
|
|
12374
12510
|
if (wantsLastGreen && hubCtx == null) throw new RunUsageError("--on-fail-explain needs a hub connection for the per-spec last-green baselines (--hub-url/--hub-token or CCQA_HUB_URL/CCQA_HUB_TOKEN), or an explicit --on-fail-explain-base <ref>");
|
|
12375
12511
|
const ledgerHub = wantsLastGreen ? hubCtx : null;
|
|
12376
|
-
if (rerunProfile !== null && hubCtx == null) throw new RunUsageError("--only-hub-rerun-needed
|
|
12377
|
-
if (opts.reportToHub && hubCtx == null) throw new RunUsageError(
|
|
12378
|
-
if (opts.learnHubLivePrompt && hubCtx == null) throw new RunUsageError("--learn-hub-live-prompt
|
|
12512
|
+
if (rerunProfile !== null && hubCtx == null) throw new RunUsageError(needsHubConnection("--only-hub-rerun-needed"));
|
|
12513
|
+
if (opts.reportToHub && hubCtx == null) throw new RunUsageError(REPORT_TO_HUB_NEEDS_CONNECTION);
|
|
12514
|
+
if (opts.learnHubLivePrompt && hubCtx == null) throw new RunUsageError(needsHubConnection("--learn-hub-live-prompt"));
|
|
12379
12515
|
const [customPrompt, triageUserPrompt, ledgerEntries, rerunReport, fetchedDeployHead] = await Promise.all([
|
|
12380
12516
|
forExecution ? fetchCustomPrompt(hubCtx) : null,
|
|
12381
12517
|
forExecution ? fetchTriageUserPrompt(hubCtx) : null,
|
|
@@ -13137,7 +13273,16 @@ async function streamFiltered(source, sink, capture) {
|
|
|
13137
13273
|
function createRunTeardown() {
|
|
13138
13274
|
const sessions = /* @__PURE__ */ new Set();
|
|
13139
13275
|
const finalizers = [];
|
|
13140
|
-
let
|
|
13276
|
+
let running = null;
|
|
13277
|
+
const tearDown = async () => {
|
|
13278
|
+
for (const fn of finalizers) try {
|
|
13279
|
+
await fn();
|
|
13280
|
+
} catch (err) {
|
|
13281
|
+
warn(`teardown finalizer failed (partial report may be incomplete): ${err instanceof Error ? err.message : String(err)}`);
|
|
13282
|
+
}
|
|
13283
|
+
await Promise.all([...sessions].map((name) => closeSession(name)));
|
|
13284
|
+
sessions.clear();
|
|
13285
|
+
};
|
|
13141
13286
|
return {
|
|
13142
13287
|
trackSession(name) {
|
|
13143
13288
|
sessions.add(name);
|
|
@@ -13148,24 +13293,18 @@ function createRunTeardown() {
|
|
|
13148
13293
|
onFinalize(fn) {
|
|
13149
13294
|
finalizers.push(fn);
|
|
13150
13295
|
},
|
|
13151
|
-
|
|
13152
|
-
|
|
13153
|
-
|
|
13154
|
-
for (const fn of finalizers) try {
|
|
13155
|
-
await fn();
|
|
13156
|
-
} catch (err) {
|
|
13157
|
-
warn(`teardown finalizer failed (partial report may be incomplete): ${err instanceof Error ? err.message : String(err)}`);
|
|
13158
|
-
}
|
|
13159
|
-
await Promise.all([...sessions].map((name) => closeSession(name)));
|
|
13160
|
-
sessions.clear();
|
|
13296
|
+
run() {
|
|
13297
|
+
running ??= tearDown();
|
|
13298
|
+
return running;
|
|
13161
13299
|
}
|
|
13162
13300
|
};
|
|
13163
13301
|
}
|
|
13164
13302
|
/**
|
|
13165
13303
|
* Install SIGINT/SIGTERM handlers that run {@link RunTeardown.run} then exit
|
|
13166
13304
|
* with the conventional signal code, and return a disposer that removes them.
|
|
13167
|
-
*
|
|
13168
|
-
*
|
|
13305
|
+
* A command must install at most one: a second handler that also exits would
|
|
13306
|
+
* race this one and could terminate mid-finalizer. A second signal while
|
|
13307
|
+
* tearing down hard-exits immediately rather than waiting.
|
|
13169
13308
|
*/
|
|
13170
13309
|
function installTeardownSignalHandlers(teardown) {
|
|
13171
13310
|
let handling = false;
|
|
@@ -14691,8 +14830,11 @@ function parseAutoFixFlag(raw) {
|
|
|
14691
14830
|
* The `generate` flow shared by `ccqa generate` and the codegen half of
|
|
14692
14831
|
* `ccqa record`: resolve the spec's target plugin, load its input (the
|
|
14693
14832
|
* recording, for input:"recording" targets), and dispatch to the plugin.
|
|
14694
|
-
* This layer owns the CLI concerns — overwrite confirmation, logging
|
|
14695
|
-
*
|
|
14833
|
+
* This layer owns the CLI concerns — overwrite confirmation, logging — while
|
|
14834
|
+
* the plugin owns the generation pipeline. A generation whose output still
|
|
14835
|
+
* fails is reported as `{ passed: false }` rather than thrown or exited: both
|
|
14836
|
+
* callers have work left that `process.exit` would skip — `ccqa record
|
|
14837
|
+
* --report-to-hub` still has to seal the run holding what the retries cost.
|
|
14696
14838
|
*/
|
|
14697
14839
|
async function runGenerate(featureName, specName, opts) {
|
|
14698
14840
|
header("generate", `${featureName}/${specName}`);
|
|
@@ -14700,7 +14842,7 @@ async function runGenerate(featureName, specName, opts) {
|
|
|
14700
14842
|
await ensureCcqaDir(cwd);
|
|
14701
14843
|
const releaseLock = await acquireSpecLock(featureName, specName, "generate", cwd);
|
|
14702
14844
|
try {
|
|
14703
|
-
await runGenerateLocked(featureName, specName, opts, cwd);
|
|
14845
|
+
return await runGenerateLocked(featureName, specName, opts, cwd);
|
|
14704
14846
|
} finally {
|
|
14705
14847
|
await releaseLock();
|
|
14706
14848
|
}
|
|
@@ -14731,7 +14873,7 @@ async function runGenerateLocked(featureName, specName, opts, cwd) {
|
|
|
14731
14873
|
if (existingOutput && !opts.force) {
|
|
14732
14874
|
if (!await confirmOverwrite(existingOutput)) {
|
|
14733
14875
|
info("aborted; pass --overwrite to replace it without prompting");
|
|
14734
|
-
return;
|
|
14876
|
+
return { passed: true };
|
|
14735
14877
|
}
|
|
14736
14878
|
}
|
|
14737
14879
|
let recording;
|
|
@@ -14760,15 +14902,14 @@ async function runGenerateLocked(featureName, specName, opts, cwd) {
|
|
|
14760
14902
|
maxRetries: opts.maxRetries,
|
|
14761
14903
|
mode: opts.fixMode,
|
|
14762
14904
|
useSnapshot: opts.useSnapshot
|
|
14763
|
-
}
|
|
14905
|
+
},
|
|
14906
|
+
teardown: opts.teardown
|
|
14764
14907
|
};
|
|
14765
14908
|
const result = await target.generate(ctx);
|
|
14766
14909
|
if (opts.updateAgentPrompt) await runGenerateAgentPromptUpdate(target, featureName, specName, result, opts, cwd);
|
|
14767
|
-
if (!result.passed)
|
|
14768
|
-
|
|
14769
|
-
|
|
14770
|
-
}
|
|
14771
|
-
hint(`run 'ccqa run ${featureName}/${specName}' to execute the test`);
|
|
14910
|
+
if (!result.passed) warn("auto-fix exhausted; test still failing");
|
|
14911
|
+
else hint(`run 'ccqa run ${featureName}/${specName}' to execute the test`);
|
|
14912
|
+
return { passed: result.passed };
|
|
14772
14913
|
}
|
|
14773
14914
|
/**
|
|
14774
14915
|
* `ccqa generate --learn-hub-codegen-prompt`: refresh the target's learned
|
|
@@ -14838,15 +14979,18 @@ async function runGenerateCli(specPath, opts) {
|
|
|
14838
14979
|
cwd
|
|
14839
14980
|
});
|
|
14840
14981
|
if (opts.learnHubCodegenPrompt && hubClient === null) {
|
|
14841
|
-
error("--learn-hub-codegen-prompt
|
|
14982
|
+
error(needsHubConnection("--learn-hub-codegen-prompt"));
|
|
14842
14983
|
process.exit(2);
|
|
14843
14984
|
}
|
|
14844
14985
|
const hubContext = hubClient && project ? {
|
|
14845
14986
|
hub: hubClient,
|
|
14846
14987
|
project
|
|
14847
14988
|
} : null;
|
|
14989
|
+
const teardown = createRunTeardown();
|
|
14990
|
+
const disposeSignalHandlers = installTeardownSignalHandlers(teardown);
|
|
14991
|
+
let passed;
|
|
14848
14992
|
try {
|
|
14849
|
-
await runGenerate(featureName, specName, {
|
|
14993
|
+
({passed} = await runGenerate(featureName, specName, {
|
|
14850
14994
|
maxRetries: parseInt(opts.autoFixMaxRetries ?? "3", 10),
|
|
14851
14995
|
fixMode: toFixMode(opts.autoFix ?? "interactive"),
|
|
14852
14996
|
force: opts.overwrite ?? false,
|
|
@@ -14856,15 +15000,20 @@ async function runGenerateCli(specPath, opts) {
|
|
|
14856
15000
|
targetOverride: opts.target,
|
|
14857
15001
|
cwd,
|
|
14858
15002
|
hubContext,
|
|
14859
|
-
updateAgentPrompt: opts.learnHubCodegenPrompt ?? false
|
|
14860
|
-
|
|
15003
|
+
updateAgentPrompt: opts.learnHubCodegenPrompt ?? false,
|
|
15004
|
+
teardown
|
|
15005
|
+
}));
|
|
14861
15006
|
} catch (e) {
|
|
14862
15007
|
if (e instanceof SpecLockedError) {
|
|
14863
15008
|
error(e.message);
|
|
14864
15009
|
process.exit(2);
|
|
14865
15010
|
}
|
|
14866
15011
|
throw e;
|
|
15012
|
+
} finally {
|
|
15013
|
+
await teardown.run();
|
|
15014
|
+
disposeSignalHandlers();
|
|
14867
15015
|
}
|
|
15016
|
+
if (!passed) process.exit(1);
|
|
14868
15017
|
}
|
|
14869
15018
|
//#endregion
|
|
14870
15019
|
//#region src/cli/record.ts
|
|
@@ -14872,7 +15021,7 @@ const VALIDATION_MODES = ["lenient", "strict"];
|
|
|
14872
15021
|
const recordCommand = addHubOptions(addProfileOption(addLanguageOption(new Command("record").argument("<feature/spec>", "Spec id in '<feature>/<spec>' form (resolves to .ccqa/features/<feature>/test-cases/<spec>/)").description("Record a test from a spec: run agent-browser to collect actions (trace), then compile them into runnable code via the spec's target (generate) — a vitest test.spec.ts for agent-browser, a @playwright/test spec for the playwright target. Recording-backed targets only; spec-input targets like runn have no trace step (use `ccqa generate`), and agent-browser live specs need no recording.").optionsGroup("How to record:").option("-m, --model <name>", "Claude model alias ('sonnet'|'opus'|'haiku') or full ID. Overrides CCQA_MODEL.").option("--trace-validation <mode>", "What to do with actions that fail post-trace validation: 'lenient' (default) tags them; 'strict' drops them.", (raw) => {
|
|
14873
15022
|
if (VALIDATION_MODES.includes(raw)) return raw;
|
|
14874
15023
|
throw new Error(`--trace-validation must be one of ${VALIDATION_MODES.join(" | ")}`);
|
|
14875
|
-
}, "lenient").option("--auto-fix <mode>", "Auto-fix behaviour during script generation: 'interactive' (default, prompt y/N; declines on non-TTY), 'auto' (apply without prompt, for CI), 'skip' (agent-browser: apply only high-confidence fixes; external targets like playwright/runn: no fix pass at all).", parseAutoFixFlag, "interactive").option("--auto-fix-max-retries <n>", "Maximum number of auto-fix retries", "3").option("--trace-only", "Stop after the trace step; do not generate test code").option("--no-session-pin", "Don't pin AGENT_BROWSER_SESSION / capture page snapshots after a failure (debug toggle)").optionsGroup("What to do with the result:").option("--overwrite", "Replace an existing test.spec.ts without warning").optionsGroup("Learning:").option("--learn-hub-trace-prompt", "After the trace finishes, ask Claude to refresh the \"record.agent\" prompt on the hub from a summary of the run. Requires a hub connection.").optionsGroup("Environment and connection:").option("--cwd <path>", "Working directory containing the .ccqa/ tree (monorepo support). Defaults to the current directory.").option("--project <name>", "Project name for the hub. Defaults to the current directory's name.")))).action(withUsageErrors(async (specPath, opts) => {
|
|
15024
|
+
}, "lenient").option("--auto-fix <mode>", "Auto-fix behaviour during script generation: 'interactive' (default, prompt y/N; declines on non-TTY), 'auto' (apply without prompt, for CI), 'skip' (agent-browser: apply only high-confidence fixes; external targets like playwright/runn: no fix pass at all).", parseAutoFixFlag, "interactive").option("--auto-fix-max-retries <n>", "Maximum number of auto-fix retries", "3").option("--trace-only", "Stop after the trace step; do not generate test code").option("--no-session-pin", "Don't pin AGENT_BROWSER_SESSION / capture page snapshots after a failure (debug toggle)").optionsGroup("What to do with the result:").option("--overwrite", "Replace an existing test.spec.ts without warning").option("--report-to-hub", "Leave a run (kind: record) on the hub saying this spec was recorded and what the recording spent on Claude, so a budget summed over the hub's runs sees it. It advances no ledger: a recording verifies nothing.").optionsGroup("Learning:").option("--learn-hub-trace-prompt", "After the trace finishes, ask Claude to refresh the \"record.agent\" prompt on the hub from a summary of the run. Requires a hub connection.").optionsGroup("Environment and connection:").option("--cwd <path>", "Working directory containing the .ccqa/ tree (monorepo support). Defaults to the current directory.").option("--project <name>", "Project name for the hub. Defaults to the current directory's name.")))).action(withUsageErrors(async (specPath, opts) => {
|
|
14876
15025
|
await withCostReporting("record", () => runRecord(specPath, opts));
|
|
14877
15026
|
}));
|
|
14878
15027
|
async function runRecord(specPath, opts) {
|
|
@@ -14911,9 +15060,10 @@ async function runRecord(specPath, opts) {
|
|
|
14911
15060
|
project: hubProject
|
|
14912
15061
|
} : null;
|
|
14913
15062
|
if (opts.learnHubTracePrompt && hubContext === null) {
|
|
14914
|
-
error("--learn-hub-trace-prompt
|
|
15063
|
+
error(needsHubConnection("--learn-hub-trace-prompt"));
|
|
14915
15064
|
process.exit(2);
|
|
14916
15065
|
}
|
|
15066
|
+
const pushConn = opts.reportToHub ? requireReportToHubConnection(hubContext) : null;
|
|
14917
15067
|
const releaseLock = await acquireSpecLock(featureName, specName, "record", cwdForProfile).catch((e) => {
|
|
14918
15068
|
if (e instanceof SpecLockedError) {
|
|
14919
15069
|
error(e.message);
|
|
@@ -14921,37 +15071,78 @@ async function runRecord(specPath, opts) {
|
|
|
14921
15071
|
}
|
|
14922
15072
|
throw e;
|
|
14923
15073
|
});
|
|
14924
|
-
|
|
15074
|
+
const push = pushConn ? await openHubRun("record", pushConn, cwdForProfile, opts.hubProfile) : null;
|
|
15075
|
+
if (push) info(`hub: record run opened (${push.runId})`);
|
|
15076
|
+
let recorded = false;
|
|
15077
|
+
let sealed = true;
|
|
15078
|
+
const teardown = createRunTeardown();
|
|
15079
|
+
teardown.onFinalize(async () => {
|
|
15080
|
+
if (push) sealed = await sealRecordPush(push, featureName, specName, recorded);
|
|
15081
|
+
});
|
|
15082
|
+
const disposeSignalHandlers = installTeardownSignalHandlers(teardown);
|
|
14925
15083
|
try {
|
|
14926
|
-
traceResult =
|
|
14927
|
-
|
|
14928
|
-
|
|
14929
|
-
|
|
14930
|
-
|
|
14931
|
-
|
|
14932
|
-
|
|
14933
|
-
|
|
14934
|
-
|
|
14935
|
-
|
|
14936
|
-
|
|
14937
|
-
|
|
14938
|
-
|
|
14939
|
-
|
|
14940
|
-
|
|
15084
|
+
let traceResult = null;
|
|
15085
|
+
let generated = true;
|
|
15086
|
+
try {
|
|
15087
|
+
traceResult = await runTrace(featureName, specName, opts.model, opts.traceValidation ?? "lenient", language, {
|
|
15088
|
+
cwd: cwdForProfile,
|
|
15089
|
+
hubContext
|
|
15090
|
+
});
|
|
15091
|
+
blank();
|
|
15092
|
+
if (!opts.traceOnly) generated = (await runGenerate(featureName, specName, {
|
|
15093
|
+
maxRetries: parseInt(opts.autoFixMaxRetries ?? "3", 10),
|
|
15094
|
+
fixMode: toFixMode(opts.autoFix ?? "interactive"),
|
|
15095
|
+
force: opts.overwrite ?? false,
|
|
15096
|
+
useSnapshot: opts.sessionPin !== false,
|
|
15097
|
+
language,
|
|
15098
|
+
model: opts.model,
|
|
15099
|
+
cwd: cwdForProfile,
|
|
15100
|
+
hubContext,
|
|
15101
|
+
teardown
|
|
15102
|
+
})).passed;
|
|
15103
|
+
} finally {
|
|
15104
|
+
await releaseLock();
|
|
15105
|
+
}
|
|
15106
|
+
if (opts.learnHubTracePrompt && traceResult !== null) {
|
|
15107
|
+
blank();
|
|
15108
|
+
await updateAgentPrompt({
|
|
15109
|
+
kind: "record",
|
|
15110
|
+
flag: "--learn-hub-trace-prompt",
|
|
15111
|
+
runSummary: buildRecordRunSummary(featureName, specName, traceResult),
|
|
15112
|
+
hubContext,
|
|
15113
|
+
...opts.model ? { model: opts.model } : {},
|
|
15114
|
+
...language ? { language } : {}
|
|
15115
|
+
});
|
|
15116
|
+
}
|
|
15117
|
+
recorded = generated;
|
|
14941
15118
|
} finally {
|
|
14942
|
-
await
|
|
14943
|
-
|
|
14944
|
-
if (opts.learnHubTracePrompt && traceResult !== null) {
|
|
14945
|
-
blank();
|
|
14946
|
-
await updateAgentPrompt({
|
|
14947
|
-
kind: "record",
|
|
14948
|
-
flag: "--learn-hub-trace-prompt",
|
|
14949
|
-
runSummary: buildRecordRunSummary(featureName, specName, traceResult),
|
|
14950
|
-
hubContext,
|
|
14951
|
-
...opts.model ? { model: opts.model } : {},
|
|
14952
|
-
...language ? { language } : {}
|
|
14953
|
-
});
|
|
15119
|
+
await teardown.run();
|
|
15120
|
+
disposeSignalHandlers();
|
|
14954
15121
|
}
|
|
15122
|
+
if (!sealed) process.exit(2);
|
|
15123
|
+
if (!recorded) process.exit(1);
|
|
15124
|
+
}
|
|
15125
|
+
/**
|
|
15126
|
+
* Close the record run with the one row this command produced, answering
|
|
15127
|
+
* whether it closed. One spec is recorded per invocation, so one row is the
|
|
15128
|
+
* whole run — enough for the runs list to say what the money bought.
|
|
15129
|
+
*/
|
|
15130
|
+
async function sealRecordPush(push, featureName, specName, recorded) {
|
|
15131
|
+
return sealHubRun(push, {
|
|
15132
|
+
rows: [emptySpecRow({
|
|
15133
|
+
feature: featureName,
|
|
15134
|
+
spec: specName,
|
|
15135
|
+
title: null,
|
|
15136
|
+
status: recorded ? "passed" : "failed"
|
|
15137
|
+
})],
|
|
15138
|
+
reportMeta: {
|
|
15139
|
+
git: {
|
|
15140
|
+
head: push.gitHead,
|
|
15141
|
+
base: null
|
|
15142
|
+
},
|
|
15143
|
+
cost: currentReportCost()
|
|
15144
|
+
}
|
|
15145
|
+
});
|
|
14955
15146
|
}
|
|
14956
15147
|
/**
|
|
14957
15148
|
* Compact summary of the trace pass for the record agent-prompt refresh.
|
|
@@ -15175,6 +15366,7 @@ Drift found:
|
|
|
15175
15366
|
"confidence": 0.0,
|
|
15176
15367
|
"surface": "spec" | "generated",
|
|
15177
15368
|
"subDiagnosis": "SELECTOR_DRIFT" | "OVER_ASSERTION" | "NONE",
|
|
15369
|
+
"specChangeKind": "FEATURE_REMOVED" | "BEHAVIOUR_CHANGED",
|
|
15178
15370
|
"headline": "<one line: what is out of sync>",
|
|
15179
15371
|
"recommendation": "<what to change to bring them back in sync>",
|
|
15180
15372
|
"reasoning": "<how you reached this label: what you looked for, what you found, why it is this label and not the other>",
|
|
@@ -15186,6 +15378,13 @@ Drift found:
|
|
|
15186
15378
|
\`\`\`
|
|
15187
15379
|
|
|
15188
15380
|
\`subDiagnosis\`: \`SELECTOR_DRIFT\` when a selector or string was renamed, \`OVER_ASSERTION\` when the spec asserts something narrower than the product ever promised, \`NONE\` otherwise.
|
|
15381
|
+
|
|
15382
|
+
\`specChangeKind\`: set it only when the label is \`SPEC_CHANGE\`, and omit the field entirely otherwise. It says which repair the spec needs — deleting it, or rewriting and re-recording it:
|
|
15383
|
+
|
|
15384
|
+
- \`FEATURE_REMOVED\` — the code no longer implements the behaviour at all: removed, moved elsewhere, or deliberately disabled. This is the stronger claim, so earn it: your evidence must point at where the implementation would be if it still existed.
|
|
15385
|
+
- \`BEHAVIOUR_CHANGED\` — the behaviour is still there, but its wording, its route, or the conditions it runs under moved.
|
|
15386
|
+
|
|
15387
|
+
When the evidence does not support "gone", answer \`BEHAVIOUR_CHANGED\`. When neither reading is supported, omit the field — there is no value for "I cannot tell", and a human decides what you leave unsaid.
|
|
15189
15388
|
`;
|
|
15190
15389
|
}
|
|
15191
15390
|
function buildDriftUserPrompt(artifacts) {
|
|
@@ -15366,10 +15565,11 @@ async function checkSpec(target, opts) {
|
|
|
15366
15565
|
continue;
|
|
15367
15566
|
}
|
|
15368
15567
|
try {
|
|
15568
|
+
const reply = DriftReplySchema.parse(JSON.parse(json));
|
|
15369
15569
|
return {
|
|
15370
15570
|
target,
|
|
15371
15571
|
ok: true,
|
|
15372
|
-
drift:
|
|
15572
|
+
drift: reply.drift ? normalizeDiagnosis(reply.drift) : null,
|
|
15373
15573
|
live: artifacts.live,
|
|
15374
15574
|
title: artifacts.title
|
|
15375
15575
|
};
|
|
@@ -15503,7 +15703,7 @@ function determineExitCode(results, threshold) {
|
|
|
15503
15703
|
//#endregion
|
|
15504
15704
|
//#region src/drift/to-report.ts
|
|
15505
15705
|
/** Tracks the drift prompt's own version — the two must never drift apart. */
|
|
15506
|
-
const DRIFT_REPORT_PROMPT_VERSION = "
|
|
15706
|
+
const DRIFT_REPORT_PROMPT_VERSION = "6";
|
|
15507
15707
|
/**
|
|
15508
15708
|
* Spec-level status under the given threshold, mirroring determineExitCode's
|
|
15509
15709
|
* per-spec logic (exit-code.ts) but scoped to a single SpecResult.
|
|
@@ -15536,7 +15736,7 @@ function driftResultToRow(result, threshold) {
|
|
|
15536
15736
|
status: specStatus(result, threshold)
|
|
15537
15737
|
}),
|
|
15538
15738
|
...result.live === void 0 ? {} : { mode: result.live ? "live" : "deterministic" },
|
|
15539
|
-
analysis: result.drift
|
|
15739
|
+
analysis: result.drift
|
|
15540
15740
|
};
|
|
15541
15741
|
}
|
|
15542
15742
|
/**
|
|
@@ -15822,14 +16022,24 @@ async function runAudit(specPath, opts) {
|
|
|
15822
16022
|
if (targets.length === 0) exitWithNoSpecs(format, "noDiffIntersection", "no specs intersect the changed file set; nothing to check");
|
|
15823
16023
|
}
|
|
15824
16024
|
const blocks = await loadAvailableBlocks(cwd);
|
|
15825
|
-
|
|
16025
|
+
let pushConn = null;
|
|
16026
|
+
if (opts.reportToHub) {
|
|
16027
|
+
const pushHub = resolveHubClient(opts);
|
|
16028
|
+
pushConn = requireReportToHubConnection(pushHub && {
|
|
16029
|
+
hub: pushHub,
|
|
16030
|
+
project: resolveProject({
|
|
16031
|
+
project: opts.project,
|
|
16032
|
+
cwd
|
|
16033
|
+
})
|
|
16034
|
+
});
|
|
16035
|
+
}
|
|
15826
16036
|
let results;
|
|
15827
16037
|
let promptCtx;
|
|
15828
16038
|
let push = null;
|
|
15829
16039
|
try {
|
|
15830
16040
|
promptCtx = await resolveAuditPromptContext(opts, cwd);
|
|
15831
|
-
if (
|
|
15832
|
-
push = await
|
|
16041
|
+
if (pushConn) {
|
|
16042
|
+
push = await openHubRun("drift", pushConn, cwd, opts.hubProfile);
|
|
15833
16043
|
if (format === "text") info(`hub: incremental drift run opened (${push.runId})`);
|
|
15834
16044
|
}
|
|
15835
16045
|
results = await analyzeDrift({
|
|
@@ -15890,56 +16100,6 @@ async function releaseSpecs(hub, project, profile, holder) {
|
|
|
15890
16100
|
}
|
|
15891
16101
|
}
|
|
15892
16102
|
/**
|
|
15893
|
-
* Fail fast when `--report-to-hub` was requested but no hub connection is
|
|
15894
|
-
* available. Called at the top of `runAudit`, before the sweep spends any
|
|
15895
|
-
* model calls — `pushDriftResults` (below) keeps its own equivalent, injectable
|
|
15896
|
-
* check so it stays safe to call standalone (e.g. in tests) without this guard
|
|
15897
|
-
* having already run.
|
|
15898
|
-
*/
|
|
15899
|
-
function requireReportToHubConnection(opts) {
|
|
15900
|
-
if (!opts.reportToHub || resolveHubClient(opts)) return;
|
|
15901
|
-
error("--report-to-hub requires a hub connection (--hub-url/--hub-token or CCQA_HUB_URL/CCQA_HUB_TOKEN)");
|
|
15902
|
-
process.exit(2);
|
|
15903
|
-
}
|
|
15904
|
-
/**
|
|
15905
|
-
* Open the drift run this sweep patches into. A failure here is fatal, as it
|
|
15906
|
-
* is for `ccqa run`: a job that asked to publish and cannot reach the hub has
|
|
15907
|
-
* not done what it was told, and the audit has no local artifact to fall back
|
|
15908
|
-
* on — the hub is its only output. Raised before any spec is checked, so
|
|
15909
|
-
* nothing is wasted. Not retried: a dropped response after the hub committed
|
|
15910
|
-
* would leave a second orphan running run.
|
|
15911
|
-
*
|
|
15912
|
-
* `resolveHub` is injectable for tests.
|
|
15913
|
-
*/
|
|
15914
|
-
async function openDriftPush(opts, cwd, resolveHub = resolveHubClient) {
|
|
15915
|
-
const hub = resolveHub(opts);
|
|
15916
|
-
if (!hub) throw new RunUsageError("--report-to-hub requires a hub connection (--hub-url/--hub-token or CCQA_HUB_URL/CCQA_HUB_TOKEN)");
|
|
15917
|
-
const project = resolveProject({
|
|
15918
|
-
project: opts.project,
|
|
15919
|
-
cwd
|
|
15920
|
-
});
|
|
15921
|
-
const [branch, gitHead] = await Promise.all([detectBranch(cwd), getGitHead(cwd)]);
|
|
15922
|
-
const ciRunId = githubRunId();
|
|
15923
|
-
const runUrl = githubRunUrl();
|
|
15924
|
-
try {
|
|
15925
|
-
return {
|
|
15926
|
-
hub,
|
|
15927
|
-
runId: (await hub.openRun({
|
|
15928
|
-
project,
|
|
15929
|
-
kind: "drift",
|
|
15930
|
-
...branch ? { branch } : {},
|
|
15931
|
-
...opts.hubProfile ? { profile: opts.hubProfile } : {},
|
|
15932
|
-
...gitHead ? { gitHead } : {},
|
|
15933
|
-
...ciRunId ? { ciRunId } : {},
|
|
15934
|
-
...runUrl ? { runUrl } : {}
|
|
15935
|
-
})).id,
|
|
15936
|
-
gitHead
|
|
15937
|
-
};
|
|
15938
|
-
} catch (err) {
|
|
15939
|
-
throw new RunUsageError(`--report-to-hub: could not open a run on the hub (${errMessage(err)})`);
|
|
15940
|
-
}
|
|
15941
|
-
}
|
|
15942
|
-
/**
|
|
15943
16103
|
* Send one finished spec. A failure is warned and swallowed rather than
|
|
15944
16104
|
* thrown: the seal resends every row, so a dropped patch costs freshness for
|
|
15945
16105
|
* the rest of the sweep, not the record.
|
|
@@ -15971,22 +16131,16 @@ async function sealDriftPush(push, args) {
|
|
|
15971
16131
|
customPromptVersion: promptCtx?.customPromptVersion ?? null,
|
|
15972
16132
|
triageUserPromptHash: promptCtx?.triageUserPromptHash ?? null
|
|
15973
16133
|
});
|
|
15974
|
-
|
|
15975
|
-
|
|
15976
|
-
|
|
15977
|
-
|
|
15978
|
-
|
|
15979
|
-
|
|
15980
|
-
|
|
15981
|
-
|
|
15982
|
-
|
|
15983
|
-
|
|
15984
|
-
}
|
|
15985
|
-
});
|
|
15986
|
-
} catch (err) {
|
|
15987
|
-
error(`hub: could not close the drift run ${push.runId}: ${errMessage(err)}`);
|
|
15988
|
-
process.exit(2);
|
|
15989
|
-
}
|
|
16134
|
+
if (!await sealHubRun(push, {
|
|
16135
|
+
rows: report.results,
|
|
16136
|
+
reportMeta: {
|
|
16137
|
+
git: report.git,
|
|
16138
|
+
promptVersion: report.promptVersion,
|
|
16139
|
+
customPromptVersion: report.customPromptVersion,
|
|
16140
|
+
...report.triageUserPromptHash ? { triageUserPromptHash: report.triageUserPromptHash } : {},
|
|
16141
|
+
cost: report.cost
|
|
16142
|
+
}
|
|
16143
|
+
})) process.exit(2);
|
|
15990
16144
|
if (format === "text") info(`pushed drift result to hub: ${(opts.hubUrl ?? process.env.CCQA_HUB_URL ?? "").replace(/\/+$/, "")}/#/runs/${push.runId}`);
|
|
15991
16145
|
}
|
|
15992
16146
|
/**
|
|
@@ -16824,6 +16978,7 @@ function gradedDriftEntry(ledger, key, runId, label) {
|
|
|
16824
16978
|
delete graded.headline;
|
|
16825
16979
|
delete graded.confidence;
|
|
16826
16980
|
}
|
|
16981
|
+
if (label !== "SPEC_CHANGE") delete graded.specChangeKind;
|
|
16827
16982
|
return graded;
|
|
16828
16983
|
}
|
|
16829
16984
|
//#endregion
|
|
@@ -16958,7 +17113,7 @@ function createPushRunHandler(config) {
|
|
|
16958
17113
|
runUrl: report.runUrl ?? null,
|
|
16959
17114
|
reportCreatedAt: report.createdAt,
|
|
16960
17115
|
createdAt: (/* @__PURE__ */ new Date()).toISOString(),
|
|
16961
|
-
...await resolveDeployedSha(config.storage, project, profile, deployedSha),
|
|
17116
|
+
...await resolveDeployedSha(config.storage, kind, project, profile, deployedSha),
|
|
16962
17117
|
deployedShaAmbiguous: false
|
|
16963
17118
|
};
|
|
16964
17119
|
await config.storage.artifacts.putDir(run.id, dir);
|
|
@@ -17009,7 +17164,7 @@ function createOpenRunHandler(config) {
|
|
|
17009
17164
|
runUrl: runUrl || null,
|
|
17010
17165
|
reportCreatedAt: now,
|
|
17011
17166
|
createdAt: now,
|
|
17012
|
-
...await resolveDeployedSha(config.storage, project, profile, deployedSha),
|
|
17167
|
+
...await resolveDeployedSha(config.storage, kind, project, profile, deployedSha),
|
|
17013
17168
|
deployedShaAmbiguous: false
|
|
17014
17169
|
};
|
|
17015
17170
|
await config.storage.runs.create(run);
|
|
@@ -17131,10 +17286,11 @@ async function updateDriftLedger(storage, run, results) {
|
|
|
17131
17286
|
for (const row of results) {
|
|
17132
17287
|
if (row.status === "skipped") continue;
|
|
17133
17288
|
const key = `${row.feature}/${row.spec}`;
|
|
17134
|
-
const diagnosis = row.analysis;
|
|
17289
|
+
const diagnosis = row.analysis ? normalizeDiagnosis(row.analysis) : null;
|
|
17135
17290
|
ledger.specs[key] = {
|
|
17136
17291
|
label: diagnosis ? diagnosis.label : null,
|
|
17137
17292
|
surface: diagnosis?.surface,
|
|
17293
|
+
specChangeKind: diagnosis?.specChangeKind,
|
|
17138
17294
|
confidence: diagnosis?.confidence,
|
|
17139
17295
|
headline: diagnosis?.headline,
|
|
17140
17296
|
gitHead,
|
|
@@ -17157,11 +17313,15 @@ async function updateDriftLedger(storage, run, results) {
|
|
|
17157
17313
|
* Best-effort: a deploy log the hub can't read leaves the run unattributed
|
|
17158
17314
|
* (re-run selection then answers `unknown`) rather than rejecting the run.
|
|
17159
17315
|
*/
|
|
17160
|
-
async function resolveDeployedSha(storage, project, profile, explicit) {
|
|
17316
|
+
async function resolveDeployedSha(storage, kind, project, profile, explicit) {
|
|
17161
17317
|
if (explicit) return {
|
|
17162
17318
|
deployedSha: explicit,
|
|
17163
17319
|
deployedShaSource: "client"
|
|
17164
17320
|
};
|
|
17321
|
+
if (kind === "record") return {
|
|
17322
|
+
deployedSha: null,
|
|
17323
|
+
deployedShaSource: null
|
|
17324
|
+
};
|
|
17165
17325
|
try {
|
|
17166
17326
|
const head = await storage.deploys.head(project, profile ?? "default");
|
|
17167
17327
|
if (head) return {
|
|
@@ -17266,17 +17426,19 @@ function createPatchRunHandler(config) {
|
|
|
17266
17426
|
sendJson(ctx.res, 200, updated);
|
|
17267
17427
|
};
|
|
17268
17428
|
}
|
|
17269
|
-
/** GET /api/v1/runs?project&branch&status&limit */
|
|
17429
|
+
/** GET /api/v1/runs?project&branch&status&kind&limit */
|
|
17270
17430
|
function createListRunsHandler(storage) {
|
|
17271
17431
|
return async (ctx) => {
|
|
17272
17432
|
const project = ctx.url.searchParams.get("project");
|
|
17273
17433
|
const branch = ctx.url.searchParams.get("branch");
|
|
17274
17434
|
const status = ctx.url.searchParams.get("status");
|
|
17275
17435
|
const limitRaw = ctx.url.searchParams.get("limit");
|
|
17436
|
+
const kindsRaw = ctx.url.searchParams.get("kind");
|
|
17276
17437
|
const runs = await storage.runs.list({
|
|
17277
17438
|
...project ? { project } : {},
|
|
17278
17439
|
...branch ? { branch } : {},
|
|
17279
17440
|
...status ? { status } : {},
|
|
17441
|
+
...kindsRaw ? { kinds: kindsRaw.split(",").map(requireKind) } : {},
|
|
17280
17442
|
...limitRaw ? { limit: Number(limitRaw) } : {}
|
|
17281
17443
|
});
|
|
17282
17444
|
sendJson(ctx.res, 200, { runs: await Promise.all(runs.map((r) => withGradedDrift(storage, r))) });
|
|
@@ -17410,16 +17572,21 @@ function parseRunScope(ctx) {
|
|
|
17410
17572
|
const profileRaw = ctx.url.searchParams.get("profile");
|
|
17411
17573
|
const profile = profileRaw ? requireSafeSegment(profileRaw, "profile") : null;
|
|
17412
17574
|
const kindRaw = ctx.url.searchParams.get("kind");
|
|
17413
|
-
if (kindRaw !== null && kindRaw !== "run" && kindRaw !== "drift") throw new HttpError(400, "invalid_param", `invalid kind: must be "run" or "drift"`);
|
|
17414
17575
|
const deployedSha = boundedParam(ctx.url.searchParams.get("deployedSha"), "deployedSha", 64);
|
|
17415
17576
|
return {
|
|
17416
17577
|
project,
|
|
17417
17578
|
branch,
|
|
17418
17579
|
profile,
|
|
17419
|
-
kind: kindRaw
|
|
17580
|
+
kind: kindRaw === null ? "run" : requireKind(kindRaw),
|
|
17420
17581
|
deployedSha
|
|
17421
17582
|
};
|
|
17422
17583
|
}
|
|
17584
|
+
/** One `kind` value, validated against the enum so the message can't drift from it. */
|
|
17585
|
+
function requireKind(raw) {
|
|
17586
|
+
const parsed = ReportKindSchema.safeParse(raw);
|
|
17587
|
+
if (!parsed.success) throw new HttpError(400, "invalid_param", `invalid kind: must be one of ${ReportKindSchema.options.map((k) => `"${k}"`).join(", ")}`);
|
|
17588
|
+
return parsed.data;
|
|
17589
|
+
}
|
|
17423
17590
|
/**
|
|
17424
17591
|
* A branch is a free-form label (e.g. `feature/foo`), so `/` is allowed —
|
|
17425
17592
|
* only length is bounded (a sanity cap; the last-green ledger separately
|
|
@@ -17965,7 +18132,7 @@ function createGetAuditNeedHandler(storage) {
|
|
|
17965
18132
|
//#endregion
|
|
17966
18133
|
//#region src/hub/api/handlers/locks.ts
|
|
17967
18134
|
/** A spec-key list and three short strings; nothing here should approach this. */
|
|
17968
|
-
const MAX_BODY_BYTES$
|
|
18135
|
+
const MAX_BODY_BYTES$2 = 1024 * 1024;
|
|
17969
18136
|
/**
|
|
17970
18137
|
* POST /api/v1/projects/:project/locks?profile=
|
|
17971
18138
|
*
|
|
@@ -17979,7 +18146,7 @@ function createAcquireLocksHandler(storage) {
|
|
|
17979
18146
|
return async (ctx) => {
|
|
17980
18147
|
const project = requireSafeSegment(ctx.params.project, "project");
|
|
17981
18148
|
const profile = requireProfileParam(ctx.url);
|
|
17982
|
-
const body = await readJsonBody(ctx.req, MAX_BODY_BYTES$
|
|
18149
|
+
const body = await readJsonBody(ctx.req, MAX_BODY_BYTES$2, AcquireLocksRequestSchema, "lock request");
|
|
17983
18150
|
let result = {
|
|
17984
18151
|
granted: [],
|
|
17985
18152
|
denied: []
|
|
@@ -18008,12 +18175,55 @@ function createReleaseLocksHandler(storage) {
|
|
|
18008
18175
|
return async (ctx) => {
|
|
18009
18176
|
const project = requireSafeSegment(ctx.params.project, "project");
|
|
18010
18177
|
const profile = requireProfileParam(ctx.url);
|
|
18011
|
-
const body = await readJsonBody(ctx.req, MAX_BODY_BYTES$
|
|
18178
|
+
const body = await readJsonBody(ctx.req, MAX_BODY_BYTES$2, ReleaseLocksRequestSchema, "release request");
|
|
18012
18179
|
await storage.locks.update(project, profile, (current) => releaseAll(current, body.holder));
|
|
18013
18180
|
ctx.res.writeHead(204).end();
|
|
18014
18181
|
};
|
|
18015
18182
|
}
|
|
18016
18183
|
//#endregion
|
|
18184
|
+
//#region src/hub/api/handlers/acks.ts
|
|
18185
|
+
/**
|
|
18186
|
+
* Pre-parse guard only. `PutAckRequestSchema` holds the real bound (5000 keys
|
|
18187
|
+
* of 256 characters); this sits above the largest body that can satisfy it —
|
|
18188
|
+
* 5000 keys of 256 `\uXXXX`-escaped characters — so a conforming client is
|
|
18189
|
+
* never answered 413 by a limit the documented bounds don't mention.
|
|
18190
|
+
*/
|
|
18191
|
+
const MAX_BODY_BYTES$1 = 8 * 1024 * 1024;
|
|
18192
|
+
function requireAckKey(ctx) {
|
|
18193
|
+
return {
|
|
18194
|
+
project: requireSafeSegment(ctx.params.project, "project"),
|
|
18195
|
+
profile: requireProfileParam(ctx.url),
|
|
18196
|
+
name: requireSafeSegment(ctx.params.name, "name")
|
|
18197
|
+
};
|
|
18198
|
+
}
|
|
18199
|
+
/**
|
|
18200
|
+
* GET /api/v1/projects/:project/acks/:name?profile= — an unset ack answers 200
|
|
18201
|
+
* with an empty set, not 404: "nothing acted on yet" is a real answer, and a
|
|
18202
|
+
* 404 there is a special case consumers get wrong.
|
|
18203
|
+
*/
|
|
18204
|
+
function createGetAckHandler(storage) {
|
|
18205
|
+
return async (ctx) => {
|
|
18206
|
+
const key = requireAckKey(ctx);
|
|
18207
|
+
const ack = await storage.acks.get(key.project, key.profile, key.name);
|
|
18208
|
+
sendJson(ctx.res, 200, {
|
|
18209
|
+
...key,
|
|
18210
|
+
...ack
|
|
18211
|
+
});
|
|
18212
|
+
};
|
|
18213
|
+
}
|
|
18214
|
+
/** PUT /api/v1/projects/:project/acks/:name?profile= */
|
|
18215
|
+
function createPutAckHandler(storage) {
|
|
18216
|
+
return async (ctx) => {
|
|
18217
|
+
const key = requireAckKey(ctx);
|
|
18218
|
+
const body = await readJsonBody(ctx.req, MAX_BODY_BYTES$1, PutAckRequestSchema, "ack body");
|
|
18219
|
+
const ack = await storage.acks.put(key.project, key.profile, key.name, body.keys);
|
|
18220
|
+
sendJson(ctx.res, 200, {
|
|
18221
|
+
...key,
|
|
18222
|
+
...ack
|
|
18223
|
+
});
|
|
18224
|
+
};
|
|
18225
|
+
}
|
|
18226
|
+
//#endregion
|
|
18017
18227
|
//#region src/hub/core/rerun.ts
|
|
18018
18228
|
/**
|
|
18019
18229
|
* When each deployed commit reached the environment. A baseline read at that
|
|
@@ -19263,7 +19473,12 @@ const CSS = `
|
|
|
19263
19473
|
.empty-note { color: var(--muted); font-size: 13px; padding: 16px 2px; }
|
|
19264
19474
|
|
|
19265
19475
|
.runid { font-family: var(--mono); font-size: 13px; font-weight: 600; }
|
|
19266
|
-
.
|
|
19476
|
+
/* Chips wrap when a run carries several labels. Laid out as inline content
|
|
19477
|
+
they wrapped to a line indented by the chips' own left margin, and the two
|
|
19478
|
+
lines sat at text leading — too tight to read as separate rows. Flex with a
|
|
19479
|
+
gap aligns every line at the same left edge and spaces them the same way
|
|
19480
|
+
horizontally and vertically. */
|
|
19481
|
+
.subline { display: flex; flex-wrap: wrap; align-items: center; gap: 6px; margin-top: 4px; }
|
|
19267
19482
|
.ci-badge { display: inline-flex; align-items: center; gap: 5px; font-size: 11px; font-family: var(--mono); color: var(--muted); background: var(--surface-3); border: 1px solid var(--border); border-radius: 5px; padding: 1px 6px; }
|
|
19268
19483
|
.ci-badge.local { color: var(--muted-2); }
|
|
19269
19484
|
a.ci-badge { text-decoration: none; }
|
|
@@ -19279,6 +19494,15 @@ const CSS = `
|
|
|
19279
19494
|
.badge.skipped .d { background: var(--muted); }
|
|
19280
19495
|
.badge.running { background: var(--amber-bg); color: var(--amber); border-color: var(--amber-border); }
|
|
19281
19496
|
.badge.running .d { background: var(--amber); }
|
|
19497
|
+
/* A drift verdict is a diagnosis, not a broken test: amber, never fail-red.
|
|
19498
|
+
Without these the badge rendered as bare text next to the pill-shaped
|
|
19499
|
+
pass/fail ones, and read as a different kind of thing. */
|
|
19500
|
+
.badge.dr-found { background: var(--amber-bg); color: var(--amber); border-color: var(--amber-border); }
|
|
19501
|
+
.badge.dr-found .d { background: var(--amber); }
|
|
19502
|
+
.badge.dr-clean { background: var(--pass-bg); color: var(--pass); border-color: var(--pass-border); }
|
|
19503
|
+
.badge.dr-clean .d { background: var(--pass); }
|
|
19504
|
+
.badge.dr-unknown { background: var(--surface-3); color: var(--muted); border-color: var(--border); }
|
|
19505
|
+
.badge.dr-unknown .d { background: var(--muted); }
|
|
19282
19506
|
.badge-live, .badge-det { display: inline-flex; align-items: center; padding: 2px 8px; border-radius: var(--radius-sm); font-size: 11px; font-weight: 600; border: 1px solid transparent; }
|
|
19283
19507
|
.badge-live { background: var(--violet-bg); color: var(--violet); border-color: var(--violet-border); }
|
|
19284
19508
|
.badge-det { background: var(--surface-3); color: var(--muted); border-color: var(--border); }
|
|
@@ -19290,8 +19514,11 @@ const CSS = `
|
|
|
19290
19514
|
order decides). One amber look for every drift label chip — a label chip
|
|
19291
19515
|
is a finding, not a severity, so it does not split into fail-red/amber
|
|
19292
19516
|
the way the old errors/warnings counts did. */
|
|
19293
|
-
.chip.kind-chip { color: var(--violet); background: var(--violet-bg); border-color: var(--violet-border); font-family: var(--font);
|
|
19294
|
-
.chip.drift-count-chip { color: var(--amber); background: var(--amber-bg); border-color: var(--amber-border);
|
|
19517
|
+
.chip.kind-chip { color: var(--violet); background: var(--violet-bg); border-color: var(--violet-border); font-family: var(--font); }
|
|
19518
|
+
.chip.drift-count-chip { color: var(--amber); background: var(--amber-bg); border-color: var(--amber-border); }
|
|
19519
|
+
/* Prose, not an identifier — and neutral, since it qualifies the label chip
|
|
19520
|
+
beside it rather than claiming a severity of its own. */
|
|
19521
|
+
.chip.spec-change-chip { font-family: var(--font); }
|
|
19295
19522
|
.drift-meta-box { display: flex; flex-direction: column; gap: 4px; }
|
|
19296
19523
|
/* The chips carry their own margin for the run list, where they sit inline
|
|
19297
19524
|
after other chips. Here the container owns the spacing, so the margin only
|
|
@@ -19765,6 +19992,7 @@ const CLIENT_JS = `
|
|
|
19765
19992
|
"meta.drift": "Drift",
|
|
19766
19993
|
"diag.cause": "Cause", "diag.fix": "Fix",
|
|
19767
19994
|
"diag.surface": "Surface", "diag.surface.spec": "spec", "diag.surface.generated": "generated code",
|
|
19995
|
+
"diag.specChangeKind.FEATURE_REMOVED": "feature gone", "diag.specChangeKind.BEHAVIOUR_CHANGED": "behaviour changed",
|
|
19768
19996
|
"acc.reasoning": "Reasoning", "acc.evidence": "Evidence", "acc.steps": "Live run steps",
|
|
19769
19997
|
"acc.assertions": "Assertions",
|
|
19770
19998
|
"acc.artifacts": "Artifacts",
|
|
@@ -19773,7 +20001,7 @@ const CLIENT_JS = `
|
|
|
19773
20001
|
"spec.kind.live": "Live", "spec.kind.det": "Deterministic",
|
|
19774
20002
|
"det.steps": "Steps",
|
|
19775
20003
|
"det.noEvidence": "No step screenshots:",
|
|
19776
|
-
"kind.run": "Test run", "kind.drift": "Drift audit",
|
|
20004
|
+
"kind.run": "Test run", "kind.drift": "Drift audit", "kind.record": "Recording",
|
|
19777
20005
|
"drift.summary.ratio": "{found} of {total} specs",
|
|
19778
20006
|
"drift.clean": "No drift issues",
|
|
19779
20007
|
"status.passed": "passed", "status.failed": "failed", "status.skipped": "skipped", "status.running": "running",
|
|
@@ -19923,6 +20151,7 @@ const CLIENT_JS = `
|
|
|
19923
20151
|
"meta.drift": "ドリフト",
|
|
19924
20152
|
"diag.cause": "原因", "diag.fix": "対処",
|
|
19925
20153
|
"diag.surface": "対象", "diag.surface.spec": "spec", "diag.surface.generated": "生成コード",
|
|
20154
|
+
"diag.specChangeKind.FEATURE_REMOVED": "機能が無い", "diag.specChangeKind.BEHAVIOUR_CHANGED": "振る舞いが変わった",
|
|
19926
20155
|
"acc.reasoning": "推論", "acc.evidence": "根拠", "acc.steps": "実行ステップ",
|
|
19927
20156
|
"acc.assertions": "アサーション",
|
|
19928
20157
|
"acc.artifacts": "成果物",
|
|
@@ -19931,7 +20160,7 @@ const CLIENT_JS = `
|
|
|
19931
20160
|
"spec.kind.live": "ライブ", "spec.kind.det": "決定的",
|
|
19932
20161
|
"det.steps": "ステップ",
|
|
19933
20162
|
"det.noEvidence": "ステップのスクリーンショットなし:",
|
|
19934
|
-
"kind.run": "テスト実行", "kind.drift": "ドリフト監査",
|
|
20163
|
+
"kind.run": "テスト実行", "kind.drift": "ドリフト監査", "kind.record": "収録",
|
|
19935
20164
|
"drift.summary.ratio": "{found} / {total} スペック",
|
|
19936
20165
|
"drift.clean": "ドリフトの問題なし",
|
|
19937
20166
|
"status.passed": "合格", "status.failed": "失敗", "status.skipped": "スキップ", "status.running": "実行中",
|
|
@@ -20310,6 +20539,21 @@ const CLIENT_JS = `
|
|
|
20310
20539
|
return run.kind === "drift" ? driftFoundBadge(driftRunState(run), "drift.run.") : statusBadge(run.status);
|
|
20311
20540
|
}
|
|
20312
20541
|
|
|
20542
|
+
// Which command left the run, and whether its spec counts are a tally of
|
|
20543
|
+
// what was verified. A recording's rows are the specs it wrote, not specs it
|
|
20544
|
+
// checked, so "1 / 1 passed" and a full meter would claim a test result.
|
|
20545
|
+
// A kind from a newer hub keeps the generic label but is read the same
|
|
20546
|
+
// cautious way, since nothing here knows what its counts mean.
|
|
20547
|
+
var KINDS = {
|
|
20548
|
+
run: { label: "kind.run", verifies: true },
|
|
20549
|
+
drift: { label: "kind.drift", verifies: true },
|
|
20550
|
+
record: { label: "kind.record", verifies: false },
|
|
20551
|
+
};
|
|
20552
|
+
function kindOf(kind) { return KINDS[kind] || { label: "kind.run", verifies: false }; }
|
|
20553
|
+
function kindChip(kind) {
|
|
20554
|
+
return el("span", "chip kind-chip", t(kindOf(kind).label));
|
|
20555
|
+
}
|
|
20556
|
+
|
|
20313
20557
|
// A run's Claude spend, in the same $x.xxxx form as the per-step badge.
|
|
20314
20558
|
// A run that billed nothing, and one stored before costs were recorded, both
|
|
20315
20559
|
// arrive as a non-number — printing $0.0000 would claim a measured zero.
|
|
@@ -20580,12 +20824,10 @@ const CLIENT_JS = `
|
|
|
20580
20824
|
runCell.appendChild(el("div", "runid", r.id.slice(0, 8)));
|
|
20581
20825
|
var sub = el("div", "subline");
|
|
20582
20826
|
sub.appendChild(ciBadge(r));
|
|
20827
|
+
sub.appendChild(kindChip(r.kind));
|
|
20583
20828
|
if (r.kind === "drift") {
|
|
20584
|
-
sub.appendChild(el("span", "chip kind-chip", t("kind.drift")));
|
|
20585
20829
|
var rowDrift = driftSummary(r);
|
|
20586
20830
|
if (rowDrift) driftChips(rowDrift).forEach(function (chip) { sub.appendChild(chip); });
|
|
20587
|
-
} else {
|
|
20588
|
-
sub.appendChild(el("span", "chip kind-chip", t("kind.run")));
|
|
20589
20831
|
}
|
|
20590
20832
|
runCell.appendChild(sub);
|
|
20591
20833
|
tr.appendChild(runCell);
|
|
@@ -20602,29 +20844,33 @@ const CLIENT_JS = `
|
|
|
20602
20844
|
statusCell.appendChild(runStatusBadge(r));
|
|
20603
20845
|
tr.appendChild(statusCell);
|
|
20604
20846
|
|
|
20605
|
-
// A drift row counts what the audit found, not what "passed" — the same
|
|
20606
|
-
// ratio its detail page shows. Reading passed/total here printed a
|
|
20607
|
-
// different number for the same run in the two places you would compare.
|
|
20608
|
-
var rowDriftSummary = r.kind === "drift" ? driftSummary(r) : null;
|
|
20609
|
-
var found = rowDriftSummary
|
|
20610
|
-
? rowDriftSummary.testDrift + rowDriftSummary.specChange + rowDriftSummary.unknown
|
|
20611
|
-
: null;
|
|
20612
|
-
var num = rowDriftSummary
|
|
20613
|
-
? found + " / " + rowDriftSummary.specs
|
|
20614
|
-
: r.specs.passed + " / " + r.specs.total;
|
|
20615
|
-
var fillTotal = rowDriftSummary ? rowDriftSummary.specs : r.specs.total;
|
|
20616
|
-
var fillPart = rowDriftSummary ? found : r.specs.passed;
|
|
20617
|
-
|
|
20618
20847
|
var specsCell = document.createElement("td");
|
|
20619
|
-
|
|
20620
|
-
|
|
20621
|
-
|
|
20622
|
-
|
|
20623
|
-
|
|
20624
|
-
|
|
20625
|
-
|
|
20626
|
-
|
|
20627
|
-
|
|
20848
|
+
if (!kindOf(r.kind).verifies) {
|
|
20849
|
+
specsCell.appendChild(el("span", "muted", "—"));
|
|
20850
|
+
} else {
|
|
20851
|
+
// A drift row counts what the audit found, not what "passed" — the same
|
|
20852
|
+
// ratio its detail page shows. Reading passed/total here printed a
|
|
20853
|
+
// different number for the same run in the two places you would compare.
|
|
20854
|
+
var rowDriftSummary = r.kind === "drift" ? driftSummary(r) : null;
|
|
20855
|
+
var found = rowDriftSummary
|
|
20856
|
+
? rowDriftSummary.testDrift + rowDriftSummary.specChange + rowDriftSummary.unknown
|
|
20857
|
+
: null;
|
|
20858
|
+
var num = rowDriftSummary
|
|
20859
|
+
? found + " / " + rowDriftSummary.specs
|
|
20860
|
+
: r.specs.passed + " / " + r.specs.total;
|
|
20861
|
+
var fillTotal = rowDriftSummary ? rowDriftSummary.specs : r.specs.total;
|
|
20862
|
+
var fillPart = rowDriftSummary ? found : r.specs.passed;
|
|
20863
|
+
|
|
20864
|
+
var specsWrap = el("div", "specs");
|
|
20865
|
+
var meter = el("span", "meter" + (rowDriftSummary ? " drift" : ""));
|
|
20866
|
+
var pct = fillTotal > 0 ? Math.round((fillPart / fillTotal) * 100) : 0;
|
|
20867
|
+
var bar = el("i");
|
|
20868
|
+
bar.style.width = pct + "%";
|
|
20869
|
+
meter.appendChild(bar);
|
|
20870
|
+
specsWrap.appendChild(meter);
|
|
20871
|
+
specsWrap.appendChild(el("span", "num muted", num));
|
|
20872
|
+
specsCell.appendChild(specsWrap);
|
|
20873
|
+
}
|
|
20628
20874
|
tr.appendChild(specsCell);
|
|
20629
20875
|
|
|
20630
20876
|
tr.appendChild(el("td", "muted num", costText(r.costUsd)));
|
|
@@ -20663,7 +20909,7 @@ const CLIENT_JS = `
|
|
|
20663
20909
|
// What kind of run this is, said once. The spec cards below used to repeat
|
|
20664
20910
|
// it per row, which read as "this spec was drift-audited" — a property of
|
|
20665
20911
|
// the run described as if it varied spec to spec. Same chip as the run list.
|
|
20666
|
-
sub.appendChild(
|
|
20912
|
+
sub.appendChild(kindChip(run.kind));
|
|
20667
20913
|
idblock.appendChild(sub);
|
|
20668
20914
|
head.appendChild(idblock);
|
|
20669
20915
|
|
|
@@ -20697,7 +20943,7 @@ const CLIENT_JS = `
|
|
|
20697
20943
|
driftBox.appendChild(el("div", "muted", ratio));
|
|
20698
20944
|
metaItem(t("meta.drift"), driftBox);
|
|
20699
20945
|
}
|
|
20700
|
-
} else {
|
|
20946
|
+
} else if (kindOf(run.kind).verifies) {
|
|
20701
20947
|
metaItem(t("meta.specs"), run.specs.passed + " / " + run.specs.total + " " + t("meta.passed"));
|
|
20702
20948
|
}
|
|
20703
20949
|
// Everything this run spent on Claude — live browsing, triage, the audit a
|
|
@@ -20806,6 +21052,12 @@ const CLIENT_JS = `
|
|
|
20806
21052
|
var a = r.analysis;
|
|
20807
21053
|
var head = el("div", "analysis-head");
|
|
20808
21054
|
head.appendChild(labelChip(a.label));
|
|
21055
|
+
// Which repair a SPEC_CHANGE needs — delete the spec, or rewrite it. The
|
|
21056
|
+
// label is re-checked rather than trusted: this chip means nothing beside
|
|
21057
|
+
// any other one, however the row reached the browser.
|
|
21058
|
+
if (a.label === "SPEC_CHANGE" && a.specChangeKind) {
|
|
21059
|
+
head.appendChild(el("span", "chip spec-change-chip", t("diag.specChangeKind." + a.specChangeKind)));
|
|
21060
|
+
}
|
|
20809
21061
|
head.appendChild(el("span", "conf", Math.round(a.confidence * 100) + "%"));
|
|
20810
21062
|
wrap.appendChild(head);
|
|
20811
21063
|
var kv = el("div", "analysis-kv");
|
|
@@ -22000,8 +22252,12 @@ const CLIENT_JS = `
|
|
|
22000
22252
|
// runs and no deploys to judge). One runs page answers both: runId -> CI URL,
|
|
22001
22253
|
// and the profiles a run was actually recorded under. A run pushed without a
|
|
22002
22254
|
// profile lands in "default", exactly as the ledger stores it.
|
|
22255
|
+
//
|
|
22256
|
+
// Only the two kinds a ledger entry can point at, so recordings — which
|
|
22257
|
+
// advance no ledger and are never looked up here — cannot crowd them out of
|
|
22258
|
+
// the window.
|
|
22003
22259
|
function fetchRunIndex() {
|
|
22004
|
-
return apiFetch("/api/v1/runs?project=" + encodeURIComponent(state.project) + "&limit=200")
|
|
22260
|
+
return apiFetch("/api/v1/runs?project=" + encodeURIComponent(state.project) + "&kind=run,drift&limit=200")
|
|
22005
22261
|
.then(function (data) {
|
|
22006
22262
|
var urls = {};
|
|
22007
22263
|
var profiles = [];
|
|
@@ -23734,7 +23990,8 @@ function createLearningWorker(deps) {
|
|
|
23734
23990
|
const runLimit = job.input.runLimit > 0 ? job.input.runLimit : DEFAULT_RUN_LIMIT;
|
|
23735
23991
|
const runs = await storage.runs.list({
|
|
23736
23992
|
project: job.project,
|
|
23737
|
-
limit: runLimit
|
|
23993
|
+
limit: runLimit,
|
|
23994
|
+
kinds: ["run"]
|
|
23738
23995
|
});
|
|
23739
23996
|
const cases = [];
|
|
23740
23997
|
let excluded = 0;
|
|
@@ -23933,6 +24190,8 @@ function registerRoutes(router, config, queue) {
|
|
|
23933
24190
|
router.get("/api/v1/projects/:project/audit-needed", createGetAuditNeedHandler(storage));
|
|
23934
24191
|
router.post("/api/v1/projects/:project/locks", createAcquireLocksHandler(storage));
|
|
23935
24192
|
router.delete("/api/v1/projects/:project/locks", createReleaseLocksHandler(storage));
|
|
24193
|
+
router.get("/api/v1/projects/:project/acks/:name", createGetAckHandler(storage));
|
|
24194
|
+
router.put("/api/v1/projects/:project/acks/:name", createPutAckHandler(storage));
|
|
23936
24195
|
const sessionConfig = {
|
|
23937
24196
|
store: storage.sessions,
|
|
23938
24197
|
encryptionKey: config.encryptionKey
|
|
@@ -23967,6 +24226,16 @@ function registerRoutes(router, config, queue) {
|
|
|
23967
24226
|
}
|
|
23968
24227
|
//#endregion
|
|
23969
24228
|
//#region src/hub/core/storage/file/fs-helpers.ts
|
|
24229
|
+
/**
|
|
24230
|
+
* Defense-in-depth for a name this layer joins into a file path. The API's
|
|
24231
|
+
* `SAFE_SEGMENT` (api/validate.ts) is deliberately stricter — it also fixes a
|
|
24232
|
+
* charset and a length — and stays the rule for what a client may name; this
|
|
24233
|
+
* only refuses traversal, so it also covers callers that never came through
|
|
24234
|
+
* HTTP (tests, a library embedding the hub).
|
|
24235
|
+
*/
|
|
24236
|
+
function assertSafeName(value, label) {
|
|
24237
|
+
if (value.length === 0 || value === "." || value === ".." || value.includes("/") || value.includes("\\")) throw new Error(`invalid ${label}: must be a bare name without path separators or '..'`);
|
|
24238
|
+
}
|
|
23970
24239
|
/** Read and JSON-parse a file, returning `null` when it doesn't exist. Malformed JSON throws. */
|
|
23971
24240
|
async function readJson(path) {
|
|
23972
24241
|
let raw;
|
|
@@ -24106,6 +24375,7 @@ function isNotFound(err) {
|
|
|
24106
24375
|
* drift-ledger/<project>/<branch>.json (DriftLedger, no profile)
|
|
24107
24376
|
* deploys/<project>/<profile>/log.json (DeployLog, ring-buffered)
|
|
24108
24377
|
* deploys/<project>/<profile>/touch.json (SpecTouchIndex derived from the log)
|
|
24378
|
+
* acks/<project>/<profile>/<name>.json (Ack: a consumer's acted-on keys)
|
|
24109
24379
|
*
|
|
24110
24380
|
* IDs and names are validated by their callers (run ids are server-minted
|
|
24111
24381
|
* UUIDs; project/profile/name come from validated request params) before
|
|
@@ -24197,6 +24467,43 @@ function deployTouchIndexPath(root, project, profile) {
|
|
|
24197
24467
|
function specLocksPath(root, project, profile) {
|
|
24198
24468
|
return join(root, "locks", project, profile, "locks.json");
|
|
24199
24469
|
}
|
|
24470
|
+
function ackPath(root, project, profile, name) {
|
|
24471
|
+
return join(root, "acks", project, profile, `${name}.json`);
|
|
24472
|
+
}
|
|
24473
|
+
//#endregion
|
|
24474
|
+
//#region src/hub/core/storage/file/ack-store.ts
|
|
24475
|
+
function assertSafeKey(project, profile, name) {
|
|
24476
|
+
assertSafeName(project, "project");
|
|
24477
|
+
assertSafeName(profile, "profile");
|
|
24478
|
+
assertSafeName(name, "name");
|
|
24479
|
+
}
|
|
24480
|
+
/**
|
|
24481
|
+
* Ack storage: one JSON document per (project, profile, name). A write
|
|
24482
|
+
* replaces the document outright, so unlike the ledgers there is no
|
|
24483
|
+
* read-modify-write to serialize — but it still goes through `writeJson`'s
|
|
24484
|
+
* temp-then-rename, so a concurrent reader never sees a half-written set.
|
|
24485
|
+
*/
|
|
24486
|
+
function createFileAckStore(root) {
|
|
24487
|
+
return {
|
|
24488
|
+
async get(project, profile, name) {
|
|
24489
|
+
assertSafeKey(project, profile, name);
|
|
24490
|
+
const parsed = AckSchema.safeParse(await readJson(ackPath(root, project, profile, name)));
|
|
24491
|
+
return parsed.success ? parsed.data : {
|
|
24492
|
+
keys: [],
|
|
24493
|
+
at: null
|
|
24494
|
+
};
|
|
24495
|
+
},
|
|
24496
|
+
async put(project, profile, name, keys) {
|
|
24497
|
+
assertSafeKey(project, profile, name);
|
|
24498
|
+
const ack = {
|
|
24499
|
+
keys,
|
|
24500
|
+
at: (/* @__PURE__ */ new Date()).toISOString()
|
|
24501
|
+
};
|
|
24502
|
+
await writeJson(ackPath(root, project, profile, name), ack);
|
|
24503
|
+
return ack;
|
|
24504
|
+
}
|
|
24505
|
+
};
|
|
24506
|
+
}
|
|
24200
24507
|
//#endregion
|
|
24201
24508
|
//#region src/hub/core/storage/file/artifact-store.ts
|
|
24202
24509
|
/**
|
|
@@ -24367,14 +24674,6 @@ function createFileSpecLedgerStore(root) {
|
|
|
24367
24674
|
//#endregion
|
|
24368
24675
|
//#region src/hub/core/storage/file/perspectives-store.ts
|
|
24369
24676
|
/**
|
|
24370
|
-
* Defense-in-depth path validation: the HTTP layer already checks the project
|
|
24371
|
-
* segment, but this builds a file path from it, so it re-checks rather than
|
|
24372
|
-
* trusting callers. Mirrors the sibling prompt/secret stores.
|
|
24373
|
-
*/
|
|
24374
|
-
function assertSafeName$2(value, label) {
|
|
24375
|
-
if (value.length === 0 || value.includes("/") || value.includes("\\") || value.split(/[\\/]/).includes("..") || value === ".") throw new Error(`invalid ${label}: must be a bare name without path separators or '..'`);
|
|
24376
|
-
}
|
|
24377
|
-
/**
|
|
24378
24677
|
* Perspectives storage: one JSON document per project, plain UTF-8 with no
|
|
24379
24678
|
* encryption (an inventory of what is tested is not a secret). No meta file —
|
|
24380
24679
|
* the document's own `generatedAt` is its timestamp.
|
|
@@ -24382,19 +24681,19 @@ function assertSafeName$2(value, label) {
|
|
|
24382
24681
|
function createFilePerspectivesStore(root) {
|
|
24383
24682
|
return {
|
|
24384
24683
|
async put(project, blob) {
|
|
24385
|
-
assertSafeName
|
|
24684
|
+
assertSafeName(project, "project");
|
|
24386
24685
|
await writeBytes(perspectivesPath(root, project), blob);
|
|
24387
24686
|
},
|
|
24388
24687
|
async get(project) {
|
|
24389
|
-
assertSafeName
|
|
24688
|
+
assertSafeName(project, "project");
|
|
24390
24689
|
return readBytesOrNull(perspectivesPath(root, project));
|
|
24391
24690
|
},
|
|
24392
24691
|
async update(project, mutate) {
|
|
24393
|
-
assertSafeName
|
|
24692
|
+
assertSafeName(project, "project");
|
|
24394
24693
|
await updateJson(perspectivesPath(root, project), mutate);
|
|
24395
24694
|
},
|
|
24396
24695
|
async delete(project) {
|
|
24397
|
-
assertSafeName
|
|
24696
|
+
assertSafeName(project, "project");
|
|
24398
24697
|
await removePath(perspectivesPath(root, project));
|
|
24399
24698
|
}
|
|
24400
24699
|
};
|
|
@@ -24402,15 +24701,6 @@ function createFilePerspectivesStore(root) {
|
|
|
24402
24701
|
//#endregion
|
|
24403
24702
|
//#region src/hub/core/storage/file/prompt-store.ts
|
|
24404
24703
|
/**
|
|
24405
|
-
* Defense-in-depth path validation: the HTTP layer already checks project/name,
|
|
24406
|
-
* but this builds file paths from them, so it re-checks rather than trusting
|
|
24407
|
-
* callers. (Which names are allowed at all is the handler's job — this only
|
|
24408
|
-
* guards against path traversal.)
|
|
24409
|
-
*/
|
|
24410
|
-
function assertSafeName$1(value, label) {
|
|
24411
|
-
if (value.length === 0 || value.includes("/") || value.includes("\\") || value.split(/[\\/]/).includes("..") || value === ".") throw new Error(`invalid ${label}: must be a bare name without path separators or '..'`);
|
|
24412
|
-
}
|
|
24413
|
-
/**
|
|
24414
24704
|
* Prompt storage, project-scoped (not per-profile — prompts are project-wide).
|
|
24415
24705
|
* The blob is plain UTF-8 text (Markdown or custom prompt JSON) with no encryption,
|
|
24416
24706
|
* so this works whether or not `CCQA_HUB_ENCRYPTION_KEY` is configured.
|
|
@@ -24418,8 +24708,8 @@ function assertSafeName$1(value, label) {
|
|
|
24418
24708
|
function createFilePromptStore(root) {
|
|
24419
24709
|
return {
|
|
24420
24710
|
async put(project, name, blob, meta = {}) {
|
|
24421
|
-
assertSafeName
|
|
24422
|
-
assertSafeName
|
|
24711
|
+
assertSafeName(project, "project");
|
|
24712
|
+
assertSafeName(name, "name");
|
|
24423
24713
|
await writeJson(promptMetaPath(root, project, name), {
|
|
24424
24714
|
meta,
|
|
24425
24715
|
updatedAt: (/* @__PURE__ */ new Date()).toISOString()
|
|
@@ -24427,8 +24717,8 @@ function createFilePromptStore(root) {
|
|
|
24427
24717
|
await writeBytes(promptBlobPath(root, project, name), blob);
|
|
24428
24718
|
},
|
|
24429
24719
|
async get(project, name) {
|
|
24430
|
-
assertSafeName
|
|
24431
|
-
assertSafeName
|
|
24720
|
+
assertSafeName(project, "project");
|
|
24721
|
+
assertSafeName(name, "name");
|
|
24432
24722
|
const blob = await readBytesOrNull(promptBlobPath(root, project, name));
|
|
24433
24723
|
if (!blob) return null;
|
|
24434
24724
|
return {
|
|
@@ -24437,7 +24727,7 @@ function createFilePromptStore(root) {
|
|
|
24437
24727
|
};
|
|
24438
24728
|
},
|
|
24439
24729
|
async list(project) {
|
|
24440
|
-
assertSafeName
|
|
24730
|
+
assertSafeName(project, "project");
|
|
24441
24731
|
const names = (await listDirOrEmpty(promptProjectDir(root, project))).filter((f) => f.endsWith(".txt")).map((f) => f.slice(0, -4));
|
|
24442
24732
|
const out = [];
|
|
24443
24733
|
for (const name of names) {
|
|
@@ -24451,8 +24741,8 @@ function createFilePromptStore(root) {
|
|
|
24451
24741
|
return out;
|
|
24452
24742
|
},
|
|
24453
24743
|
async delete(project, name) {
|
|
24454
|
-
assertSafeName
|
|
24455
|
-
assertSafeName
|
|
24744
|
+
assertSafeName(project, "project");
|
|
24745
|
+
assertSafeName(name, "name");
|
|
24456
24746
|
await removePath(promptBlobPath(root, project, name));
|
|
24457
24747
|
await removePath(promptMetaPath(root, project, name));
|
|
24458
24748
|
},
|
|
@@ -24494,7 +24784,7 @@ function createFileRunStore(root) {
|
|
|
24494
24784
|
};
|
|
24495
24785
|
});
|
|
24496
24786
|
},
|
|
24497
|
-
async list({ project, branch, status, limit }) {
|
|
24787
|
+
async list({ project, branch, status, kinds, limit }) {
|
|
24498
24788
|
const ids = await listSubdirsOrEmpty(runsDir(root));
|
|
24499
24789
|
const runs = [];
|
|
24500
24790
|
for (const id of ids) {
|
|
@@ -24503,6 +24793,7 @@ function createFileRunStore(root) {
|
|
|
24503
24793
|
if (project !== void 0 && run.project !== project) continue;
|
|
24504
24794
|
if (branch !== void 0 && run.branch !== branch) continue;
|
|
24505
24795
|
if (status !== void 0 && run.status !== status) continue;
|
|
24796
|
+
if (kinds !== void 0 && !kinds.includes(run.kind)) continue;
|
|
24506
24797
|
runs.push(run);
|
|
24507
24798
|
}
|
|
24508
24799
|
runs.sort((a, b) => b.createdAt.localeCompare(a.createdAt));
|
|
@@ -24521,14 +24812,6 @@ function createFileRunStore(root) {
|
|
|
24521
24812
|
}
|
|
24522
24813
|
//#endregion
|
|
24523
24814
|
//#region src/hub/core/storage/file/secret-store.ts
|
|
24524
|
-
/**
|
|
24525
|
-
* Defense-in-depth: project/profile/name are expected to already be validated
|
|
24526
|
-
* by the HTTP layer (`requireSafeSegment`), but this store builds file paths
|
|
24527
|
-
* directly from them, so it re-checks rather than trusting callers blindly.
|
|
24528
|
-
*/
|
|
24529
|
-
function assertSafeName(value, label) {
|
|
24530
|
-
if (value.length === 0 || value.includes("/") || value.includes("\\") || value.split(/[\\/]/).includes("..") || value === ".") throw new Error(`invalid ${label}: must be a bare name without path separators or '..'`);
|
|
24531
|
-
}
|
|
24532
24815
|
function assertSafeScope(scope) {
|
|
24533
24816
|
assertSafeName(scope.project, "project");
|
|
24534
24817
|
assertSafeName(scope.profile, "profile");
|
|
@@ -24622,7 +24905,8 @@ function createFileHubStorage(dataDir) {
|
|
|
24622
24905
|
ledger: createFileSpecLedgerStore(dataDir),
|
|
24623
24906
|
driftLedger: createFileDriftLedgerStore(dataDir),
|
|
24624
24907
|
deploys: createFileDeployStore(dataDir),
|
|
24625
|
-
locks: createFileLockStore(dataDir)
|
|
24908
|
+
locks: createFileLockStore(dataDir),
|
|
24909
|
+
acks: createFileAckStore(dataDir)
|
|
24626
24910
|
};
|
|
24627
24911
|
}
|
|
24628
24912
|
//#endregion
|