fix(pricing): add openai:gpt-5.2 canonical entry for the new default slot A

DEFAULT_SLOTS slot A moved to openai:gpt-5.2, which had no CANONICAL_PRICING entry — estimateCost silently dropped slot A from the --max-usd pre-flight and est_cost_usd audit rows (~1/3 under-count on the default panel). Rates from the OpenAI recipe chat touchpoint (verified 2026-04-20). Also refresh the --slot-a-model help text default and pin a pricing-presence assertion in the DEFAULT_SLOTS consistency test. Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
fix(autopilot,eval): nightly quality probe enable path works end-to-end + wire conversation-parser probe
2026-08-17 10:22:34 +00:00 · 2026-07-22 11:18:09 -07:00 · 2026-07-21 14:24:41 -07:00
31 changed files with 795 additions and 621 deletions
@@ -527,6 +527,9 @@ export async function runAutopilot(engine: BrainEngine, args: string[]) {
  process.on('SIGINT',  () => { void shutdown('SIGINT'); });

  let consecutiveErrors = 0;
+  // Parser-probe fixture warning is once-per-process, not once-per-cycle
+  // (compiled-binary installs have no source tree; don't spam the log).
+  let parserProbeFixtureWarned = false;
  // v0.37.7.0 #1162 — counter for consecutive reconnect failures.
  // Reset on every successful health probe or reconnect. Threshold
  // controlled by GBRAIN_AUTOPILOT_MAX_RECONNECT_FAILS env (default 30).
@@ -1073,17 +1076,36 @@ export async function runAutopilot(engine: BrainEngine, args: string[]) {
    // loop. Probe runs even when cycleOk=false (probe may surface signal
    // explaining why the cycle is failing).
    try {
-      const probeEnabled = cfg?.autopilot?.nightly_quality_probe?.enabled === true;
+      const { resolveProbeEnabled, resolveProbeMaxUsd, runNightlyQualityProbe } = await import('../core/cycle/nightly-quality-probe.ts');
+      // Dual-plane read: `gbrain config set` (what the doctor enable hint
+      // prints) writes the DB plane; ~/.gbrain/config.json is the fallback.
+      let dbEnabled: string | null = null;
+      let dbMaxUsd: string | null = null;
+      try {
+        dbEnabled = await engine.getConfig('autopilot.nightly_quality_probe.enabled');
+        dbMaxUsd = await engine.getConfig('autopilot.nightly_quality_probe.max_usd');
+      } catch { /* DB unavailable → file plane only */ }
+      const probeEnabled = resolveProbeEnabled(dbEnabled, cfg?.autopilot?.nightly_quality_probe?.enabled);
      if (probeEnabled) {
-        const { runNightlyQualityProbe } = await import('../core/cycle/nightly-quality-probe.ts');
        const { runLongMemEvalForProbe, runCrossModalBatchForProbe } = await import('../core/cycle/nightly-probe-adapters.ts');
        const { isAvailable } = await import('../core/ai/gateway.ts');
-        const maxUsd = Number(cfg?.autopilot?.nightly_quality_probe?.max_usd ?? 5);
+        const { existsSync } = await import('node:fs');
+        const { fileURLToPath } = await import('node:url');
+        const { join } = await import('node:path');
+        const maxUsd = resolveProbeMaxUsd(dbMaxUsd, cfg?.autopilot?.nightly_quality_probe?.max_usd);
+        // The committed fixture (test/fixtures/longmemeval-nightly.jsonl)
+        // lives in the gbrain PACKAGE, not the brain repo — repoPath is
+        // sync.repo_path (the user's brain), where the fixture never
+        // exists, so the probe error'd on every real install. Resolve the
+        // package root from the module location; keep repoPath as the
+        // fallback for setups that vendor the fixture into the brain repo.
+        const pkgRoot = fileURLToPath(new URL('../..', import.meta.url));
+        const fixtureAtPkgRoot = existsSync(join(pkgRoot, 'test', 'fixtures', 'longmemeval-nightly.jsonl'));
        await runNightlyQualityProbe({
          isEnabled: () => true, // already gated above; phase re-checks for defense-in-depth
          hasEmbeddingProvider: () => isAvailable('embedding'),
          resolveMaxUsd: () => maxUsd,
-          resolveRepoRoot: () => repoPath ?? gbrainHomePath('.'),
+          resolveRepoRoot: () => (fixtureAtPkgRoot ? pkgRoot : repoPath ?? gbrainHomePath('.')),
          runLongMemEval: runLongMemEvalForProbe,
          runCrossModalBatch: runCrossModalBatchForProbe,
          now: () => new Date(),
@@ -1095,6 +1117,62 @@ export async function runAutopilot(engine: BrainEngine, args: string[]) {
      // informational; autopilot loop continues.
    }

+    // 4.6 — Nightly conversation-parser probe (v0.41.16.0 phase module;
+    // the scheduler wire-up was deferred at ship and is added here). Same
+    // posture as 4.5: the phase owns its gates (enabled/mode-gate, LLM
+    // key), the wiring owns invocation + the audit row, and a probe
+    // failure NEVER crashes the autopilot loop. Per D10 the probe is
+    // default-ON for search.mode=tokenmax, opt-in otherwise.
+    try {
+      const { runConversationParserNightlyProbe } = await import('../core/conversation-parser/nightly-probe.ts');
+      const { logParserProbeEvent, parserProbeRanWithin } = await import('../core/audit-parser-probe.ts');
+      const { isAvailable } = await import('../core/ai/gateway.ts');
+      const { existsSync } = await import('node:fs');
+      const { fileURLToPath } = await import('node:url');
+      const { join } = await import('node:path');
+      // Flag reads dual-plane: the DB row (`gbrain config set …`) wins,
+      // ~/.gbrain/config.json is the fallback. search.mode lives on the
+      // DB plane only (mode.ts owns it).
+      let parserDbEnabled: string | null = null;
+      let dbSearchMode: string | null = null;
+      try {
+        parserDbEnabled = await engine.getConfig('autopilot.conversation_parser_probe.enabled');
+        dbSearchMode = await engine.getConfig('search.mode');
+      } catch { /* DB unavailable → file plane only */ }
+      const parserEnabled = parserDbEnabled != null
+        ? parserDbEnabled === 'true'
+        : cfg?.autopilot?.conversation_parser_probe?.enabled === true;
+      const searchMode = dbSearchMode ?? '';
+      // Fixtures are committed in the gbrain package (test/fixtures/…),
+      // NOT the brain repo — resolve from the module location. Compiled
+      // binaries carry no source tree: skip quietly instead of writing
+      // failure rows that would flip doctor to WARN on every binary install.
+      const pkgRoot = fileURLToPath(new URL('../..', import.meta.url));
+      const fixturePath = join(pkgRoot, 'test', 'fixtures', 'conversation-formats', 'all.jsonl');
+      const adversarialPath = join(pkgRoot, 'test', 'fixtures', 'conversation-formats', 'adversarial.jsonl');
+      const shouldInvoke = parserEnabled || searchMode === 'tokenmax';
+      if (shouldInvoke && existsSync(fixturePath) && existsSync(adversarialPath)) {
+        const result = await runConversationParserNightlyProbe({
+          isEnabled: () => parserEnabled,
+          searchMode: () => searchMode,
+          hasLlmKey: () => isAvailable('chat'),
+          resolveFixturePath: () => fixturePath,
+          resolveAdversarialPath: () => adversarialPath,
+          now: () => new Date(),
+          shouldSkipForRateLimit: () => parserProbeRanWithin(24 * 60 * 60 * 1000),
+        });
+        // rate_limited is a non-run: the loop ticks every few minutes, so
+        // logging every skip would flood the audit file with no-signal rows.
+        if (result.outcome !== 'rate_limited') logParserProbeEvent(result);
+      } else if (shouldInvoke && !parserProbeFixtureWarned) {
+        parserProbeFixtureWarned = true;
+        console.error(`[parser-probe] fixtures not found under ${pkgRoot}; skipping (probe needs a source-checkout install)`);
+      }
+    } catch (e) {
+      logError('autopilot.parser_probe', e);
+      // Informational, like 4.5: do NOT bump consecutiveErrors.
+    }
+
    // Wait for next cycle
    await new Promise(r => setTimeout(r, interval * 1000));
  }
@@ -2960,6 +2960,54 @@ function _resolveSyncFreshnessHours(varName: string, fallback: number): number {
 * branch (disabled / enabled-no-events / enabled-all-pass / enabled-with-failures)
 * without spinning up the audit JSONL or a real config file.
 */
+/**
+ * Pure function form of the conversation_parser_probe_health check.
+ * Mirrors computeNightlyQualityProbeHealthCheck: skip-with-hint when the
+ * probe is off and silent, surface the last 7 days of audit events when
+ * it has run, WARN on any non-pass outcome.
+ *
+ * `effectiveEnabled` folds the D10 mode-gate in: explicitly enabled OR
+ * search.mode=tokenmax (where the probe is default-on).
+ */
+export function computeConversationParserProbeHealthCheck(
+  effectiveEnabled: boolean,
+  events: ReadonlyArray<{ outcome: string; ts: string; reason?: string }>,
+): Check {
+  const name = 'conversation_parser_probe_health';
+  if (!effectiveEnabled && events.length === 0) {
+    return {
+      name,
+      status: 'ok',
+      message:
+        'disabled (opt-in; default-on only for search.mode=tokenmax). Enable with: ' +
+        '`gbrain config set autopilot.conversation_parser_probe.enabled true`',
+    };
+  }
+  if (events.length === 0) {
+    return {
+      name,
+      status: 'ok',
+      message: 'enabled but no probe events in the last 7 days (next run by autopilot; fixtures require a source-checkout install).',
+    };
+  }
+  const bad = events.filter(e => e.outcome !== 'pass');
+  const latest = events[events.length - 1]!;
+  if (bad.length > 0) {
+    return {
+      name,
+      status: 'warn',
+      message:
+        `${bad.length}/${events.length} probe run(s) in the last 7 days did not pass; ` +
+        `latest: ${latest.outcome}${latest.reason ? ` (${latest.reason})` : ''}`,
+    };
+  }
+  return {
+    name,
+    status: 'ok',
+    message: `${events.length} probe run(s) in the last 7 days, all pass (latest ${latest.ts}).`,
+  };
+}
+
 export function computeNightlyQualityProbeHealthCheck(
  probeEnabled: boolean,
  events: ReadonlyArray<{ outcome: string; ts: string; detail?: string }>,
@@ -4843,10 +4891,17 @@ export async function buildChecks(
  try {
    const { readRecentQualityProbeEvents } = await import('../core/audit-quality-probe.ts');
    const { loadConfig } = await import('../core/config.ts');
+    const { resolveProbeEnabled } = await import('../core/cycle/nightly-quality-probe.ts');
    let probeEnabled = false;
    try {
+      // Dual-plane read, matching the autopilot gate: the DB row (what the
+      // enable hint's `gbrain config set` writes) wins; file plane fallback.
+      let dbVal: string | null = null;
+      try {
+        dbVal = engine ? await engine.getConfig('autopilot.nightly_quality_probe.enabled') : null;
+      } catch { /* DB unavailable → file plane only */ }
      const cfg = loadConfig();
-      probeEnabled = Boolean((cfg as any)?.autopilot?.nightly_quality_probe?.enabled);
+      probeEnabled = resolveProbeEnabled(dbVal, (cfg as any)?.autopilot?.nightly_quality_probe?.enabled);
    } catch { /* config unavailable → treat as disabled */ }
    const events = readRecentQualityProbeEvents(7);
    const check = computeNightlyQualityProbeHealthCheck(probeEnabled, events);
@@ -5030,19 +5085,29 @@ export async function buildChecks(

  // 3d.5 v0.41.13.0 — conversation_parser_probe_health. Mode-gated
  // per D10: ON when search.mode=tokenmax, opt-in for other modes.
-  // Surface the last 7 days of nightly-probe events; warn on FAIL /
-  // BUDGET_EXCEEDED / adversarial_false_positive.
-  //
-  // v0.41.13.0 ships the probe as opt-in (autopilot wiring deferred
-  // to T7 in the cathedral plan); this check skips with an enable
-  // hint until the probe has at least one audit event written.
-  checks.push({
-    name: 'conversation_parser_probe_health',
-    status: 'ok',
-    message:
-      'Skipped (nightly probe is opt-in; enable with ' +
-      '`gbrain config set autopilot.conversation_parser_probe.enabled true`)',
-  });
+  // Surfaces the last 7 days of nightly-probe audit events; warn on any
+  // non-pass outcome (fail / budget_exceeded / adversarial_false_positive).
+  // (Until the autopilot wire-up this was a hardcoded "Skipped" stub.)
+  try {
+    const { readRecentParserProbeEvents } = await import('../core/audit-parser-probe.ts');
+    let parserProbeEnabled = false;
+    try {
+      let dbVal: string | null = null;
+      let dbMode: string | null = null;
+      try {
+        dbVal = engine ? await engine.getConfig('autopilot.conversation_parser_probe.enabled') : null;
+        dbMode = engine ? await engine.getConfig('search.mode') : null;
+      } catch { /* DB unavailable → file plane only */ }
+      const { loadConfig } = await import('../core/config.ts');
+      const fileVal = (loadConfig() as any)?.autopilot?.conversation_parser_probe?.enabled;
+      const flagOn = dbVal != null ? dbVal === 'true' : fileVal === true;
+      parserProbeEnabled = flagOn || dbMode === 'tokenmax';
+    } catch { /* config unavailable → treat as disabled */ }
+    const parserEvents = readRecentParserProbeEvents(7);
+    checks.push(computeConversationParserProbeHealthCheck(parserProbeEnabled, parserEvents));
+  } catch {
+    // Best-effort; audit-log read failure shouldn't stop doctor.
+  }

  // 3e. home_dir_in_worktree (v0.35.8.0). Walks up from `gbrainPath()`
  // looking for a `.git` directory OR file. If found, warns: `~/.gbrain/`
@@ -1,7 +1,6 @@
 import type { BrainEngine } from '../core/engine.ts';
 import { embedBatch, currentEmbeddingSignature } from '../core/embedding.ts';
-import type { ChunkInput, ResolvedColumn } from '../core/types.ts';
-import { resolveWriteColumnForEngine } from '../core/search/embedding-column.ts';
+import type { ChunkInput } from '../core/types.ts';
 import { chunkText } from '../core/chunkers/recursive.ts';
 import { createProgress, type ProgressReporter } from '../core/progress.ts';
 import { getCliOptions, cliOptsToProgressOptions } from '../core/cli-options.ts';
@@ -184,13 +183,8 @@ export class EmbeddingDimMismatchError extends Error {
 * fresh-install bug class at the very first invocation instead of letting
 * the worker pool hammer N pages with raw 22000 errors.
 */
-async function preflightDimMismatch(engine: BrainEngine, dryRun: boolean, embeddingColumn?: ResolvedColumn): Promise<void> {
+async function preflightDimMismatch(engine: BrainEngine, dryRun: boolean): Promise<void> {
  if (dryRun) return; // dry-run never embeds, no risk
-  // #1262: an alt-column brain writes to `embeddingColumn`, not the legacy
-  // `embedding` column — the legacy column's dims are irrelevant, and the
-  // registry entry (validated at resolve time) pins the target's dims. Only
-  // the legacy default path needs the schema-vs-gateway dim comparison.
-  if (embeddingColumn && embeddingColumn.name !== 'embedding') return;
  const { readContentChunksEmbeddingDim, embeddingMismatchMessage } = await import('../core/embedding-dim-check.ts');
  const { getEmbeddingDimensions, getEmbeddingModel } = await import('../core/ai/gateway.ts');
  let existing;
@@ -244,12 +238,7 @@ export async function runEmbedCore(engine: BrainEngine, opts: EmbedOpts): Promis
  // v0.37.11.0 (Lane D.2): pre-flight dim-mismatch check. Catches the headline
  // fresh-install bug class before the worker pool spends 20 parallel calls
  // hitting raw Postgres dimension errors.
-  // #1262: resolve the write-side embedding column ONCE at the boundary
-  // (merged config + gateway model) and thread the descriptor through every
-  // upsertChunks / stale-scan below. undefined => legacy `embedding` column.
-  const embeddingColumn = await resolveWriteColumnForEngine(engine);
-
-  await preflightDimMismatch(engine, !!opts.dryRun, embeddingColumn);
+  await preflightDimMismatch(engine, !!opts.dryRun);

  const result: EmbedResult = {
    embedded: 0,
@@ -264,7 +253,7 @@ export async function runEmbedCore(engine: BrainEngine, opts: EmbedOpts): Promis
    for (const s of opts.slugs) {
      if (isAborted(opts.signal)) break; // #1737: stop the per-slug loop on abort
      try {
-        await embedPage(engine, s, !!opts.dryRun, result, opts.sourceId, opts.signal, embeddingColumn);
+        await embedPage(engine, s, !!opts.dryRun, result, opts.sourceId, opts.signal);
      } catch (e: unknown) {
        serr(`  Error embedding ${s}: ${e instanceof Error ? e.message : e}`);
      }
@@ -358,7 +347,7 @@ export async function runEmbedCore(engine: BrainEngine, opts: EmbedOpts): Promis
        catchUp: opts.catchUp,
        pacer,
        paceMaxConcurrency,
-      }, opts.signal, embeddingColumn);
+      }, opts.signal);
    } finally {
      // E1: surface pacing telemetry (human + structured) when pacing was on.
      const snap = pacer.snapshot();
@@ -387,7 +376,7 @@ export async function runEmbedCore(engine: BrainEngine, opts: EmbedOpts): Promis
    return result;
  }
  if (opts.slug) {
-    await embedPage(engine, opts.slug, !!opts.dryRun, result, opts.sourceId, opts.signal, embeddingColumn);
+    await embedPage(engine, opts.slug, !!opts.dryRun, result, opts.sourceId, opts.signal);
    return result;
  }
  throw new Error('No embed target specified. Pass { slug }, { slugs }, { all }, or { stale }.');
@@ -532,13 +521,8 @@ async function embedPage(
  result: EmbedResult,
  sourceId?: string,
  signal?: AbortSignal,
-  embeddingColumn?: ResolvedColumn,
 ) {
  const opts = sourceId ? { sourceId } : undefined;
-  // #1262: write-side descriptor rides only on WRITE calls (upsertChunks).
-  const chunkOpts = (sourceId || embeddingColumn)
-    ? { ...(sourceId && { sourceId }), ...(embeddingColumn && { embeddingColumn }) }
-    : undefined;
  const page = await engine.getPage(slug, opts);
  if (!page) {
    throw new Error(`Page not found: ${slug}`);
@@ -570,7 +554,7 @@ async function embedPage(
    }

    if (inputs.length > 0) {
-      await engine.upsertChunks(slug, inputs, chunkOpts);
+      await engine.upsertChunks(slug, inputs, opts);
      chunks = await engine.getChunks(slug, opts);
    }
  }
@@ -605,7 +589,7 @@ async function embedPage(
    token_count: c.token_count || Math.ceil(c.chunk_text.length / 4),
  }));

-  await engine.upsertChunks(slug, updated, chunkOpts);
+  await engine.upsertChunks(slug, updated, opts);
  // v0.41.31: stamp provenance so a later model/dims swap is detectable as
  // stale. embedPage is the per-slug path used by `gbrain embed <slug>` AND
  // by `gbrain sync`'s post-import embed step (runEmbedCore({slugs})).
@@ -638,7 +622,6 @@ async function embedAll(
    paceMaxConcurrency?: number;
  },
  signal?: AbortSignal,
-  embeddingColumn?: ResolvedColumn,
 ) {
  // v0.41.31: current embedding provenance signature. Stamped onto pages
  // when their chunks are (re)embedded so a later model/dimension swap is
@@ -661,7 +644,7 @@ async function embedAll(
    // D7: thread sourceId so `gbrain embed --stale --source X` actually scopes.
    // v0.41.18.0 (A13): thread batchSize/priority/catchUp into the stale path.
    // #1737: thread the external abort signal so the cycle embed phase bails.
-    return await embedAllStale(engine, sourceId, dryRun, result, onProgress, staleOpts, signature, signal, embeddingColumn);
+    return await embedAllStale(engine, sourceId, dryRun, result, onProgress, staleOpts, signature, signal);
  }

  // --all path: pacer (no-op when off). E-1: lower the worker count to the
@@ -742,10 +725,7 @@ async function embedAll(
        embedding: embeddingMap.get(c.chunk_index) ?? undefined,
        token_count: c.token_count || Math.ceil(c.chunk_text.length / 4),
      }));
-      await observed(pacer, () => engine.upsertChunks(page.slug, updated, {
-        ...(pageSourceId && { sourceId: pageSourceId }),
-        ...(embeddingColumn && { embeddingColumn }),
-      }));
+      await observed(pacer, () => engine.upsertChunks(page.slug, updated, pageOpts));
      // v0.41.31: stamp embedding provenance so a later model swap is
      // detectable as stale.
      await observed(pacer, () =>
@@ -825,16 +805,10 @@ async function embedAllStale(
  },
  signature?: string,
  externalSignal?: AbortSignal,
-  embeddingColumn?: ResolvedColumn,
 ) {
  // D7: thread sourceId so source-scoped runs only count + visit
  // that source's NULL embeddings.
-  // #1262: the stale predicate follows the write-side column — without it an
-  // alt-column brain would perpetually re-select (and re-pay for) chunks whose
-  // target column is already populated.
-  const sourceOpt = (sourceId || embeddingColumn)
-    ? { ...(sourceId && { sourceId }), ...(embeddingColumn && { embeddingColumn }) }
-    : undefined;
+  const sourceOpt = sourceId ? { sourceId } : undefined;

  // v0.41.31: re-embed pages whose embedding_signature drifted (model/dims
  // swap). dry-run must NOT mutate, so it counts signature-stale via the
@@ -993,7 +967,6 @@ async function embedAllStale(
            afterUpdatedAt,
          }),
          ...(sourceId && { sourceId }),
-          ...(embeddingColumn && { embeddingColumn }),
        }),
      );
      if (batch.length === 0) {
@@ -1046,10 +1019,7 @@ async function embedAllStale(
            embedding: staleIdxToEmbedding.get(c.chunk_index) ?? undefined,
            token_count: c.token_count || Math.ceil(c.chunk_text.length / 4),
          }));
-          await observed(pacer, () => engine.upsertChunks(slug, merged, {
-            sourceId: keySourceId,
-            ...(embeddingColumn && { embeddingColumn }),
-          }));
+          await observed(pacer, () => engine.upsertChunks(slug, merged, { sourceId: keySourceId }));
          // v0.41.31: stamp provenance after the page's chunks are embedded —
          // but only when EVERY chunk was stale (fully re-embedded this pass).
          // A partially-stale page keeps preserved chunks of unknown/old
@@ -1120,7 +1090,7 @@ async function embedAllStale(
  // as a clean run — re-running won't help until the underlying failure is fixed.
  if (staleOpts?.catchUp && !effectiveSignal.aborted && embedFailures > 0) {
    const remaining = await engine.countStaleChunks(
-      signature ? { signature, ...sourceOpt } : sourceOpt,
+      signature ? { signature, ...(sourceId ? { sourceId } : {}) } : (sourceId ? { sourceId } : undefined),
    );
    if (remaining > 0) {
      serr(`\n  [embed] catch-up finished but ${remaining} chunk(s) remain stale after ${embedFailures} embed failure(s). These are not embeddable as-is; re-running won't clear them until the underlying error is resolved.`);
@@ -76,7 +76,7 @@ FLAGS:
                           dimensions (goal, depth, sourcing, specificity, useful).
  --cycles N               1-3. Default: 3 in TTY, 1 in non-TTY (T11). Each
                           cycle is 3 model calls; verdict aggregates over them.
-  --slot-a-model <id>      Override default 'openai:gpt-4o'.
+  --slot-a-model <id>      Override default 'openai:gpt-5.2'.
  --slot-b-model <id>      Override default 'anthropic:claude-opus-4-7'.
  --slot-c-model <id>      Override default 'google:gemini-1.5-pro'.
  --receipt-dir <path>     Default: gbrainPath('eval-receipts').
@@ -468,6 +468,14 @@ interface BatchRow {
  question_id: string;
  question: string;
  hypothesis: string;
+  /**
+   * Gold answer from the benchmark dataset, when the upstream eval emits
+   * it (eval-longmemeval does). Folded into the judge task so CORRECTNESS
+   * is verifiable — without it a judge panel that sees only
+   * {question, hypothesis} cannot validate a terse factual answer against
+   * a haystack it never saw.
+   */
+  answer?: string;
 }

 /**
@@ -581,6 +589,7 @@ function readBatchRows(path: string): BatchReadResult {
      question_id: typeof obj.question_id === 'string' ? obj.question_id : `line-${lineNo}`,
      question: obj.question,
      hypothesis: obj.hypothesis,
+      ...(typeof obj.answer === 'string' && obj.answer.length > 0 ? { answer: obj.answer } : {}),
    });
  }
  if (summarySkipped > 0) {
@@ -697,7 +706,11 @@ async function runBatchMode(parsed: ParsedArgs, opts: RunCrossModalOpts): Promis
      fn: async (row, idx) => {
        process.stderr.write(`[eval cross-modal batch] ${idx + 1}/${rows.length} ${row.question_id} starting...\n`);
        return await runEvalFn({
-          task: row.question,
+          // With a gold answer the judges can actually verify correctness;
+          // without one they see only {question, hypothesis} and cannot.
+          task: row.answer
+            ? `${row.question}\n\nExpected answer (gold label from the benchmark dataset): ${row.answer}`
+            : row.question,
          output: row.hypothesis,
          slug: row.question_id,
          dimensions,
@@ -33,6 +33,7 @@ import {
  type AliasMap,
 } from '../eval/longmemeval/extract.ts';
 import { extractCandidateEntities } from '../core/think/entity-extract.ts';
+import { splitProviderModelId } from '../core/model-id.ts';
 import { resolveEntitySlugWithSource, type ResolutionSource } from '../core/entities/resolve.ts';
 import { formatTrajectoryBlock } from '../core/trajectory-format.ts';

@@ -469,14 +470,22 @@ export async function runEvalLongMemEval(args: string[], runOpts: RunOpts = {}):
  });

  // Wrap Anthropic SDK so its `.messages.create` shape matches ThinkLLMClient.
-  // Same pattern as src/core/think/index.ts:247-249.
+  // Same pattern as src/core/think/index.ts:247-249 — EXCEPT think's default
+  // client routes through the gateway, which parses `provider:model` recipe
+  // ids. This eval's client is a raw SDK by design (hermetic, no gateway
+  // dependency), and resolveModel returns RECIPE ids (`anthropic:claude-…`);
+  // passing one through unstripped 404s every answer/extractor call, which
+  // surfaces downstream as all-upstream_error batches in the nightly probe.
+  const toSdkModel = (m: string): string => splitProviderModelId(m).model || m;
  const realClient = new Anthropic();
  const client: ThinkLLMClient = runOpts.client ?? {
-    create: (params, callOpts) => realClient.messages.create(params, callOpts),
+    create: (params, callOpts) =>
+      realClient.messages.create({ ...params, model: toSdkModel(params.model) }, callOpts),
  };
  // v0.40.2.0 — separate extractor client (defaults to same SDK).
  const extractorClient: ThinkLLMClient = runOpts.extractorClient ?? {
-    create: (params, callOpts) => realClient.messages.create(params, callOpts),
+    create: (params, callOpts) =>
+      realClient.messages.create({ ...params, model: toSdkModel(params.model) }, callOpts),
  };
  const trajectoryEnabled = !opts.noTrajectory;
  const extractorModel = trajectoryEnabled
@@ -751,6 +760,11 @@ async function runOneQuestion(
    // v0.40.1.0 (Track D / T2) — copy question_type into the row so the
    // by_type_summary can be rebuilt from the file on resume runs.
    question_type: q.question_type,
+    // Gold answer for downstream consumers that verify correctness (the
+    // cross-modal --batch judge folds it into the task; evaluate_qa.py
+    // ignores unknown fields). Without it a judge can't validate a terse
+    // factual hypothesis against a haystack it never saw.
+    ...(q.answer !== undefined ? { answer: q.answer } : {}),
    hypothesis,
    retrieved_session_ids: retrievedSessionIds,
    ...(recallHit !== undefined ? { recall_hit: recallHit } : {}),
@@ -576,16 +576,9 @@ async function runInlineCostGate(

  // Stale backlog: cheap single SQL; fail-open to 0 so a transient DB hiccup
  // never blocks the sync. Signature-aware (model/dims swap surfaces here).
-  // #1262: follow the write-side embedding column — otherwise an alt-column
-  // brain's fully-embedded corpus counts as phantom backlog on every gate.
  let staleChars = 0;
  try {
-    const { resolveWriteColumnForEngine } = await import('../core/search/embedding-column.ts');
-    const embeddingColumn = await resolveWriteColumnForEngine(engine);
-    staleChars = await engine.sumStaleChunkChars({
-      signature: currentEmbeddingSignature(),
-      ...(embeddingColumn && { embeddingColumn }),
-    });
+    staleChars = await engine.sumStaleChunkChars({ signature: currentEmbeddingSignature() });
  } catch {
    staleChars = 0;
  }
@@ -0,0 +1,63 @@
+/**
+ * Nightly conversation-parser probe audit trail.
+ *
+ * One event per REAL probe run lands in
+ * `~/.gbrain/audit/parser-probe-YYYY-Www.jsonl` (ISO-week rotation via the
+ * shared audit-writer primitive; honors `GBRAIN_AUDIT_DIR`).
+ * Scheduler-cadence skips (`rate_limited`) are NOT logged — the autopilot
+ * loop ticks every few minutes, so logging every skip would flood the
+ * audit file with rows that carry no signal.
+ *
+ * Read by `gbrain doctor`'s `conversation_parser_probe_health` check and
+ * by the autopilot wiring's 24h rate-limit gate (`parserProbeRanWithin`).
+ */
+
+import { createAuditWriter } from './audit/audit-writer.ts';
+import type { NightlyProbeResult } from './conversation-parser/nightly-probe.ts';
+
+export type ParserProbeAuditEvent = NightlyProbeResult;
+
+const writer = createAuditWriter<ParserProbeAuditEvent>({
+  featureName: 'parser-probe',
+  errorLabel: 'gbrain',
+  errorMessagePrefix: 'parser-probe audit ',
+  errorTrailer: '; probe continues',
+});
+
+/** Append one parser-probe event. Best-effort; never throws. */
+export function logParserProbeEvent(event: ParserProbeAuditEvent): void {
+  writer.log(event);
+}
+
+/**
+ * Read recent parser-probe events (current + previous ISO week, filtered
+ * to the window). Missing files and corrupt rows are skipped silently.
+ */
+export function readRecentParserProbeEvents(
+  days = 7,
+  now: Date = new Date(),
+): ParserProbeAuditEvent[] {
+  return writer.readRecent(days, now);
+}
+
+/** Exposed for tests pinning the rotation edge cases. */
+export function computeParserProbeAuditFilename(now: Date = new Date()): string {
+  return writer.computeFilename(now);
+}
+
+/**
+ * 24h rate-limit gate for the autopilot wiring: true when any audited run
+ * happened within `windowMs` of `now`. Only REAL outcomes are audited (see
+ * module header), so a pass/fail today blocks re-runs until tomorrow while
+ * scheduler-cadence skips never extend the window.
+ */
+export function parserProbeRanWithin(
+  windowMs: number,
+  now: Date = new Date(),
+): boolean {
+  const cutoff = now.getTime() - windowMs;
+  return readRecentParserProbeEvents(2, now).some((ev) => {
+    const ts = Date.parse(ev.ts);
+    return Number.isFinite(ts) && ts >= cutoff;
+  });
+}
@@ -105,6 +105,16 @@ export interface GBrainConfig {
       */
      max_usd?: number;
    };
+    /**
+     * v0.41.16.0 — nightly conversation-parser probe. Per D10: default ON
+     * for `search.mode=tokenmax` brains, opt-in for conservative/balanced.
+     * ~$0.05/night with the committed fixtures × Haiku polish. Gated
+     * INSIDE the autopilot tick body, like nightly_quality_probe.
+     */
+    conversation_parser_probe?: {
+      /** Enable for non-tokenmax modes. Defaults to false. */
+      enabled?: boolean;
+    };
    /**
     * v0.42.x (#1685 GAP D) — extract_atoms backlog auto-drain. Default ON so a
     * pack-gated silent backlog never piles up unseen; daily-spend-capped so the
@@ -61,7 +61,6 @@ import {
  type SynopsisFailureKind,
 } from './audit-synopsis.ts';
 import type { BrainEngine } from './engine.ts';
-import { resolveWriteColumnForEngine } from './search/embedding-column.ts';
 import type { ChunkInput, CRMode, Page } from './types.ts';
 import type { SourceRow } from './sources-ops.ts';

@@ -287,13 +286,9 @@ export async function reembedPageWithContextualRetrieval(

      // ── PHASE 2: single DB transaction ───────────────────────────
      try {
-        // #1262: contextual re-embeds write TEXT embeddings — thread the
-        // caller-resolved write column like every other embed path.
-        const embeddingColumn = await resolveWriteColumnForEngine(args.engine);
        await args.engine.transaction(async (tx) => {
          await tx.upsertChunks(args.pageSlug, phase1.embeddedChunks, {
            sourceId: args.sourceId,
-            ...(embeddingColumn && { embeddingColumn }),
          });
          await tx.updatePageContextualRetrievalState(
            args.pageSlug,
@@ -17,11 +17,11 @@
 * Cost: ~$0.05/night with default fixtures × Haiku polish. Bounded
 * by the active BudgetTracker the autopilot loop creates per-tick.
 *
- * **Wiring into the autopilot loop is deferred to a follow-up**
- * (filed in TODOS.md). v0.41.16.0 ships the phase as a callable
- * module so doctor + future cron drivers can invoke it; the
- * scheduler wire-up follows the same shape as
- * `src/core/cycle/nightly-quality-probe.ts` (v0.40.1.0 Track D / T6).
+ * Wired into the autopilot loop (step 4.6 in autopilot.ts), following
+ * the same shape as `src/core/cycle/nightly-quality-probe.ts`
+ * (v0.40.1.0 Track D / T6): the wiring resolves fixtures from the
+ * gbrain package root, writes real outcomes to the parser-probe audit
+ * trail (`audit-parser-probe.ts`), and never crashes the loop.
 *
 * Test seam: all dependencies are injected via NightlyProbeDeps so
 * unit tests don't touch real LLMs or real fixtures.
@@ -44,7 +44,12 @@ export const DEFAULT_DIMENSIONS: string[] = [
 * `--slot-a-model`, `--slot-b-model`, `--slot-c-model` on the CLI.
 */
 export const DEFAULT_SLOTS: SlotConfig[] = [
-  { id: 'A', model: 'openai:gpt-4o' },
+  // Every default MUST be listed in its recipe's chat touchpoint (pinned by
+  // test/cross-modal-default-slots.test.ts) — `openai:gpt-4o` sat here after
+  // the OpenAI recipe dropped it, so slot A errored "not listed for OpenAI
+  // chat" on every install and the 3-slot panel could never reach its
+  // 2-model quorum without a Google key (verdict: permanently inconclusive).
+  { id: 'A', model: 'openai:gpt-5.2' },
  { id: 'B', model: 'anthropic:claude-opus-4-7' },
  { id: 'C', model: 'google:gemini-1.5-pro' },
 ];
@@ -70,6 +70,28 @@ export async function runLongMemEvalForProbe(args: LongMemEvalProbeArgs): Promis
 * the batch input) or unparseable (cross-modal wrote garbage). Both
 * cases are paste-ready in the error message.
 */
+/**
+ * QA-shaped judge dimensions for the nightly probe. The batch judge's
+ * DEFAULT_DIMENSIONS rubric (DEPTH / SOURCING / SPECIFICITY / …) is built
+ * for rich agent responses; LongMemEval hypotheses are deliberately terse
+ * factual answers ("in widget-co") that can never score ≥7 on DEPTH or
+ * SOURCING — so with the default rubric the probe FAILs every night even
+ * when retrieval + answering are perfectly healthy. The probe owns its
+ * invocation of the eval tool and passes dimensions matching the
+ * fixture's QA shape instead.
+ *
+ * NOTE: the `--dimensions` CLI flag splits on commas, so these dimension
+ * descriptions must stay comma-free.
+ */
+export const PROBE_QA_DIMENSIONS: string[] = [
+  // No faithfulness/grounding dimension on purpose: the judge never sees
+  // the haystack, so any accurate detail beyond the terse gold label reads
+  // as "invented" and correct answers fail (verified empirically — a
+  // correct "before + dates" answer scored 4/10 on such a dimension).
+  'CORRECTNESS — Does the hypothesis state the same fact as the expected answer? A terse direct answer is ideal.',
+  'DIRECTNESS — Does it answer THIS question without hedging or padding or answering something else?',
+];
+
 export async function runCrossModalBatchForProbe(
  args: CrossModalProbeArgs,
 ): Promise<{ exitCode: number; summary: CrossModalBatchSummary }> {
@@ -81,6 +103,8 @@ export async function runCrossModalBatchForProbe(
    args.summaryPath,
    '--max-usd',
    String(args.maxUsd),
+    '--dimensions',
+    PROBE_QA_DIMENSIONS.join(','),
    '--yes',
    '--json',
  ]);
@@ -62,6 +62,42 @@ export interface NightlyProbeDeps {
  now: () => Date;
 }

+/**
+ * Dual-plane flag resolution (same precedent as `mcp.publish_skills` in
+ * serve-http.ts): the DB config row — what `gbrain config set` writes —
+ * wins when present; the file plane (~/.gbrain/config.json) is the
+ * fallback. Doctor's paste-ready enable hint says `gbrain config set
+ * autopilot.nightly_quality_probe.enabled true`, so the gate MUST read
+ * the DB plane — a file-only read turns that hint into a silent no-op.
+ */
+export function resolveProbeEnabled(
+  dbVal: string | null | undefined,
+  fileVal: unknown,
+): boolean {
+  if (dbVal != null) return dbVal === 'true';
+  return fileVal === true;
+}
+
+/**
+ * Same dual-plane rule for the per-run cost cap. Malformed or negative
+ * values on either plane fall through to the next plane / the default.
+ */
+export function resolveProbeMaxUsd(
+  dbVal: string | null | undefined,
+  fileVal: unknown,
+  fallback: number = DEFAULT_MAX_USD,
+): number {
+  if (dbVal != null) {
+    const n = Number(dbVal);
+    if (Number.isFinite(n) && n >= 0) return n;
+  }
+  if (fileVal != null) {
+    const n = Number(fileVal);
+    if (Number.isFinite(n) && n >= 0) return n;
+  }
+  return fallback;
+}
+
 /**
 * Pure function: decide whether the probe should run given the audit
 * history. Returns reason when skipping.
@@ -101,21 +137,17 @@ export async function runNightlyQualityProbe(deps: NightlyProbeDeps): Promise<Ni
    return { outcome: 'disabled', exit_code: 0, detail: 'feature flag off' };
  }

-  // 24h rate limit — skip + audit "rate_limited".
+  // 24h rate limit — skip WITHOUT an audit row. The autopilot loop invokes
+  // the probe every cycle (~5-10 min), so all but one invocation per day
+  // lands here; logging each skip floods the audit file (~hundreds of
+  // rows/day) and — because doctor treats any non-pass outcome as bad
+  // signal — flips nightly_quality_probe_health to a permanent WARN the
+  // moment the probe is enabled. A skip is a non-event: the real runs are
+  // the signal, and their rows are what gates the next 24h window.
  const now = deps.now();
  const recent = readRecentQualityProbeEvents(2, now); // 2-day window is enough for 24h check
  const decision = shouldRunNightly(now, recent);
  if (!decision.run) {
-    logQualityProbeEvent({
-      outcome: 'rate_limited',
-      exit_code: 0,
-      pass_count: 0,
-      fail_count: 0,
-      inconclusive_count: 0,
-      error_count: 0,
-      est_cost_usd: 0,
-      detail: 'already ran within 24h window',
-    });
    return { outcome: 'rate_limited', exit_code: 0, detail: 'already ran within 24h' };
  }

@@ -18,7 +18,7 @@
 */

 import type { BrainEngine } from './engine.ts';
-import type { ChunkInput, ResolvedColumn } from './types.ts';
+import type { ChunkInput } from './types.ts';
 import { embedBatchWithBackoff } from '../commands/embed.ts';
 import { type DbPacer, createNoopPacer, observed } from './db-pacer.ts';
 import { AbortError } from './abort-check.ts';
@@ -61,13 +61,6 @@ export interface EmbedStaleOpts {
   * Omit to keep the legacy `embedding IS NULL`-only behavior.
   */
  embeddingSignature?: string;
-  /**
-   * #1262: caller-resolved write-side embedding column. Threaded into BOTH
-   * listStaleChunks (staleness predicate) and upsertChunks (write target) so
-   * an alt-column brain converges instead of re-selecting embedded rows.
-   * Resolve at the boundary via `resolveWriteColumnForEngine()`.
-   */
-  embeddingColumn?: ResolvedColumn;
  /**
   * DB-contention pacer (paced-backfill). When enabled it (a) supplies the
   * worker count via the caller passing `concurrency = bundle.maxConcurrency`
@@ -163,7 +156,6 @@ export async function embedStaleForSource(
        afterPageId,
        afterChunkIndex,
        sourceId,
-        ...(opts.embeddingColumn && { embeddingColumn: opts.embeddingColumn }),
      }),
    );
    if (batch.length === 0) {
@@ -231,10 +223,7 @@ export async function embedStaleForSource(
          doc_comment: c.doc_comment ?? undefined,
          symbol_name_qualified: c.symbol_name_qualified ?? undefined,
        }));
-        await observed(pacer, () => engine.upsertChunks(slug, merged, {
-          sourceId: keySourceId,
-          ...(opts.embeddingColumn && { embeddingColumn: opts.embeddingColumn }),
-        }));
+        await observed(pacer, () => engine.upsertChunks(slug, merged, { sourceId: keySourceId }));
        // v0.41.31: stamp provenance only when EVERY chunk was stale (fully
        // re-embedded this pass) — a partially-stale page keeps preserved
        // chunks of unknown provenance, so don't claim current. After the
@@ -12,7 +12,6 @@ import type {
  BrainStats, BrainHealth,
  IngestLogEntry, IngestLogInput,
  EngineConfig,
-  ResolvedColumn,
  CodeEdgeInput, CodeEdgeResult,
  EvalCandidate, EvalCandidateInput,
  EvalCaptureFailure, EvalCaptureFailureReason,
@@ -988,13 +987,8 @@ export interface BrainEngine {
   * — Postgres rolls back automatically on conn drop, so commit-ambiguous
   * failure replays to the same end state. Callers MUST NOT wrap externally;
   * see {@link BatchOpts} retry-contract block.
-   *
-   * `opts.embeddingColumn` (optional) selects the content_chunks column that
-   * receives TEXT embeddings (#1262). The caller resolves the descriptor at
-   * the import/embed boundary via `resolveWriteColumn()`; engines never read
-   * config or choose columns themselves. Omitted => legacy `embedding`.
   */
-  upsertChunks(slug: string, chunks: ChunkInput[], opts?: { sourceId?: string; embeddingColumn?: ResolvedColumn } & BatchOpts): Promise<void>;
+  upsertChunks(slug: string, chunks: ChunkInput[], opts?: { sourceId?: string } & BatchOpts): Promise<void>;
  /**
   * Read every chunk for a page. `opts.sourceId` source-scopes the page
   * lookup; without it, multi-source brains return chunks from every
@@ -1011,13 +1005,8 @@ export interface BrainEngine {
   * counts across every source in the brain. Operators running
   * `gbrain embed --stale --source media-corpus` expect only that
   * source's NULLs touched; the caller threads `sourceId` here.
-   *
-   * `opts.embeddingColumn` switches the staleness predicate from the legacy
-   * `embedding` column to the resolved write-side column, so alt-column
-   * brains do not perpetually re-select rows whose target column is already
-   * populated (#1262). Must match the eventual upsertChunks target.
   */
-  countStaleChunks(opts?: { sourceId?: string; signature?: string; embeddingColumn?: ResolvedColumn }): Promise<number>;
+  countStaleChunks(opts?: { sourceId?: string; signature?: string }): Promise<number>;
  /**
   * Sum of LENGTH(chunk_text) over stale chunks — the character-count
   * backlog the embed phase / embed-backfill will process. Sibling of
@@ -1031,13 +1020,8 @@ export interface BrainEngine {
   * model signature (a model/dims swap). NULL signature is GRANDFATHERED
   * (never counted) so the post-migration corpus isn't flagged en masse.
   * Omit `signature` for the legacy `embedding IS NULL`-only count.
-   *
-   * `opts.embeddingColumn` switches the staleness predicate to the resolved
-   * write-side column (#1262) — same contract as countStaleChunks — so the
-   * sync cost gate doesn't count an alt-column brain's fully-embedded corpus
-   * as phantom backlog.
   */
-  sumStaleChunkChars(opts?: { sourceId?: string; signature?: string; embeddingColumn?: ResolvedColumn }): Promise<number>;
+  sumStaleChunkChars(opts?: { sourceId?: string; signature?: string }): Promise<number>;
  /**
   * Stamp `pages.embedding_signature = signature` for one page. Called after
   * a page's chunks are (re)embedded so a later model swap can detect it as
@@ -1085,9 +1069,6 @@ export interface BrainEngine {
    // both round-trip TIMESTAMPTZ as Date | string; ISO string is the
    // common denominator on the wire).
    afterUpdatedAt?: string | null;
-    // #1262: staleness predicate targets this column when set (must match
-    // countStaleChunks and the eventual upsertChunks write target).
-    embeddingColumn?: ResolvedColumn;
  }): Promise<StaleChunkRow[]>;
  /**
   * Delete every chunk for a page. Internal page-id lookup is sourceId-scoped
@@ -10,8 +10,7 @@ import { findChunkForOffset } from './chunkers/edge-extractor.ts';
 import { extractCodeRefs, imageOfCandidates } from './link-extraction.ts';
 import { embedBatch, embedMultimodal, currentEmbeddingSignature } from './embedding.ts';
 import { slugifyPath, slugifyCodePath, isCodeFilePath } from './sync.ts';
-import type { ChunkInput, PageInput, PageType, ResolvedColumn } from './types.ts';
-import { resolveWriteColumnForEngine } from './search/embedding-column.ts';
+import type { ChunkInput, PageInput, PageType } from './types.ts';
 import { computeEffectiveDate } from './effective-date.ts';
 import { MARKDOWN_CHUNKER_VERSION } from './chunkers/recursive.ts';
 import { logSlugFallback } from './audit-slug-fallback.ts';
@@ -741,14 +740,6 @@ export async function importFromContent(
  // schema DEFAULT — required for multi-source brains; harmless ('default')
  // for single-source callers.
  const txOpts = sourceId ? { sourceId } : undefined;
-  // #1262: resolve the write-side embedding column once (merged config +
-  // gateway model) BEFORE the transaction; the descriptor rides only on
-  // upsertChunks so text embeddings land in the registered column.
-  const chunkWriteColumn = await resolveWriteColumnForEngine(engine);
-  const chunkOpts: { sourceId?: string; embeddingColumn?: ResolvedColumn } | undefined =
-    (sourceId || chunkWriteColumn)
-      ? { ...(sourceId && { sourceId }), ...(chunkWriteColumn && { embeddingColumn: chunkWriteColumn }) }
-      : undefined;
  await engine.transaction(async (tx) => {
    if (existing) await tx.createVersion(slug, txOpts);

@@ -833,7 +824,7 @@ export async function importFromContent(
    }

    if (chunks.length > 0) {
-      await tx.upsertChunks(slug, chunks, chunkOpts);
+      await tx.upsertChunks(slug, chunks, txOpts);
      // v0.41.31: stamp embedding provenance when this import actually
      // embedded (not --no-embed), so a later model/dims swap is detectable
      // as stale via embed --stale. The deferred/backfill + per-slug embed
@@ -1073,12 +1064,6 @@ export async function importCodeFile(
  const title = `${relativePath} (${lang})`;
  const sourceId = opts.sourceId;
  const txOpts = sourceId ? { sourceId } : undefined;
-  // #1262: write-side embedding column descriptor (rides only on upsertChunks).
-  const chunkWriteColumn = await resolveWriteColumnForEngine(engine);
-  const chunkOpts: { sourceId?: string; embeddingColumn?: ResolvedColumn } | undefined =
-    (sourceId || chunkWriteColumn)
-      ? { ...(sourceId && { sourceId }), ...(chunkWriteColumn && { embeddingColumn: chunkWriteColumn }) }
-      : undefined;

  const byteLength = Buffer.byteLength(content, 'utf-8');
  if (byteLength > MAX_FILE_SIZE) {
@@ -1198,7 +1183,7 @@ export async function importCodeFile(
    await tx.addTag(slug, lang, txOpts);

    if (chunks.length > 0) {
-      await tx.upsertChunks(slug, chunks, chunkOpts);
+      await tx.upsertChunks(slug, chunks, txOpts);
      // v0.41.31: stamp embedding provenance ONLY when every chunk was
      // freshly embedded with the current model this call (no reuse-by-hash
      // carrying old-model vectors). Mixed pages stay unstamped rather than
@@ -1347,12 +1332,6 @@ export async function withImportTransaction(
 ): Promise<void> {
  const sourceId = spec.sourceId ?? 'default';
  const txOpts = spec.sourceId ? { sourceId: spec.sourceId } : undefined;
-  // #1262: write-side embedding column descriptor (rides only on upsertChunks).
-  const chunkWriteColumn = await resolveWriteColumnForEngine(engine);
-  const chunkOpts: { sourceId?: string; embeddingColumn?: ResolvedColumn } | undefined =
-    (spec.sourceId || chunkWriteColumn)
-      ? { ...(spec.sourceId && { sourceId: spec.sourceId }), ...(chunkWriteColumn && { embeddingColumn: chunkWriteColumn }) }
-      : undefined;
  await engine.transaction(async (tx) => {
    if (spec.hadExisting) await tx.createVersion(spec.slug, txOpts);
    await tx.putPage(spec.slug, spec.page, txOpts);
@@ -1368,7 +1347,7 @@ export async function withImportTransaction(
    }
    if (spec.chunks !== undefined) {
      if (spec.chunks.length > 0) {
-        await tx.upsertChunks(spec.slug, spec.chunks, chunkOpts);
+        await tx.upsertChunks(spec.slug, spec.chunks, txOpts);
      } else {
        await tx.deleteChunks(spec.slug, txOpts);
      }
@@ -35,7 +35,6 @@ import { tryAcquireDbLock } from '../../db-lock.ts';
 import { BudgetTracker, BudgetExhausted } from '../../budget/budget-tracker.ts';
 import { withBudgetTracker } from '../../ai/gateway.ts';
 import { embedStaleForSource } from '../../embed-stale.ts';
-import { resolveWriteColumnForEngine } from '../../search/embedding-column.ts';
 import { currentEmbeddingSignature } from '../../embedding.ts';
 import { type DbPacer, createDbPacer, createNoopPacer } from '../../db-pacer.ts';
 import { resolvePaceMode, loadPaceModeConfig, readPaceEnv } from '../../pace-mode.ts';
@@ -165,16 +164,12 @@ export function makeEmbedBackfillHandler(engine: BrainEngine) {
    // the supervisor, so pacing it is the headline win.
    const { pacer, concurrency } = await resolveBackfillPacer(engine, job.data);

-    // #1262: resolve the write-side embedding column once at the job boundary.
-    const embeddingColumn = await resolveWriteColumnForEngine(engine);
-
    try {
      const result = await withBudgetTracker(tracker, async () =>
        embedStaleForSource(engine, sourceId, {
          batchSize,
          signal: job.signal,
          pacer,
-          ...(embeddingColumn && { embeddingColumn }),
          ...(concurrency !== undefined && { concurrency }),
          // v0.41.31: re-embed pages whose model signature drifted + stamp
          // provenance as chunks land.
@@ -75,6 +75,11 @@ export const CANONICAL_PRICING: Record<string, ModelPricing> = {
  'openai:gpt-4o':                        { input:  2.50, output: 10.00 },
  'openai:gpt-4o-mini':                   { input:  0.15, output:  0.60 },
  'openai:gpt-5':                         { input:  5.00, output: 20.00 },
+  // gpt-5.2: rates from the OpenAI recipe chat touchpoint (verified
+  // 2026-04-20). Needed here because it's the cross-modal DEFAULT_SLOTS
+  // slot-A model — without a canonical entry estimateCost silently drops
+  // slot A from the --max-usd pre-flight and est_cost_usd audit rows.
+  'openai:gpt-5.2':                       { input:  1.25, output: 10.00 },
  'openai:gpt-5.5':                       { input:  4.00, output: 16.00 },

  // ── Google ─────────────────────────────────────────────────────────────
@@ -40,7 +40,6 @@ import type {
  BrainStats, BrainHealth,
  IngestLogEntry, IngestLogInput,
  EngineConfig,
-  ResolvedColumn,
  EvalCandidate, EvalCandidateInput,
  EvalCaptureFailure, EvalCaptureFailureReason,
  SalienceOpts, SalienceResult, AnomaliesOpts, AnomalyResult,
@@ -2231,20 +2230,12 @@ export class PGLiteEngine implements BrainEngine {
  }

  // Chunks
-  async upsertChunks(slug: string, chunks: ChunkInput[], opts?: { sourceId?: string; embeddingColumn?: ResolvedColumn } & BatchOpts): Promise<void> {
+  async upsertChunks(slug: string, chunks: ChunkInput[], opts?: { sourceId?: string } & BatchOpts): Promise<void> {
    return this.batchRetry(opts?.auditSite ?? 'upsertChunks', opts?.signal, () => this._upsertChunksOnce(slug, chunks, opts), chunks.length);
  }

-  private async _upsertChunksOnce(slug: string, chunks: ChunkInput[], opts?: { sourceId?: string; embeddingColumn?: ResolvedColumn }): Promise<void> {
+  private async _upsertChunksOnce(slug: string, chunks: ChunkInput[], opts?: { sourceId?: string }): Promise<void> {
    const sourceId = opts?.sourceId ?? 'default';
-    // #1262: caller-resolved write target for TEXT embeddings. Descriptor
-    // names are identifier-validated + quoted by buildVectorCastFragment;
-    // omitted => legacy `embedding vector`. Mirrors postgres-engine.ts.
-    const targetFragment = opts?.embeddingColumn
-      ? buildVectorCastFragment(opts.embeddingColumn)
-      : undefined;
-    const targetCol = targetFragment?.col ?? 'embedding';
-    const embeddingCast = targetFragment?.castSql.replace('$1::', '') ?? 'vector';

    // Source-scope the page-id lookup so duplicate slugs in different sources
    // do not return multiple rows or target the wrong page.
@@ -2279,7 +2270,7 @@ export class PGLiteEngine implements BrainEngine {
    // list. Image chunks pass embedding=null + embedding_image=Float32Array
    // (1024-dim Voyage). Text/code chunks pass embedding=Float32Array +
    // embedding_image=null. Default modality='text' when omitted.
-    const cols = `(page_id, chunk_index, chunk_text, chunk_source, ${targetCol}, model, token_count, embedded_at, language, symbol_name, symbol_type, start_line, end_line, parent_symbol_path, doc_comment, symbol_name_qualified, modality, embedding_image)`;
+    const cols = '(page_id, chunk_index, chunk_text, chunk_source, embedding, model, token_count, embedded_at, language, symbol_name, symbol_type, start_line, end_line, parent_symbol_path, doc_comment, symbol_name_qualified, modality, embedding_image)';
    const rowParts: string[] = [];
    const params: unknown[] = [];
    let paramIdx = 1;
@@ -2297,7 +2288,7 @@ export class PGLiteEngine implements BrainEngine {
      const modality = chunk.modality ?? 'text';

      // Inline ::vector NULL literals to avoid a per-branch placeholder.
-      const embeddingPh = embeddingStr ? `$${paramIdx++}::${embeddingCast}` : 'NULL';
+      const embeddingPh = embeddingStr ? `$${paramIdx++}::vector` : 'NULL';
      const embeddedAtPh = embeddingStr ? 'now()' : 'NULL';
      const embeddingImagePh = embeddingImageStr ? `$${paramIdx++}::vector` : 'NULL';

@@ -2336,19 +2327,19 @@ export class PGLiteEngine implements BrainEngine {
       ON CONFLICT (page_id, chunk_index) DO UPDATE SET
         chunk_text = EXCLUDED.chunk_text,
         chunk_source = EXCLUDED.chunk_source,
-         ${targetCol} = CASE
-           WHEN EXCLUDED.chunk_text != content_chunks.chunk_text THEN EXCLUDED.${targetCol}
-           WHEN content_chunks.${targetCol} IS NULL THEN EXCLUDED.${targetCol}
+         embedding = CASE
+           WHEN EXCLUDED.chunk_text != content_chunks.chunk_text THEN EXCLUDED.embedding
+           WHEN content_chunks.embedding IS NULL THEN EXCLUDED.embedding
           WHEN EXCLUDED.embedded_at IS NOT NULL
                AND (content_chunks.embedded_at IS NULL OR EXCLUDED.embedded_at > content_chunks.embedded_at)
-                THEN EXCLUDED.${targetCol}
-           ELSE content_chunks.${targetCol}
+                THEN EXCLUDED.embedding
+           ELSE content_chunks.embedding
         END,
         model = COALESCE(EXCLUDED.model, content_chunks.model),
         token_count = EXCLUDED.token_count,
         embedded_at = CASE
-           WHEN EXCLUDED.chunk_text != content_chunks.chunk_text AND EXCLUDED.${targetCol} IS NULL THEN NULL
-           WHEN content_chunks.${targetCol} IS NULL AND EXCLUDED.${targetCol} IS NOT NULL THEN EXCLUDED.embedded_at
+           WHEN EXCLUDED.chunk_text != content_chunks.chunk_text AND EXCLUDED.embedding IS NULL THEN NULL
+           WHEN content_chunks.embedding IS NULL AND EXCLUDED.embedding IS NOT NULL THEN EXCLUDED.embedded_at
           WHEN EXCLUDED.embedded_at IS NOT NULL
                AND (content_chunks.embedded_at IS NULL OR EXCLUDED.embedded_at > content_chunks.embedded_at)
                THEN EXCLUDED.embedded_at
@@ -2386,19 +2377,14 @@ export class PGLiteEngine implements BrainEngine {
   * drift (NULL grandfathered → never stale). Shared by countStaleChunks +
   * sumStaleChunkChars so they can't drift.
   */
-  private buildStaleChunkWhere(opts?: { sourceId?: string; signature?: string; embeddingColumn?: ResolvedColumn }): { where: string; params: unknown[] } {
-    // #1262: staleness targets the caller-resolved write column when set
-    // (identifier-validated + quoted); legacy `embedding` otherwise.
-    const staleCol = opts?.embeddingColumn
-      ? buildVectorCastFragment(opts.embeddingColumn).col
-      : 'embedding';
+  private buildStaleChunkWhere(opts?: { sourceId?: string; signature?: string }): { where: string; params: unknown[] } {
    const params: unknown[] = [];
    const conds: string[] = [];
    if (opts?.signature !== undefined) {
      params.push(opts.signature);
-      conds.push(`(cc.${staleCol} IS NULL OR (p.embedding_signature IS NOT NULL AND p.embedding_signature <> $${params.length}))`);
+      conds.push(`(cc.embedding IS NULL OR (p.embedding_signature IS NOT NULL AND p.embedding_signature <> $${params.length}))`);
    } else {
-      conds.push(`cc.${staleCol} IS NULL`);
+      conds.push(`cc.embedding IS NULL`);
    }
    conds.push(`NOT (COALESCE(p.frontmatter, '{}'::jsonb) ? 'embed_skip')`);
    if (opts?.sourceId !== undefined) {
@@ -2408,7 +2394,7 @@ export class PGLiteEngine implements BrainEngine {
    return { where: conds.join(' AND '), params };
  }

-  async countStaleChunks(opts?: { sourceId?: string; signature?: string; embeddingColumn?: ResolvedColumn }): Promise<number> {
+  async countStaleChunks(opts?: { sourceId?: string; signature?: string }): Promise<number> {
    // D7: source-scoped count for `gbrain embed --stale --source X`. Always
    // JOIN pages so embed-skip + signature predicates apply. PGLite is
    // PostgreSQL 17.5 in WASM and supports the full JSONB operator set.
@@ -2424,7 +2410,7 @@ export class PGLiteEngine implements BrainEngine {
    return Number(count);
  }

-  async sumStaleChunkChars(opts?: { sourceId?: string; signature?: string; embeddingColumn?: ResolvedColumn }): Promise<number> {
+  async sumStaleChunkChars(opts?: { sourceId?: string; signature?: string }): Promise<number> {
    // Sibling of countStaleChunks: same stale predicate, summing chunk_text
    // length for the sync cost preview. ::bigint guards int4 overflow.
    const { where, params } = this.buildStaleChunkWhere(opts);
@@ -2477,17 +2463,11 @@ export class PGLiteEngine implements BrainEngine {
    sourceId?: string;
    orderBy?: 'page_id' | 'updated_desc';
    afterUpdatedAt?: string | null;
-    embeddingColumn?: ResolvedColumn;
  }): Promise<StaleChunkRow[]> {
    const limit = opts?.batchSize ?? 2000;
    const afterPid = opts?.afterPageId ?? 0;
    const afterIdx = opts?.afterChunkIndex ?? -1;
    const orderBy = opts?.orderBy ?? 'page_id';
-    // #1262: staleness follows the caller-resolved write column (validated +
-    // quoted identifier); legacy `embedding` otherwise.
-    const staleCol = opts?.embeddingColumn
-      ? buildVectorCastFragment(opts.embeddingColumn).col
-      : 'embedding';

    // v0.41.18.0 (A13, codex #9): --priority recent path. See postgres-engine
    // sibling for full rationale. Same composite cursor + ORDER BY.
@@ -2501,7 +2481,7 @@ export class PGLiteEngine implements BrainEngine {
                  p.updated_at
             FROM content_chunks cc
             JOIN pages p ON p.id = cc.page_id
-            WHERE cc.${staleCol} IS NULL
+            WHERE cc.embedding IS NULL
              AND NOT (COALESCE(p.frontmatter, '{}'::jsonb) ? 'embed_skip')
            ORDER BY p.updated_at DESC NULLS LAST, p.id ASC, cc.chunk_index ASC
            LIMIT $1`,
@@ -2512,7 +2492,7 @@ export class PGLiteEngine implements BrainEngine {
                  p.updated_at
             FROM content_chunks cc
             JOIN pages p ON p.id = cc.page_id
-            WHERE cc.${staleCol} IS NULL
+            WHERE cc.embedding IS NULL
              AND NOT (COALESCE(p.frontmatter, '{}'::jsonb) ? 'embed_skip')
              AND (
                p.updated_at < $1::timestamptz
@@ -2531,7 +2511,7 @@ export class PGLiteEngine implements BrainEngine {
                p.updated_at
           FROM content_chunks cc
           JOIN pages p ON p.id = cc.page_id
-          WHERE cc.${staleCol} IS NULL
+          WHERE cc.embedding IS NULL
            AND p.source_id = $1
            AND NOT (COALESCE(p.frontmatter, '{}'::jsonb) ? 'embed_skip')
          ORDER BY p.updated_at DESC NULLS LAST, p.id ASC, cc.chunk_index ASC
@@ -2543,7 +2523,7 @@ export class PGLiteEngine implements BrainEngine {
                p.updated_at
           FROM content_chunks cc
           JOIN pages p ON p.id = cc.page_id
-          WHERE cc.${staleCol} IS NULL
+          WHERE cc.embedding IS NULL
            AND p.source_id = $1
            AND NOT (COALESCE(p.frontmatter, '{}'::jsonb) ? 'embed_skip')
            AND (
@@ -2568,7 +2548,7 @@ export class PGLiteEngine implements BrainEngine {
                cc.model, cc.token_count, p.source_id, cc.page_id
           FROM content_chunks cc
           JOIN pages p ON p.id = cc.page_id
-          WHERE cc.${staleCol} IS NULL
+          WHERE cc.embedding IS NULL
            AND NOT (COALESCE(p.frontmatter, '{}'::jsonb) ? 'embed_skip')
            AND (cc.page_id, cc.chunk_index) > ($1, $2)
          ORDER BY cc.page_id, cc.chunk_index
@@ -2582,7 +2562,7 @@ export class PGLiteEngine implements BrainEngine {
              cc.model, cc.token_count, p.source_id, cc.page_id
         FROM content_chunks cc
         JOIN pages p ON p.id = cc.page_id
-        WHERE cc.${staleCol} IS NULL
+        WHERE cc.embedding IS NULL
          AND p.source_id = $1
          AND NOT (COALESCE(p.frontmatter, '{}'::jsonb) ? 'embed_skip')
          AND (cc.page_id, cc.chunk_index) > ($2, $3)
@@ -50,7 +50,6 @@ import type {
  BrainStats, BrainHealth,
  IngestLogEntry, IngestLogInput,
  EngineConfig,
-  ResolvedColumn,
  EvalCandidate, EvalCandidateInput,
  EvalCaptureFailure, EvalCaptureFailureReason,
  SalienceOpts, SalienceResult, AnomaliesOpts, AnomalyResult,
@@ -2381,21 +2380,13 @@ export class PostgresEngine implements BrainEngine {
  }

  // Chunks
-  async upsertChunks(slug: string, chunks: ChunkInput[], opts?: { sourceId?: string; embeddingColumn?: ResolvedColumn } & BatchOpts): Promise<void> {
+  async upsertChunks(slug: string, chunks: ChunkInput[], opts?: { sourceId?: string } & BatchOpts): Promise<void> {
    return this.batchRetry(opts?.auditSite ?? 'upsertChunks', opts?.signal, () => this._upsertChunksOnce(slug, chunks, opts), chunks.length);
  }

-  private async _upsertChunksOnce(slug: string, chunks: ChunkInput[], opts?: { sourceId?: string; embeddingColumn?: ResolvedColumn }): Promise<void> {
+  private async _upsertChunksOnce(slug: string, chunks: ChunkInput[], opts?: { sourceId?: string }): Promise<void> {
    const sql = this.sql;
    const sourceId = opts?.sourceId ?? 'default';
-    // #1262: caller-resolved write target for TEXT embeddings. Descriptor
-    // names are identifier-validated + quoted by buildVectorCastFragment;
-    // omitted => legacy `embedding vector`.
-    const targetFragment = opts?.embeddingColumn
-      ? buildVectorCastFragment(opts.embeddingColumn)
-      : undefined;
-    const targetCol = targetFragment?.col ?? 'embedding';
-    const embeddingCast = targetFragment?.castSql.replace('$1::', '') ?? 'vector';

    // Source-scope the page-id lookup. Without this filter, multi-source
    // brains where the slug exists in 2+ sources return >1 row and the
@@ -2422,7 +2413,7 @@ export class PostgresEngine implements BrainEngine {
    // scope metadata through upserts.
    // v0.27.1 (Phase 8): added `modality` + `embedding_image` to the column
    // list. Image chunks pass embedding=null + embedding_image=Float32Array.
-    const cols = `(page_id, chunk_index, chunk_text, chunk_source, ${targetCol}, model, token_count, embedded_at, language, symbol_name, symbol_type, start_line, end_line, parent_symbol_path, doc_comment, symbol_name_qualified, modality, embedding_image)`;
+    const cols = '(page_id, chunk_index, chunk_text, chunk_source, embedding, model, token_count, embedded_at, language, symbol_name, symbol_type, start_line, end_line, parent_symbol_path, doc_comment, symbol_name_qualified, modality, embedding_image)';
    const rows: string[] = [];
    const params: unknown[] = [];
    let paramIdx = 1;
@@ -2439,7 +2430,7 @@ export class PostgresEngine implements BrainEngine {
        : null;
      const modality = chunk.modality ?? 'text';

-      const embeddingPh = embeddingStr ? `$${paramIdx++}::${embeddingCast}` : 'NULL';
+      const embeddingPh = embeddingStr ? `$${paramIdx++}::vector` : 'NULL';
      const embeddedAtPh = embeddingStr ? 'now()' : 'NULL';
      const embeddingImagePh = embeddingImageStr ? `$${paramIdx++}::vector` : 'NULL';

@@ -2487,19 +2478,19 @@ export class PostgresEngine implements BrainEngine {
       ON CONFLICT (page_id, chunk_index) DO UPDATE SET
         chunk_text = EXCLUDED.chunk_text,
         chunk_source = EXCLUDED.chunk_source,
-         ${targetCol} = CASE
-           WHEN EXCLUDED.chunk_text != content_chunks.chunk_text THEN EXCLUDED.${targetCol}
-           WHEN content_chunks.${targetCol} IS NULL THEN EXCLUDED.${targetCol}
+         embedding = CASE
+           WHEN EXCLUDED.chunk_text != content_chunks.chunk_text THEN EXCLUDED.embedding
+           WHEN content_chunks.embedding IS NULL THEN EXCLUDED.embedding
           WHEN EXCLUDED.embedded_at IS NOT NULL
                AND (content_chunks.embedded_at IS NULL OR EXCLUDED.embedded_at > content_chunks.embedded_at)
-                THEN EXCLUDED.${targetCol}
-           ELSE content_chunks.${targetCol}
+                THEN EXCLUDED.embedding
+           ELSE content_chunks.embedding
         END,
         model = COALESCE(EXCLUDED.model, content_chunks.model),
         token_count = EXCLUDED.token_count,
         embedded_at = CASE
-           WHEN EXCLUDED.chunk_text != content_chunks.chunk_text AND EXCLUDED.${targetCol} IS NULL THEN NULL
-           WHEN content_chunks.${targetCol} IS NULL AND EXCLUDED.${targetCol} IS NOT NULL THEN EXCLUDED.embedded_at
+           WHEN EXCLUDED.chunk_text != content_chunks.chunk_text AND EXCLUDED.embedding IS NULL THEN NULL
+           WHEN content_chunks.embedding IS NULL AND EXCLUDED.embedding IS NOT NULL THEN EXCLUDED.embedded_at
           WHEN EXCLUDED.embedded_at IS NOT NULL
                AND (content_chunks.embedded_at IS NULL OR EXCLUDED.embedded_at > content_chunks.embedded_at)
                THEN EXCLUDED.embedded_at
@@ -2539,19 +2530,14 @@ export class PostgresEngine implements BrainEngine {
   * embedding_signature drift (NULL grandfathered). Shared by
   * countStaleChunks + sumStaleChunkChars (parity with the PGLite sibling).
   */
-  private buildStaleChunkWhere(opts?: { sourceId?: string; signature?: string; embeddingColumn?: ResolvedColumn }): { where: string; params: unknown[] } {
-    // #1262: staleness targets the caller-resolved write column when set
-    // (identifier-validated + quoted); legacy `embedding` otherwise.
-    const staleCol = opts?.embeddingColumn
-      ? buildVectorCastFragment(opts.embeddingColumn).col
-      : 'embedding';
+  private buildStaleChunkWhere(opts?: { sourceId?: string; signature?: string }): { where: string; params: unknown[] } {
    const params: unknown[] = [];
    const conds: string[] = [];
    if (opts?.signature !== undefined) {
      params.push(opts.signature);
-      conds.push(`(cc.${staleCol} IS NULL OR (p.embedding_signature IS NOT NULL AND p.embedding_signature <> $${params.length}))`);
+      conds.push(`(cc.embedding IS NULL OR (p.embedding_signature IS NOT NULL AND p.embedding_signature <> $${params.length}))`);
    } else {
-      conds.push(`cc.${staleCol} IS NULL`);
+      conds.push(`cc.embedding IS NULL`);
    }
    conds.push(`NOT (COALESCE(p.frontmatter, '{}'::jsonb) ? 'embed_skip')`);
    if (opts?.sourceId !== undefined) {
@@ -2561,7 +2547,7 @@ export class PostgresEngine implements BrainEngine {
    return { where: conds.join(' AND '), params };
  }

-  async countStaleChunks(opts?: { sourceId?: string; signature?: string; embeddingColumn?: ResolvedColumn }): Promise<number> {
+  async countStaleChunks(opts?: { sourceId?: string; signature?: string }): Promise<number> {
    // Always JOIN pages so the embed_skip + signature predicates apply.
    // D7: source_id scoping. v0.41.31: optional signature widens staleness
    // to embedding_signature drift (NULL grandfathered).
@@ -2579,7 +2565,7 @@ export class PostgresEngine implements BrainEngine {
    });
  }

-  async sumStaleChunkChars(opts?: { sourceId?: string; signature?: string; embeddingColumn?: ResolvedColumn }): Promise<number> {
+  async sumStaleChunkChars(opts?: { sourceId?: string; signature?: string }): Promise<number> {
    // Sibling of countStaleChunks: same stale predicate, summing chunk_text
    // length for the sync cost preview. ::bigint guards int4 overflow.
    const { where, params } = this.buildStaleChunkWhere(opts);
@@ -2632,18 +2618,11 @@ export class PostgresEngine implements BrainEngine {
    sourceId?: string;
    orderBy?: 'page_id' | 'updated_desc';
    afterUpdatedAt?: string | null;
-    embeddingColumn?: ResolvedColumn;
  }): Promise<StaleChunkRow[]> {
    const limit = opts?.batchSize ?? 2000;
    const afterPid = opts?.afterPageId ?? 0;
    const afterIdx = opts?.afterChunkIndex ?? -1;
    const orderBy = opts?.orderBy ?? 'page_id';
-    // #1262: staleness follows the caller-resolved write column (validated +
-    // quoted identifier); legacy `embedding` otherwise. Interpolated below as
-    // an unsafe FRAGMENT (identifiers can't be bound parameters).
-    const staleCol = opts?.embeddingColumn
-      ? buildVectorCastFragment(opts.embeddingColumn).col
-      : 'embedding';

    // RLS scope binding (opt-in via GBRAIN_RLS_SCOPE_BINDING).
    return await this.withScopedReadTransaction(undefined, opts?.sourceId, async (tx) => {
@@ -2660,7 +2639,7 @@ export class PostgresEngine implements BrainEngine {
                   p.updated_at
            FROM content_chunks cc
            JOIN pages p ON p.id = cc.page_id
-            WHERE ${tx.unsafe(`cc.${staleCol} IS NULL`)}
+            WHERE cc.embedding IS NULL
              AND NOT (COALESCE(p.frontmatter, '{}'::jsonb) ? 'embed_skip')
            ORDER BY p.updated_at DESC NULLS LAST, p.id ASC, cc.chunk_index ASC
            LIMIT ${limit}
@@ -2670,7 +2649,7 @@ export class PostgresEngine implements BrainEngine {
                   p.updated_at
            FROM content_chunks cc
            JOIN pages p ON p.id = cc.page_id
-            WHERE ${tx.unsafe(`cc.${staleCol} IS NULL`)}
+            WHERE cc.embedding IS NULL
              AND NOT (COALESCE(p.frontmatter, '{}'::jsonb) ? 'embed_skip')
              AND (
                p.updated_at < ${afterUpdated}::timestamptz
@@ -2688,7 +2667,7 @@ export class PostgresEngine implements BrainEngine {
                 p.updated_at
          FROM content_chunks cc
          JOIN pages p ON p.id = cc.page_id
-          WHERE ${tx.unsafe(`cc.${staleCol} IS NULL`)}
+          WHERE cc.embedding IS NULL
            AND p.source_id = ${opts.sourceId}
            AND NOT (COALESCE(p.frontmatter, '{}'::jsonb) ? 'embed_skip')
          ORDER BY p.updated_at DESC NULLS LAST, p.id ASC, cc.chunk_index ASC
@@ -2699,7 +2678,7 @@ export class PostgresEngine implements BrainEngine {
                 p.updated_at
          FROM content_chunks cc
          JOIN pages p ON p.id = cc.page_id
-          WHERE ${tx.unsafe(`cc.${staleCol} IS NULL`)}
+          WHERE cc.embedding IS NULL
            AND p.source_id = ${opts.sourceId}
            AND NOT (COALESCE(p.frontmatter, '{}'::jsonb) ? 'embed_skip')
            AND (
@@ -2719,7 +2698,7 @@ export class PostgresEngine implements BrainEngine {
                 cc.model, cc.token_count, p.source_id, cc.page_id
          FROM content_chunks cc
          JOIN pages p ON p.id = cc.page_id
-          WHERE ${tx.unsafe(`cc.${staleCol} IS NULL`)}
+          WHERE cc.embedding IS NULL
            AND NOT (COALESCE(p.frontmatter, '{}'::jsonb) ? 'embed_skip')
            AND (cc.page_id, cc.chunk_index) > (${afterPid}, ${afterIdx})
          ORDER BY cc.page_id, cc.chunk_index
@@ -2732,7 +2711,7 @@ export class PostgresEngine implements BrainEngine {
               cc.model, cc.token_count, p.source_id, cc.page_id
        FROM content_chunks cc
        JOIN pages p ON p.id = cc.page_id
-        WHERE ${tx.unsafe(`cc.${staleCol} IS NULL`)}
+        WHERE cc.embedding IS NULL
          AND p.source_id = ${opts.sourceId}
          AND NOT (COALESCE(p.frontmatter, '{}'::jsonb) ? 'embed_skip')
          AND (cc.page_id, cc.chunk_index) > (${afterPid}, ${afterIdx})
@@ -443,80 +443,6 @@ export function resolveEmbeddingColumn(
  };
 }

-/**
- * Resolves the WRITE-side embedding column for the currently configured
- * embedding model (#1262). The read-side resolver above answers "which
- * column does this query search?"; this one answers "which column should
- * newly produced text embeddings land in?".
- *
- * Unlike read-side search, writes take no per-call column override. The
- * import/embed boundary resolves once from merged config + gateway state
- * and passes the descriptor into `engine.upsertChunks`; engines stay
- * config-free (same contract as the read-side descriptor).
- *
- * Behavior:
- *   - no user-declared `embedding_columns` => undefined (legacy brain,
- *     writes keep targeting the default `embedding` column)
- *   - a user-declared entry whose `provider` matches the current
- *     embedding model => that entry's descriptor
- *   - no provider match => undefined (fall back to legacy `embedding`)
- *
- * Only USER-declared entries are consulted — never the cfg-derived
- * builtins. The `embedding_image` builtin's provider is the multimodal
- * model; matching it here would misroute text embeddings into the image
- * column. The no-match fallback is intentional: switching models before
- * registering a matching column must not silently write vectors into an
- * arbitrary column.
- */
-export function resolveWriteColumn(cfg: GBrainConfig): ResolvedColumn | undefined {
-  const userColumns = cfg.embedding_columns;
-  if (
-    !userColumns ||
-    typeof userColumns !== 'object' ||
-    Array.isArray(userColumns) ||
-    Object.keys(userColumns).length === 0
-  ) {
-    return undefined;
-  }
-
-  // Same model-resolution chain as the registry builtin: cfg > gateway > default.
-  let gwModel: string | undefined;
-  try {
-    const gw = require('../ai/gateway.ts') as typeof import('../ai/gateway.ts');
-    gwModel = gw.getEmbeddingModel();
-  } catch {
-    // Gateway unconfigured — fall through to the canonical default.
-  }
-  const currentModel = cfg.embedding_model ?? gwModel ?? DEFAULT_EMBEDDING_MODEL;
-
-  for (const [name, entry] of Object.entries(userColumns)) {
-    if (!entry) continue;
-    validateColumnKey(name);
-    validateColumnConfig(name, entry);
-    if (entry.provider !== currentModel) continue;
-    return {
-      name,
-      type: entry.type,
-      dimensions: entry.dimensions,
-      embeddingModel: entry.provider,
-    };
-  }
-  return undefined;
-}
-
-/**
- * Engine-boundary convenience: merged config (file/env + DB plane) →
- * resolveWriteColumn. Dynamic import keeps config.ts out of this module's
- * static graph (mirrors the gateway require above).
- */
-export async function resolveWriteColumnForEngine(
-  engine: { getConfig(key: string): Promise<string | null | undefined> },
-): Promise<ResolvedColumn | undefined> {
-  const { loadConfigWithEngine } = await import('../config.ts');
-  const cfg = await loadConfigWithEngine(engine);
-  return cfg ? resolveWriteColumn(cfg) : undefined;
-}
-
 /**
 * True when the resolved column is the default `embedding` name.
 * Name-based check; does not compare embedding space.
@@ -0,0 +1,102 @@
+/**
+ * Tests for the parser-probe audit trail + the 24h rate-limit gate.
+ *
+ * Uses GBRAIN_AUDIT_DIR override pointed at a tmpdir for hermeticity
+ * (same pattern as audit-slug-fallback.serial.test.ts). Serial because
+ * the env override is process-global.
+ */
+import { afterEach, beforeEach, describe, expect, test } from 'bun:test';
+import { mkdtempSync, rmSync, readdirSync } from 'node:fs';
+import { tmpdir } from 'node:os';
+import { join } from 'node:path';
+
+import {
+  computeParserProbeAuditFilename,
+  logParserProbeEvent,
+  parserProbeRanWithin,
+  readRecentParserProbeEvents,
+  type ParserProbeAuditEvent,
+} from '../src/core/audit-parser-probe.ts';
+
+let auditDir: string;
+let savedEnv: string | undefined;
+
+beforeEach(() => {
+  auditDir = mkdtempSync(join(tmpdir(), 'parser-probe-audit-'));
+  savedEnv = process.env.GBRAIN_AUDIT_DIR;
+  process.env.GBRAIN_AUDIT_DIR = auditDir;
+});
+
+afterEach(() => {
+  if (savedEnv === undefined) delete process.env.GBRAIN_AUDIT_DIR;
+  else process.env.GBRAIN_AUDIT_DIR = savedEnv;
+  rmSync(auditDir, { recursive: true, force: true });
+});
+
+function makeEvent(overrides: Partial<ParserProbeAuditEvent> = {}): ParserProbeAuditEvent {
+  return {
+    schema_version: 1,
+    ts: new Date().toISOString(),
+    outcome: 'pass',
+    fixtures_total: 12,
+    fixtures_passed: 12,
+    recall_mean: 0.98,
+    participants_recall_mean: 0.97,
+    adversarial_false_positives: 0,
+    failed_fixture_ids: [],
+    ...overrides,
+  };
+}
+
+describe('parser-probe audit trail', () => {
+  test('log + readRecent round-trip', () => {
+    logParserProbeEvent(makeEvent({ outcome: 'fail', reason: '2 fixture(s) failed' }));
+    const events = readRecentParserProbeEvents(7);
+    expect(events.length).toBe(1);
+    expect(events[0]!.outcome).toBe('fail');
+    expect(events[0]!.reason).toBe('2 fixture(s) failed');
+    const files = readdirSync(auditDir);
+    expect(files.length).toBe(1);
+    expect(files[0]).toMatch(/^parser-probe-\d{4}-W\d{2}\.jsonl$/);
+  });
+
+  test('filename uses ISO-week rotation with the parser-probe prefix', () => {
+    // Year-boundary edge pinned by the shared writer's own tests; here we
+    // pin the prefix wiring.
+    expect(computeParserProbeAuditFilename(new Date('2026-07-06T12:00:00Z'))).toBe(
+      'parser-probe-2026-W28.jsonl',
+    );
+  });
+
+  test('readRecent filters by window', () => {
+    const old = new Date(Date.now() - 10 * 86400000).toISOString();
+    logParserProbeEvent(makeEvent({ ts: old }));
+    expect(readRecentParserProbeEvents(7).length).toBe(0);
+  });
+});
+
+describe('parserProbeRanWithin — 24h rate-limit gate', () => {
+  const DAY_MS = 24 * 60 * 60 * 1000;
+
+  test('false when no runs are audited', () => {
+    expect(parserProbeRanWithin(DAY_MS)).toBe(false);
+  });
+
+  test('true when a run landed within the window', () => {
+    logParserProbeEvent(makeEvent({ ts: new Date(Date.now() - 60_000).toISOString() }));
+    expect(parserProbeRanWithin(DAY_MS)).toBe(true);
+  });
+
+  test('false when the last run is older than the window', () => {
+    logParserProbeEvent(makeEvent({ ts: new Date(Date.now() - 25 * 3600_000).toISOString() }));
+    expect(parserProbeRanWithin(DAY_MS)).toBe(false);
+  });
+
+  test('non-pass outcomes also hold the window (mirrors quality-probe semantics)', () => {
+    logParserProbeEvent(makeEvent({
+      outcome: 'no_embedding_key',
+      ts: new Date(Date.now() - 3600_000).toISOString(),
+    }));
+    expect(parserProbeRanWithin(DAY_MS)).toBe(true);
+  });
+});
@@ -31,10 +31,15 @@ describe('autopilot wiring: nightly quality probe', () => {
    expect(SOURCE).toContain(`runCrossModalBatchForProbe`);
  });

-  test('feature flag gate present: cfg.autopilot.nightly_quality_probe.enabled', () => {
+  test('feature flag gate present: dual-plane read (DB row wins, file plane fallback)', () => {
    // Per D10: the scheduler ONLY checks the feature flag. The 24h rate-limit
    // lives inside runNightlyQualityProbe itself (no scheduler-side precheck).
-    expect(SOURCE).toContain(`nightly_quality_probe?.enabled === true`);
+    // The flag resolves through resolveProbeEnabled so `gbrain config set
+    // autopilot.nightly_quality_probe.enabled true` (the doctor hint, DB
+    // plane) and ~/.gbrain/config.json (file plane) BOTH work — a file-only
+    // read made the printed hint a silent no-op.
+    expect(SOURCE).toContain(`getConfig('autopilot.nightly_quality_probe.enabled')`);
+    expect(SOURCE).toMatch(/resolveProbeEnabled\(dbEnabled,\s*cfg\?\.autopilot\?\.nightly_quality_probe\?\.enabled\)/);
  });

  test('NO scheduler-side rate-limit check (D10 simplification)', () => {
@@ -64,12 +69,23 @@ describe('autopilot wiring: nightly quality probe', () => {
    expect(SOURCE).toContain(`now:`);
  });

+  test('resolveRepoRoot prefers the gbrain package root (committed fixture home), not the brain repoPath', () => {
+    // The DI harness in nightly-quality-probe.test.ts passes process.cwd()
+    // (= the gbrain repo in CI), which papered over the wiring passing
+    // repoPath (= sync.repo_path, the user's BRAIN repo, where the fixture
+    // never exists). Pin the package-root resolution + existence check.
+    expect(SOURCE).toMatch(/fileURLToPath\(new URL\('\.\.\/\.\.', import\.meta\.url\)\)/);
+    expect(SOURCE).toContain(`'longmemeval-nightly.jsonl'`);
+    expect(SOURCE).toMatch(/fixtureAtPkgRoot \? pkgRoot : repoPath/);
+  });
+
  test('hasEmbeddingProvider reads from gateway.isAvailable("embedding") (codex round-2 #12 — in-process, not subprocess)', () => {
    expect(SOURCE).toContain(`isAvailable('embedding')`);
    expect(SOURCE).toContain(`gateway`);
  });

-  test('max_usd default = 5 when config unset (matches plan default per D10)', () => {
-    expect(SOURCE).toMatch(/max_usd\s*\?\?\s*5/);
+  test('max_usd resolves dual-plane (default = 5 pinned by resolveProbeMaxUsd unit tests)', () => {
+    expect(SOURCE).toContain(`getConfig('autopilot.nightly_quality_probe.max_usd')`);
+    expect(SOURCE).toMatch(/resolveProbeMaxUsd\(dbMaxUsd,\s*cfg\?\.autopilot\?\.nightly_quality_probe\?\.max_usd\)/);
  });
 });
@@ -0,0 +1,77 @@
+/**
+ * Source-shape regression tests for the autopilot wiring of
+ * `runConversationParserNightlyProbe` (step 4.6).
+ *
+ * Same rationale as autopilot-nightly-probe-wiring.test.ts: the loop is
+ * hard to drive end-to-end, so these pin the structural protections —
+ * the dual-plane flag read, the D10 tokenmax mode-gate, the package-root
+ * fixture resolution, the audit-flood guard, and the try/catch posture.
+ *
+ * The probe's own gate/scoring logic is pinned by the module's unit
+ * tests; the audit trail by audit-parser-probe.serial.test.ts.
+ */
+
+import { describe, test, expect } from 'bun:test';
+import { readFileSync } from 'node:fs';
+import { resolve } from 'node:path';
+
+const AUTOPILOT_SRC = resolve('src/commands/autopilot.ts');
+const SOURCE = readFileSync(AUTOPILOT_SRC, 'utf-8');
+
+describe('autopilot wiring: conversation-parser probe', () => {
+  test('invokes the phase module and the audit trail', () => {
+    expect(SOURCE).toContain(`runConversationParserNightlyProbe`);
+    expect(SOURCE).toContain(`conversation-parser/nightly-probe`);
+    expect(SOURCE).toContain(`logParserProbeEvent`);
+    expect(SOURCE).toContain(`audit-parser-probe`);
+  });
+
+  test('flag reads dual-plane: DB row (gbrain config set) wins, file plane fallback', () => {
+    expect(SOURCE).toContain(`getConfig('autopilot.conversation_parser_probe.enabled')`);
+    expect(SOURCE).toContain(`cfg?.autopilot?.conversation_parser_probe?.enabled === true`);
+  });
+
+  test('D10 mode-gate present: tokenmax brains run the probe by default', () => {
+    expect(SOURCE).toMatch(/parserEnabled \|\| searchMode === 'tokenmax'/);
+  });
+
+  test('fixtures resolve from the gbrain package root, NOT the brain repoPath', () => {
+    // The committed fixtures live in the gbrain source tree; resolving
+    // them against sync.repo_path would point into the user's brain repo.
+    expect(SOURCE).toMatch(/fileURLToPath\(new URL\('\.\.\/\.\.', import\.meta\.url\)\)/);
+    expect(SOURCE).toContain(`'conversation-formats', 'all.jsonl'`);
+    expect(SOURCE).toContain(`'conversation-formats', 'adversarial.jsonl'`);
+  });
+
+  test('missing fixtures skip quietly (no audit row, once-per-process stderr note)', () => {
+    // Compiled-binary installs carry no source tree; writing failure rows
+    // would flip doctor to WARN on every binary install.
+    expect(SOURCE).toContain(`parserProbeFixtureWarned`);
+  });
+
+  test('rate_limited outcomes are NOT audit-logged (flood guard)', () => {
+    expect(SOURCE).toMatch(/outcome !== 'rate_limited'\) logParserProbeEvent\(result\)/);
+  });
+
+  test('rate-limit gate delegates to the audit module, not inline event reads', () => {
+    expect(SOURCE).toContain(`parserProbeRanWithin(24 * 60 * 60 * 1000)`);
+  });
+
+  test('LLM-key gate reads gateway.isAvailable("chat") in-process', () => {
+    expect(SOURCE).toContain(`isAvailable('chat')`);
+  });
+
+  test('probe call wrapped in try/catch that does NOT bump consecutiveErrors', () => {
+    expect(SOURCE).toMatch(/catch[\s\S]*?autopilot\.parser_probe[\s\S]*?do NOT bump consecutiveErrors/);
+  });
+
+  test('DI shape: the exact 7 fields of the parser probe NightlyProbeDeps', () => {
+    expect(SOURCE).toContain(`isEnabled:`);
+    expect(SOURCE).toContain(`searchMode:`);
+    expect(SOURCE).toContain(`hasLlmKey:`);
+    expect(SOURCE).toContain(`resolveFixturePath:`);
+    expect(SOURCE).toContain(`resolveAdversarialPath:`);
+    expect(SOURCE).toContain(`shouldSkipForRateLimit:`);
+    expect(SOURCE).toContain(`now:`);
+  });
+});
@@ -0,0 +1,47 @@
+/**
+ * Consistency guard: every cross-modal DEFAULT_SLOTS model must be listed
+ * in its recipe's chat touchpoint. `openai:gpt-4o` drifted out of the
+ * OpenAI recipe while remaining the slot-A default — the gateway then
+ * rejected slot A ("not listed for OpenAI chat") on every install, and the
+ * 3-slot judge panel could never reach its 2-model quorum without a Google
+ * key, pinning every batch verdict at inconclusive (which the nightly
+ * quality probe surfaces as a doctor WARN).
+ */
+import { describe, expect, test } from 'bun:test';
+
+import { DEFAULT_SLOTS } from '../src/core/cross-modal-eval/runner.ts';
+import { getRecipe } from '../src/core/ai/recipes/index.ts';
+import { splitProviderModelId } from '../src/core/model-id.ts';
+import { canonicalLookup } from '../src/core/model-pricing.ts';
+
+describe('cross-modal DEFAULT_SLOTS ↔ recipe consistency', () => {
+  test('every default slot model is listed in its recipe chat touchpoint', () => {
+    for (const slot of DEFAULT_SLOTS) {
+      const { provider, model } = splitProviderModelId(slot.model);
+      expect(provider).not.toBeNull();
+      const recipe = getRecipe(provider!);
+      expect(recipe, `slot ${slot.id}: unknown recipe "${provider}"`).toBeDefined();
+      const chatModels = recipe!.touchpoints.chat?.models ?? [];
+      expect(
+        chatModels,
+        `slot ${slot.id}: "${model}" not listed for ${provider} chat — the judge slot can never run`,
+      ).toContain(model);
+    }
+  });
+
+  test('every default slot model has a canonical pricing entry', () => {
+    // Without one, estimateCost silently drops the slot from the
+    // --max-usd pre-flight and est_cost_usd audit rows (~1/3 under-count).
+    for (const slot of DEFAULT_SLOTS) {
+      expect(
+        canonicalLookup(slot.model),
+        `slot ${slot.id}: "${slot.model}" missing from CANONICAL_PRICING`,
+      ).toBeDefined();
+    }
+  });
+
+  test('slots span three distinct providers (uncorrelated blind spots)', () => {
+    const providers = new Set(DEFAULT_SLOTS.map(s => splitProviderModelId(s.model).provider));
+    expect(providers.size).toBe(3);
+  });
+});
@@ -0,0 +1,51 @@
+/**
+ * Tests for computeConversationParserProbeHealthCheck — the pure function
+ * behind doctor's conversation_parser_probe_health check, which replaced
+ * the v0.41.13.0 hardcoded "Skipped" stub when the autopilot wiring
+ * landed. Mirrors the branch coverage style of the quality-probe check.
+ */
+import { describe, expect, test } from 'bun:test';
+
+import { computeConversationParserProbeHealthCheck } from '../src/commands/doctor.ts';
+
+const ev = (outcome: string, reason?: string, ts = new Date().toISOString()) => ({
+  outcome,
+  ts,
+  ...(reason !== undefined ? { reason } : {}),
+});
+
+describe('computeConversationParserProbeHealthCheck', () => {
+  test('disabled + no events → ok with paste-ready enable hint', () => {
+    const check = computeConversationParserProbeHealthCheck(false, []);
+    expect(check.status).toBe('ok');
+    expect(check.message).toContain('gbrain config set autopilot.conversation_parser_probe.enabled true');
+  });
+
+  test('enabled + no events yet → ok, next run by autopilot', () => {
+    const check = computeConversationParserProbeHealthCheck(true, []);
+    expect(check.status).toBe('ok');
+    expect(check.message).toContain('no probe events');
+  });
+
+  test('disabled flag but events exist (tokenmax mode-gate ran it) → events win over the hint', () => {
+    const check = computeConversationParserProbeHealthCheck(false, [ev('pass')]);
+    expect(check.status).toBe('ok');
+    expect(check.message).toContain('all pass');
+  });
+
+  test('any non-pass outcome in the window → warn, latest surfaced with reason', () => {
+    const check = computeConversationParserProbeHealthCheck(true, [
+      ev('pass'),
+      ev('adversarial_false_positive', '1 adversarial fixture(s) parsed to non-empty'),
+    ]);
+    expect(check.status).toBe('warn');
+    expect(check.message).toContain('adversarial_false_positive');
+    expect(check.message).toContain('parsed to non-empty');
+  });
+
+  test('all pass → ok with run count', () => {
+    const check = computeConversationParserProbeHealthCheck(true, [ev('pass'), ev('pass')]);
+    expect(check.status).toBe('ok');
+    expect(check.message).toContain('2 probe run(s)');
+  });
+});
@@ -241,136 +241,3 @@ describe('buildVectorCastFragment — engine SQL composer (D3)', () => {
    expect(castSql).toBe('$1::halfvec(2560)');
  });
 });
-
-describe('PGLite engine: upsertChunks write-side ResolvedColumn descriptor (#1262)', () => {
-  test('halfvec descriptor writes the text embedding to the alternate column, not legacy embedding', async () => {
-    await engine.putPage('docs/write-alt-pglite', {
-      type: 'concept',
-      title: 'Write alt column PGLite',
-      compiled_truth: 'PGLite write-side alternate embedding column test.',
-    });
-
-    const descriptor: ResolvedColumn = {
-      name: 'embedding_ze',
-      type: 'halfvec',
-      dimensions: 2560,
-      embeddingModel: 'zeroentropyai:zembed-1',
-    };
-    await engine.upsertChunks('docs/write-alt-pglite', [
-      {
-        chunk_index: 0,
-        chunk_text: 'PGLite write-side alternate embedding column test.',
-        chunk_source: 'compiled_truth',
-        embedding: new Float32Array(2560).fill(0.25),
-      },
-    ], { embeddingColumn: descriptor });
-
-    const rows = await engine.executeRaw<{
-      has_default: boolean;
-      has_ze: boolean;
-      has_embedded_at: boolean;
-    }>(
-      `SELECT embedding IS NOT NULL AS has_default,
-              embedding_ze IS NOT NULL AS has_ze,
-              embedded_at IS NOT NULL AS has_embedded_at
-         FROM content_chunks cc
-         JOIN pages p ON p.id = cc.page_id
-        WHERE p.slug = 'docs/write-alt-pglite'`,
-    );
-    expect(rows.length).toBe(1);
-    expect(rows[0].has_default).toBe(false);
-    expect(rows[0].has_ze).toBe(true);
-    expect(rows[0].has_embedded_at).toBe(true);
-  });
-
-  test('text-unchanged re-upsert without a vector preserves the alternate-column embedding', async () => {
-    const descriptor: ResolvedColumn = {
-      name: 'embedding_ze',
-      type: 'halfvec',
-      dimensions: 2560,
-      embeddingModel: 'zeroentropyai:zembed-1',
-    };
-    // Same chunk_text, no embedding: the ON CONFLICT CASE must keep the
-    // existing alternate-column vector (D24 semantics follow the column).
-    await engine.upsertChunks('docs/write-alt-pglite', [
-      {
-        chunk_index: 0,
-        chunk_text: 'PGLite write-side alternate embedding column test.',
-        chunk_source: 'compiled_truth',
-      },
-    ], { embeddingColumn: descriptor });
-    const rows = await engine.executeRaw<{ has_ze: boolean }>(
-      `SELECT embedding_ze IS NOT NULL AS has_ze
-         FROM content_chunks cc
-         JOIN pages p ON p.id = cc.page_id
-        WHERE p.slug = 'docs/write-alt-pglite'`,
-    );
-    expect(rows).toEqual([{ has_ze: true }]);
-  });
-});
-
-describe('PGLite: embed --stale converges on an alt-column brain (#1262)', () => {
-  test('boundary resolves the write column; stale scan does not re-select embedded rows', async () => {
-    const { runEmbedCore } = await import('../../src/commands/embed.ts');
-    const local = new PGLiteEngine();
-    const previousHome = process.env.GBRAIN_HOME;
-    process.env.GBRAIN_HOME = `/tmp/gbrain-write-col-stale-${Date.now()}`;
-    try {
-      await local.connect({});
-      await local.initSchema();
-      await (local as any).db.exec(
-        `ALTER TABLE content_chunks ADD COLUMN IF NOT EXISTS embedding_ze halfvec(2560)`,
-      );
-
-      const descriptor: ResolvedColumn = {
-        name: 'embedding_ze',
-        type: 'halfvec',
-        dimensions: 2560,
-        embeddingModel: 'zeroentropyai:zembed-1',
-      };
-      await local.setConfig('embedding_columns', JSON.stringify({
-        embedding_ze: { provider: 'zeroentropyai:zembed-1', dimensions: 2560, type: 'halfvec' },
-      }));
-      configureGateway({
-        embedding_model: 'zeroentropyai:zembed-1',
-        embedding_dimensions: 2560,
-        env: {},
-      });
-
-      await local.putPage('docs/stale-alt-pglite', {
-        type: 'concept',
-        title: 'Dynamic stale column',
-        compiled_truth: 'A chunk that is embedded only in the dynamic column.',
-      });
-      await local.upsertChunks('docs/stale-alt-pglite', [
-        {
-          chunk_index: 0,
-          chunk_text: 'A chunk that is embedded only in the dynamic column.',
-          chunk_source: 'compiled_truth',
-          embedding: new Float32Array(2560).fill(0.25),
-        },
-      ], { embeddingColumn: descriptor });
-
-      // Engine-level contrast: legacy predicate still sees the row as stale;
-      // the alt-column predicate does not.
-      expect(await local.countStaleChunks()).toBe(1);
-      expect(await local.countStaleChunks({ embeddingColumn: descriptor })).toBe(0);
-      // sumStaleChunkChars feeds the sync cost gate — same predicate contract.
-      expect(await local.sumStaleChunkChars()).toBeGreaterThan(0);
-      expect(await local.sumStaleChunkChars({ embeddingColumn: descriptor })).toBe(0);
-      expect(await local.listStaleChunks({ embeddingColumn: descriptor, batchSize: 100 })).toHaveLength(0);
-      expect(await local.listStaleChunks({ batchSize: 100 })).toHaveLength(1);
-
-      // Boundary-level: `embed --stale --dry-run` resolves the write column
-      // from merged config + gateway and reports NOTHING to embed. Without
-      // the fix this reports 1 (perpetual re-embed loop).
-      const result = await runEmbedCore(local, { stale: true, dryRun: true });
-      expect(result.would_embed).toBe(0);
-    } finally {
-      await local.disconnect();
-      if (previousHome === undefined) delete process.env.GBRAIN_HOME;
-      else process.env.GBRAIN_HOME = previousHome;
-      resetGateway();
-    }
-  });
-});
@@ -224,54 +224,4 @@ if (!dbUrl) {
      await engine.executeRaw(`UPDATE content_chunks SET embedding_voyage = '${v}'::vector WHERE id = ${dogId}`);
    });
  });
-
-  describe('Postgres: upsertChunks write-side ResolvedColumn descriptor (#1262)', () => {
-    const descriptor: ResolvedColumn = {
-      name: 'embedding_ze',
-      type: 'halfvec',
-      dimensions: 2560,
-      embeddingModel: 'zeroentropyai:zembed-1',
-    };
-
-    test('halfvec descriptor writes the text embedding to the alternate column, not legacy embedding', async () => {
-      await engine.putPage('docs/write-alt-postgres', {
-        type: 'concept',
-        title: 'Write alt column Postgres',
-        compiled_truth: 'Postgres write-side alternate embedding column test.',
-      });
-      await engine.upsertChunks('docs/write-alt-postgres', [
-        {
-          chunk_index: 0,
-          chunk_text: 'Postgres write-side alternate embedding column test.',
-          chunk_source: 'compiled_truth',
-          embedding: new Float32Array(2560).fill(0.25),
-        },
-      ], { embeddingColumn: descriptor });
-
-      const rows = await engine.executeRaw<{
-        has_default: boolean;
-        has_ze: boolean;
-      }>(
-        `SELECT embedding IS NOT NULL AS has_default,
-                embedding_ze IS NOT NULL AS has_ze
-           FROM content_chunks cc
-           JOIN pages p ON p.id = cc.page_id
-          WHERE p.slug = 'docs/write-alt-postgres'`,
-      );
-      expect(rows.length).toBe(1);
-      expect(rows[0].has_default).toBe(false);
-      expect(rows[0].has_ze).toBe(true);
-    }, 30_000);
-
-    test('stale scan follows the write-side column (count + list parity with the write target)', async () => {
-      // Legacy predicate: cat/dog/write-alt rows all have embedding NULL.
-      expect(await engine.countStaleChunks()).toBeGreaterThan(0);
-      // Alt-column predicate: every chunk has embedding_ze populated.
-      expect(await engine.countStaleChunks({ embeddingColumn: descriptor })).toBe(0);
-      expect(await engine.listStaleChunks({ embeddingColumn: descriptor, batchSize: 100 })).toHaveLength(0);
-      expect((await engine.listStaleChunks({ batchSize: 100 })).length).toBeGreaterThan(0);
-      // updated_desc arm uses the same predicate.
-      expect(await engine.listStaleChunks({ embeddingColumn: descriptor, orderBy: 'updated_desc', batchSize: 100 })).toHaveLength(0);
-    }, 30_000);
-  });
 }
@@ -0,0 +1,74 @@
+// Regression test for the nightly-quality-probe config-plane split-brain.
+//
+// The doctor check prints a paste-ready enable hint — `gbrain config set
+// autopilot.nightly_quality_probe.enabled true` — which writes the DB config
+// plane. But both the autopilot gate and the doctor check used to read ONLY
+// the file plane (~/.gbrain/config.json via loadConfig), so following the
+// hint was a silent no-op: the probe never ran and doctor kept reporting
+// "disabled (opt-in)".
+//
+// resolveProbeEnabled / resolveProbeMaxUsd pin the dual-plane rule (same
+// precedent as `mcp.publish_skills` in serve-http.ts): DB row wins when
+// present, file plane is the fallback.
+import { describe, expect, test } from 'bun:test';
+
+import {
+  resolveProbeEnabled,
+  resolveProbeMaxUsd,
+} from '../src/core/cycle/nightly-quality-probe.ts';
+
+describe('resolveProbeEnabled — dual-plane flag resolution', () => {
+  test('DB plane "true" enables regardless of file plane (the doctor hint path)', () => {
+    expect(resolveProbeEnabled('true', undefined)).toBe(true);
+    expect(resolveProbeEnabled('true', false)).toBe(true);
+  });
+
+  test('explicit DB "false" wins over file-plane true (config set off sticks)', () => {
+    expect(resolveProbeEnabled('false', true)).toBe(false);
+  });
+
+  test('file plane is the fallback when no DB row exists', () => {
+    expect(resolveProbeEnabled(null, true)).toBe(true);
+    expect(resolveProbeEnabled(undefined, true)).toBe(true);
+    expect(resolveProbeEnabled(null, undefined)).toBe(false);
+    expect(resolveProbeEnabled(null, false)).toBe(false);
+  });
+
+  test('file plane stays strict boolean — string "true" in config.json does not enable', () => {
+    // Matches the pre-fix autopilot gate (`=== true`); the doctor check used
+    // Boolean(...) and could disagree with autopilot on a string value.
+    // Both call sites now share this helper, so they can no longer diverge.
+    expect(resolveProbeEnabled(null, 'true')).toBe(false);
+    expect(resolveProbeEnabled(null, 1)).toBe(false);
+  });
+
+  test('non-"true" DB strings are off (mcp.publish_skills semantics)', () => {
+    expect(resolveProbeEnabled('1', true)).toBe(false);
+    expect(resolveProbeEnabled('yes', true)).toBe(false);
+    expect(resolveProbeEnabled('', true)).toBe(false);
+  });
+});
+
+describe('resolveProbeMaxUsd — dual-plane cost cap resolution', () => {
+  test('DB plane wins when parseable', () => {
+    expect(resolveProbeMaxUsd('2.5', 10)).toBe(2.5);
+    expect(resolveProbeMaxUsd('0', 10)).toBe(0);
+  });
+
+  test('malformed or negative DB value falls through to file plane', () => {
+    expect(resolveProbeMaxUsd('banana', 3)).toBe(3);
+    expect(resolveProbeMaxUsd('-1', 3)).toBe(3);
+  });
+
+  test('file plane used when no DB row; default when both absent/invalid', () => {
+    expect(resolveProbeMaxUsd(null, 7)).toBe(7);
+    expect(resolveProbeMaxUsd(null, '4')).toBe(4);
+    expect(resolveProbeMaxUsd(null, undefined)).toBe(5);
+    expect(resolveProbeMaxUsd(null, 'banana')).toBe(5);
+    expect(resolveProbeMaxUsd(undefined, -2)).toBe(5);
+  });
+
+  test('explicit fallback override is honored', () => {
+    expect(resolveProbeMaxUsd(null, undefined, 12)).toBe(12);
+  });
+});
@@ -132,17 +132,20 @@ describe('runNightlyQualityProbe (DI stub harness)', () => {
    });
  });

-  test('enabled + recent run within 24h → outcome: rate_limited', async () => {
+  test('enabled + recent run within 24h → outcome: rate_limited, NO audit row', async () => {
    // Pre-seed a recent audit event by running the probe once first.
    await withEnv({ GBRAIN_AUDIT_DIR: auditTmp }, async () => {
      // First run succeeds.
      await runNightlyQualityProbe(makeDeps());
-      // Second run, same hour → rate_limited.
+      // Second run, same hour → rate_limited. A skip is a non-event: the
+      // autopilot loop invokes the probe every cycle (~5-10 min), so
+      // logging each skip would flood the audit file and flip doctor's
+      // any-non-pass-is-bad filter to a permanent WARN.
      const r2 = await runNightlyQualityProbe(makeDeps());
      expect(r2.outcome).toBe('rate_limited');
      const events = await readEvents();
-      expect(events.length).toBe(2);
-      expect(events[1].outcome).toBe('rate_limited');
+      expect(events.length).toBe(1);
+      expect(events[0].outcome).toBe('pass');
    });
  });

@@ -13,10 +13,9 @@
 *     throw on unknown string.
 */

-import { describe, test, expect, afterAll, afterEach } from 'bun:test';
+import { describe, test, expect } from 'bun:test';
 import {
  resolveEmbeddingColumn,
-  resolveWriteColumn,
  getEmbeddingColumnRegistry,
  buildVectorCastFragment,
  quoteIdentifier,
@@ -35,28 +34,6 @@ import {
 } from '../../src/core/search/embedding-column.ts';
 import type { GBrainConfig } from '../../src/core/config.ts';
 import type { ResolvedColumn } from '../../src/core/types.ts';
-import { configureGateway, resetGateway } from '../../src/core/ai/gateway.ts';
-
-/**
- * Teardown: reset AND re-apply the legacy preload config
- * (test/helpers/legacy-embedding-preload.ts). A bare resetGateway() would
- * leave the slot empty for the NEXT file's beforeAll (the preload's
- * per-test beforeEach only fires before tests, not before beforeAll), which
- * would make sibling PGLite fixtures initSchema at the 1280 default instead
- * of the legacy 1536 their seed vectors assume.
- */
-function restorePreloadGateway() {
-  resetGateway();
-  configureGateway({
-    embedding_model: 'openai:text-embedding-3-large',
-    embedding_dimensions: 1536,
-    env: { ...process.env },
-  });
-}
-
-afterAll(() => {
-  restorePreloadGateway();
-});

 function cfg(overrides: Partial<GBrainConfig> = {}): GBrainConfig {
  return { engine: 'pglite', ...overrides };
@@ -545,89 +522,3 @@ describe('codex /ship #4 — isCacheSafe (embedding-space-based skip)', () => {
    expect(isCacheSafe(r, cfg())).toBe(true);
  });
 });
-
-describe('resolveWriteColumn — write-side boundary resolution (#1262)', () => {
-  afterEach(() => {
-    restorePreloadGateway();
-  });
-
-  test('no registry / empty registry returns undefined (legacy single-column brain)', () => {
-    expect(resolveWriteColumn(cfg())).toBeUndefined();
-    expect(resolveWriteColumn(cfg({ embedding_columns: {} }))).toBeUndefined();
-  });
-
-  test('provider match via cfg.embedding_model returns the descriptor', () => {
-    const r = resolveWriteColumn(cfg({
-      embedding_model: 'voyage:voyage-3-large',
-      embedding_dimensions: 1024,
-      embedding_columns: {
-        embedding_voyage: { provider: 'voyage:voyage-3-large', dimensions: 1024, type: 'vector' },
-      },
-    }));
-    expect(r).toEqual({
-      name: 'embedding_voyage',
-      type: 'vector',
-      dimensions: 1024,
-      embeddingModel: 'voyage:voyage-3-large',
-    });
-  });
-
-  test('provider match via gateway state (cfg.embedding_model unset) returns descriptor', () => {
-    configureGateway({
-      embedding_model: 'zeroentropyai:zembed-1',
-      embedding_dimensions: 2560,
-      env: {},
-    });
-    const r = resolveWriteColumn(cfg({
-      embedding_columns: {
-        embedding_ze: { provider: 'zeroentropyai:zembed-1', dimensions: 2560, type: 'halfvec' },
-      },
-    }));
-    expect(r).toEqual({
-      name: 'embedding_ze',
-      type: 'halfvec',
-      dimensions: 2560,
-      embeddingModel: 'zeroentropyai:zembed-1',
-    });
-  });
-
-  test('no provider match returns undefined instead of guessing a column', () => {
-    configureGateway({
-      embedding_model: 'zeroentropyai:zembed-1',
-      embedding_dimensions: 2560,
-      env: {},
-    });
-    const r = resolveWriteColumn(cfg({
-      embedding_columns: {
-        embedding_voyage: { provider: 'voyage:voyage-3-large', dimensions: 1024, type: 'vector' },
-      },
-    }));
-    expect(r).toBeUndefined();
-  });
-
-  test('only USER-declared columns are consulted — multimodal builtin never captures text writes', () => {
-    // Current model equals the embedding_image BUILTIN's provider; a registry
-    // walk that consulted builtins would misroute text writes into the image
-    // column. resolveWriteColumn must return undefined here.
-    configureGateway({
-      embedding_model: 'voyage:voyage-multimodal-3',
-      embedding_dimensions: 1024,
-      env: {},
-    });
-    const r = resolveWriteColumn(cfg({
-      embedding_columns: {
-        embedding_other: { provider: 'openai:text-embedding-3-large', dimensions: 1536, type: 'vector' },
-      },
-    }));
-    expect(r).toBeUndefined();
-  });
-
-  test('malformed registry entry throws loud (same validation as the read side)', () => {
-    expect(() => resolveWriteColumn(cfg({
-      embedding_model: 'voyage:voyage-3-large',
-      embedding_columns: {
-        'bad"col': { provider: 'voyage:voyage-3-large', dimensions: 1024, type: 'vector' },
-      } as never,
-    }))).toThrow(EmbeddingColumnConfigError);
-  });
-});