fix(dream): deterministic last-writer attribution for colliding slugs

collectChildPutPageSlugs paired each slug to a jobId first-seen over an
unordered result set. When two children collide on a final slug, the pages
row holds the LAST put_page write, so the provenance stamp
(transcript_id/transcript_source/date) could attribute an arbitrary other
transcript. ORDER BY id + last-writer-wins aligns the stamp with the
surviving content; pinned by a collision test.

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

src/core/ai/recipes/minimax.ts

@@ -1,100 +1,8 @@
 import type { Recipe } from '../types.ts';
 
 /**
- * MiniMax transport shim (#1977). MiniMax's `/v1/embeddings` endpoint is NOT
- * OpenAI-compatible at the wire level despite the recipe's
- * `implementation: 'openai-compatible'`:
- *  - Request: requires `texts` (the AI SDK sends `input`) plus an optional
- *    `type: 'db' | 'query'` asymmetric-retrieval field, and rejects OpenAI's
- *    `encoding_format`.
- *  - Response: returns `{vectors: number[][], total_tokens}` where the AI
- *    SDK's Zod schema expects `{data: [{embedding, index}], usage}`.
- *
- * Chat (`/chat/completions`) IS OpenAI-compatible, and this same fetch is
- * applied to every openai-compatible touchpoint by `applyOpenAICompatConfig`,
- * so everything outside the embeddings path passes through untouched — and
- * the response rewrite parses via `resp.clone()` only (never consume the
- * body of a response we return as-is; the DeepSeek shim rule). Fail-open:
- * any rewrite error returns the original request/response.
- *
- * @internal exported for tests.
- */
-// Cast through `unknown` because Bun's `typeof fetch` carries a `preconnect`
-// member the arrow function does not implement (matches deepseek.ts).
-export const minimaxCompatFetch = (async (
-  input: RequestInfo | URL,
-  init?: RequestInit,
-): Promise<Response> => {
-  const url =
-    typeof input === 'string' ? input : input instanceof URL ? input.href : input.url;
-  const isEmbeddings = url.includes('/embeddings');
-
-  // OUTBOUND (embeddings only): `input` → `texts`, default `type: 'db'`
-  // (the recipe's documented symmetric default — the AI SDK adapter strips
-  // the `type` threaded via providerOptions before it reaches the wire,
-  // same class as #1400), and drop `encoding_format` (not a MiniMax param).
-  if (isEmbeddings && init?.body && typeof init.body === 'string') {
-    try {
-      const parsed = JSON.parse(init.body);
-      if (
-        parsed && typeof parsed === 'object' &&
-        parsed.input !== undefined && parsed.texts === undefined
-      ) {
-        parsed.texts = Array.isArray(parsed.input) ? parsed.input : [parsed.input];
-        delete parsed.input;
-        delete parsed.encoding_format;
-        if (parsed.type === undefined) parsed.type = 'db';
-        // Drop Content-Length so fetch recomputes from the new body.
-        const headers = new Headers(init.headers ?? {});
-        headers.delete('content-length');
-        init = { ...init, body: JSON.stringify(parsed), headers };
-      }
-    } catch {
-      // Body wasn't JSON — pass through untouched.
-    }
-  }
-
-  const res = await fetch(input as any, init as any);
-
-  // INBOUND (embeddings only): `{vectors: [[...]]}` → `{data: [{embedding}]}`.
-  // Anything else (chat completions, MiniMax base_resp errors, non-JSON)
-  // returns the ORIGINAL response with its body unread.
-  if (!isEmbeddings || !res.ok) return res;
-  const ctype = res.headers.get('content-type') ?? '';
-  if (!ctype.toLowerCase().includes('application/json')) return res;
-  try {
-    const json = await res.clone().json();
-    if (!json || typeof json !== 'object' || !Array.isArray(json.vectors)) return res;
-    const totalTokens = typeof json.total_tokens === 'number' ? json.total_tokens : 0;
-    const rewritten = {
-      object: 'list',
-      data: (json.vectors as number[][]).map((embedding, index) => ({
-        object: 'embedding',
-        embedding,
-        index,
-      })),
-      model: typeof json.model === 'string' ? json.model : 'embo-01',
-      usage: { prompt_tokens: totalTokens, total_tokens: totalTokens },
-    };
-    // Fresh header set: the body changed, so upstream content-length /
-    // content-encoding would now be wrong.
-    const headers = new Headers(res.headers);
-    headers.delete('content-length');
-    headers.delete('content-encoding');
-    return new Response(JSON.stringify(rewritten), {
-      status: res.status,
-      statusText: res.statusText,
-      headers,
-    });
-  } catch {
-    return res;
-  }
-}) as unknown as typeof fetch;
-
-/**
- * MiniMax (海螺AI). `/embeddings` endpoint at api.minimaxi.com (wire shape
- * normalized by `minimaxCompatFetch` above); OpenAI-compatible
- * `/chat/completions`. The flagship embedding model is `embo-01` (1536 dims).
+ * MiniMax (海螺AI). OpenAI-compatible /embeddings endpoint at
+ * api.minimax.chat. The flagship embedding model is `embo-01` (1536 dims).
  *
  * MiniMax's API takes an extra `type: 'db' | 'query'` field for asymmetric
  * retrieval. gbrain currently has no notion of "this is a document vs a
@@ -130,25 +38,7 @@ export const minimax: Recipe = {
       // halving in the gateway catches token-limit errors at runtime.
       max_batch_tokens: 4096,
     },
-    chat: {
-      // Model list from MiniMax's /v1/models (#1977). Chat is genuinely
-      // OpenAI-compatible — no wire rewrite needed (minimaxCompatFetch
-      // passes non-embedding requests through untouched).
-      models: [
-        'MiniMax-M3',
-        'MiniMax-M2.7',
-        'MiniMax-M2.7-highspeed',
-        'MiniMax-M2.5',
-        'MiniMax-M2.5-highspeed',
-        'MiniMax-M2.1',
-        'MiniMax-M2.1-highspeed',
-        'MiniMax-M2',
-      ],
-      supports_tools: false,
-      supports_subagent_loop: false,
-    },
   },
   setup_hint:
     'Get an API key at https://www.minimaxi.com, then `export MINIMAX_API_KEY=...`',
-  compat: { fetch: minimaxCompatFetch },
 };

src/core/cycle/synthesize.ts

@@ -428,8 +428,13 @@ export async function runPhaseSynthesize(
 
     const queue = new MinionQueue(engine);
     const childIds: number[] = [];
-    /** Map child job_id → chunk metadata for D6 orchestrator-side slug rewrite. */
-    const chunkInfo = new Map<number, { idx: number; hash6: string }>();
+    /**
+     * Map child job_id → transcript metadata. Drives D6 orchestrator-side
+     * slug rewrite for chunked transcripts AND the deterministic frontmatter
+     * stampDreamProvenance merges into each written page. Populated for
+     * every child (single-chunk children carry chunkTotal=1).
+     */
+    const childMeta = new Map<number, ChildMeta>();
     /** Skip reasons for the cycle report (D5 cap hits, D8 legacy-key skips). */
     const skipReports: Array<{ filePath: string; reason: string }> = [];
 
@@ -513,9 +518,14 @@ export async function runPhaseSynthesize(
           { allowProtectedSubmit: true },
         );
         childIds.push(child.id);
-        if (isChunked) {
-          chunkInfo.set(child.id, { idx: i, hash6 });
-        }
+        childMeta.set(child.id, {
+          idx: i,
+          hash6,
+          chunkTotal: chunks.length,
+          transcriptSource: t.transcriptSource,
+          transcriptId: stripContentVersionSuffix(t.basename),
+          inferredDate: t.inferredDate,
+        });
       }
     }
 
@@ -544,14 +554,14 @@ export async function runPhaseSynthesize(
 
     // Collect slugs from put_page tool executions across the children
     // (codex finding #2: deterministic provenance, NOT pages.updated_at).
-    // D6 orchestrator slug rewrite: chunkInfo drives post-hoc rewrite of
+    // D6 orchestrator slug rewrite: childMeta drives post-hoc rewrite of
     // bare-hash slugs to `<hash6>-c<idx>` so chunked siblings can't collide
     // even if Sonnet drops the chunk suffix.
     // v0.32.8: refs carry source_id so reverseWriteRefs picks the correct
     // (source, slug) row. #1586: refs are stamped with the cycle's resolved
     // source (children write there via SubagentHandlerData.source_id).
     const cycleSourceId = opts.sourceId ?? 'default';
-    const writtenRefs = await collectChildPutPageSlugs(engine, childIds, chunkInfo, cycleSourceId);
+    const writtenRefs = await collectChildPutPageSlugs(engine, childIds, childMeta, cycleSourceId);
 
     const summaryDate = opts.date ?? today();
 
@@ -559,7 +569,12 @@ export async function runPhaseSynthesize(
     // of every child-written page BEFORE reverse-rendering, so generated pages
     // are queryable (`frontmatter->>'dream_generated'`) and a later put_page
     // write-through (which re-renders from the DB row) can't erase the stamp.
-    await stampDreamProvenance(engine, writtenRefs, summaryDate);
+    // #2285: the stamp also carries the orchestrator-owned deterministic
+    // frontmatter (transcript_id, transcript_source, transcript_hash, date,
+    // chunk) derived from childMeta — subagent drift on those fields can't
+    // leak, and reverseWriteRefs below re-reads the row so the same fields
+    // land in the on-disk markdown.
+    await stampDreamProvenance(engine, writtenRefs, summaryDate, childMeta);
 
     // Dual-write: reverse-render each DB row → markdown file.
     const reverseWriteCount = await reverseWriteRefs(engine, opts.brainDir, writtenRefs, cycleSourceId);
@@ -1095,15 +1110,17 @@ function sanitizeForSlug(s: string): string {
  * fake"): we no longer need detection because the rewrite enforces
  * uniqueness at slug-write time.
  *
- * `chunkInfo` maps child job_id → { chunk_index, hash6 }. Single-chunk
- * children are absent from the map and pass through unchanged.
+ * `childMeta` maps child job_id → per-child transcript metadata. Chunked
+ * children (chunkTotal > 1) get the slug rewrite; single-chunk children
+ * pass through unchanged. Each returned ref carries the job_id that wrote
+ * it so stampDreamProvenance can pair the slug back to its childMeta entry.
  */
 async function collectChildPutPageSlugs(
   engine: BrainEngine,
   childIds: number[],
-  chunkInfo: Map<number, { idx: number; hash6: string }>,
+  childMeta: Map<number, ChildMeta>,
   sourceId = 'default',
-): Promise<Array<{ slug: string; source_id: string }>> {
+): Promise<Array<{ slug: string; source_id: string; jobId: number }>> {
   if (childIds.length === 0) return [];
   // Raw fetch — NO SELECT DISTINCT. Preserves per-child slug duplicates so
   // the orchestrator sees what each child wrote. COALESCE handles both
@@ -1122,16 +1139,73 @@ async function collectChildPutPageSlugs(
        FROM subagent_tool_executions
       WHERE job_id = ANY($1::int[])
         AND tool_name = 'brain_put_page'
-        AND status = 'complete'`,
+        AND status = 'complete'
+      ORDER BY id`,
     [childIds],
   );
-  const rewritten = new Set<string>();
+  const rewritten = new Map<string, number>();
   for (const r of rows) {
     if (typeof r.slug !== 'string' || r.slug.length === 0) continue;
-    const ci = chunkInfo.get(r.job_id);
-    rewritten.add(ci ? rewriteChunkedSlug(r.slug, ci.hash6, ci.idx) : r.slug);
+    const meta = childMeta.get(r.job_id);
+    const finalSlug = meta && meta.chunkTotal > 1
+      ? rewriteChunkedSlug(r.slug, meta.hash6, meta.idx)
+      : r.slug;
+    // Last writer wins, in execution-row order (ORDER BY id): if two children
+    // collide on a final slug, the pages row holds the LAST put_page write, so
+    // the stamp must attribute that child's transcript — not an arbitrary one.
+    rewritten.set(finalSlug, r.job_id);
   }
-  return Array.from(rewritten).sort().map(slug => ({ slug, source_id: sourceId }));
+  return [...rewritten.entries()]
+    .sort(([a], [b]) => a.localeCompare(b))
+    .map(([slug, jobId]) => ({ slug, source_id: sourceId, jobId }));
+}
+
+/**
+ * Per-child orchestrator state. Drives D6 chunked-slug rewrite (idx + hash6)
+ * AND the deterministic frontmatter stampDreamProvenance merges into each
+ * written page. Populated for every child, not just chunked ones.
+ */
+interface ChildMeta {
+  idx: number;
+  hash6: string;
+  chunkTotal: number;
+  transcriptSource: string | null;
+  transcriptId: string;
+  inferredDate: string | null;
+}
+
+/**
+ * Strip the content-version suffix that claude-code-archive appends when a
+ * conversation is edited (`<uuid>--<contentHash>.md`). The session UUID is
+ * the stable transcript identifier; the suffix changes with content. Used to
+ * populate `transcript_id` so edits of the same session collapse to one id.
+ */
+function stripContentVersionSuffix(basename: string): string {
+  return basename.replace(/--[a-f0-9]+$/i, '');
+}
+
+/**
+ * Deterministic frontmatter for one synthesized page (#2285). Every field
+ * here is owned by the orchestrator — the subagent's value for any of these
+ * is overwritten. The subagent retains authority over type / title / tags /
+ * body. `date` feeds the effective-date precedence chain
+ * (src/core/effective-date.ts) so re-imports keep the conversation date even
+ * when sync tools re-stamp file mtimes.
+ */
+function buildDeterministicFrontmatter(
+  meta: ChildMeta,
+  cycleDate: string,
+): Record<string, unknown> {
+  const overrides: Record<string, unknown> = {
+    dream_generated: true,
+    dream_cycle_date: cycleDate,
+    transcript_id: meta.transcriptId,
+    transcript_hash: meta.hash6,
+  };
+  if (meta.transcriptSource) overrides.transcript_source = meta.transcriptSource;
+  if (meta.chunkTotal > 1) overrides.chunk = `${meta.idx + 1}/${meta.chunkTotal}`;
+  if (meta.inferredDate) overrides.date = meta.inferredDate;
+  return overrides;
 }
 
 /**
@@ -1177,12 +1251,19 @@ async function hasLegacySingleChunkCompletion(
  */
 async function stampDreamProvenance(
   engine: BrainEngine,
-  refs: Array<{ slug: string; source_id: string }>,
+  refs: Array<{ slug: string; source_id: string; jobId?: number }>,
   cycleDate: string,
+  childMeta?: Map<number, ChildMeta>,
 ): Promise<void> {
   if (refs.length === 0) return;
   const { executeRawJsonb } = await import('../sql-query.ts');
-  for (const { slug, source_id } of refs) {
+  for (const { slug, source_id, jobId } of refs) {
+    // #2285: when the ref pairs back to a child, the stamp also carries the
+    // orchestrator-owned deterministic frontmatter for that transcript.
+    const meta = jobId !== undefined ? childMeta?.get(jobId) : undefined;
+    const stamp = meta
+      ? buildDeterministicFrontmatter(meta, cycleDate)
+      : { dream_generated: true, dream_cycle_date: cycleDate };
     try {
       await executeRawJsonb(
         engine,
@@ -1190,7 +1271,7 @@ async function stampDreamProvenance(
             SET frontmatter = COALESCE(frontmatter, '{}'::jsonb) || $3::jsonb
           WHERE slug = $1 AND source_id = $2`,
         [slug, source_id],
-        [{ dream_generated: true, dream_cycle_date: cycleDate }],
+        [stamp],
       );
     } catch (e) {
       const msg = e instanceof Error ? e.message : String(e);

src/core/cycle/transcript-discovery.ts

@@ -10,7 +10,7 @@
  */
 
 import { readFileSync, readdirSync, statSync } from 'node:fs';
-import { join, basename } from 'node:path';
+import { join, basename, dirname } from 'node:path';
 import { createHash } from 'node:crypto';
 import { pruneDir } from '../sync.ts';
 
@@ -23,8 +23,22 @@ export interface DiscoveredTranscript {
   content: string;
   /** Filename basename without extension; used as a topic-slug seed. */
   basename: string;
-  /** Inferred date if the basename matches `YYYY-MM-DD...` (or null). */
+  /**
+   * Inferred conversation date (YYYY-MM-DD) or null. Precedence: the
+   * `| First message | <ISO> |` row in the transcript's `## Metadata`
+   * table (stable across mtime-restamping re-syncs) wins; a leading
+   * `YYYY-MM-DD` in the basename is the fallback.
+   */
   inferredDate: string | null;
+  /**
+   * Transcript source archive name, derived from the path's grandparent
+   * directory (the immediate parent of the date directory). For the
+   * canonical layout `<corpus>/<source>/<date>/<id>.md` this yields the
+   * source-name segment — e.g. `claude-code` for the claude-code-archive
+   * output, `meetings` for meeting recordings. Null when the file does
+   * not live under a `<source>/<date>/` pair (ad-hoc inputs).
+   */
+  transcriptSource: string | null;
 }
 
 export interface DiscoverOpts {
@@ -161,6 +175,36 @@ function matchesAnyExclude(text: string, patterns: RegExp[]): boolean {
   return false;
 }
 
+/**
+ * Content-based conversation date: the `| First message | <ISO timestamp> |`
+ * row claude-code-archive writes into the transcript's `## Metadata` table.
+ * Stable across rsync/Dropbox/Syncthing/B2 re-syncs that re-stamp mtime,
+ * unlike anything derived from file metadata. Returns YYYY-MM-DD or null.
+ */
+const FIRST_MESSAGE_RE = /^\|\s*First message\s*\|\s*(\d{4}-\d{2}-\d{2})/im;
+
+export function inferContentDate(content: string): string | null {
+  const m = FIRST_MESSAGE_RE.exec(content);
+  return m ? m[1] : null;
+}
+
+/**
+ * Derive the archive source name from a transcript path. Returns the basename
+ * of the directory two levels above the file when the immediate parent is a
+ * date directory and the grandparent looks like a source-name slug (lowercase
+ * alphanumeric segments separated by hyphens); otherwise null. This pins the
+ * canonical claude-code-archive layout `<corpus>/<source>/<date>/<id>.md`
+ * without claiming a source for ad-hoc inputs that don't match.
+ */
+export function deriveTranscriptSource(filePath: string): string | null {
+  const parentName = basename(dirname(filePath));
+  if (!/^\d{4}-\d{2}-\d{2}/.test(parentName)) return null;
+  const grandparentName = basename(dirname(dirname(filePath)));
+  if (!grandparentName) return null;
+  if (!/^[a-z0-9]+(-[a-z0-9]+)*$/.test(grandparentName)) return null;
+  return grandparentName;
+}
+
 function listTextFiles(dir: string): string[] {
   // Recursive walk with descent-time pruning (closes codex C12/C13 spec gap).
   // Accepts BOTH .txt and .md per transcript-discovery's domain rules — does
@@ -225,8 +269,11 @@ export function discoverTranscripts(opts: DiscoverOpts): DiscoveredTranscript[]
       const ext = filePath.endsWith('.md') ? '.md' : '.txt';
       const baseName = basename(filePath, ext);
       const dateMatch = DATE_RE.exec(baseName);
-      const inferredDate = dateMatch ? dateMatch[1] : null;
-      if (!isInDateRange(inferredDate, opts)) continue;
+      const filenameDate = dateMatch ? dateMatch[1] : null;
+      // Fast path: date-named files outside the window skip before the read.
+      // ponytail: a date-named file whose content date differs is filtered on
+      // its filename date — acceptable; archive layouts use UUID basenames.
+      if (filenameDate && !isInDateRange(filenameDate, opts)) continue;
 
       let content: string;
       try {
@@ -241,12 +288,17 @@ export function discoverTranscripts(opts: DiscoverOpts): DiscoveredTranscript[]
       }
       if (matchesAnyExclude(content, excludeRes)) continue;
 
+      // Content-metadata date wins (survives mtime restamps); filename next.
+      const inferredDate = inferContentDate(content) ?? filenameDate;
+      if (!isInDateRange(inferredDate, opts)) continue;
+
       results.push({
         filePath,
         contentHash: hashContent(content),
         content,
         basename: baseName,
         inferredDate,
+        transcriptSource: deriveTranscriptSource(filePath),
       });
     }
   }
@@ -290,6 +342,7 @@ export function readSingleTranscript(
     contentHash: hashContent(content),
     content,
     basename: baseName,
-    inferredDate: dateMatch ? dateMatch[1] : null,
+    inferredDate: inferContentDate(content) ?? (dateMatch ? dateMatch[1] : null),
+    transcriptSource: deriveTranscriptSource(filePath),
   };
 }

test/ai/recipe-minimax.test.ts

@@ -6,14 +6,10 @@
  *  - default auth: MINIMAX_API_KEY → "Bearer <key>"; missing → AIConfigError
  *  - dimsProviderOptions threads `type: 'db'` for embo-01 (the asymmetric
  *    retrieval field default) — pins the v1 indexing-only behavior
- *  - #1977: chat touchpoint declared; minimaxCompatFetch rewrites the
- *    embedding wire shape both directions, passes chat through with the
- *    response body UNREAD (the consumed-body regression), fail-open.
  */
 
-import { afterEach, describe, expect, test } from 'bun:test';
+import { describe, expect, test } from 'bun:test';
 import { getRecipe } from '../../src/core/ai/recipes/index.ts';
-import { minimaxCompatFetch } from '../../src/core/ai/recipes/minimax.ts';
 import { defaultResolveAuth } from '../../src/core/ai/gateway.ts';
 import { dimsProviderOptions } from '../../src/core/ai/dims.ts';
 import { AIConfigError } from '../../src/core/ai/errors.ts';
@@ -60,106 +56,4 @@ describe('recipe: minimax', () => {
     expect(dimsProviderOptions('openai-compatible', 'voyage-3-lite', 512)).toBeUndefined();
     expect(dimsProviderOptions('openai-compatible', 'nomic-embed-text', 768)).toBeUndefined();
   });
-
-  test('chat touchpoint declared (#1977) so assertTouchpoint permits gbrain think', () => {
-    const r = getRecipe('minimax')!;
-    expect(r.touchpoints.chat).toBeDefined();
-    expect(r.touchpoints.chat!.models).toContain('MiniMax-M3');
-    expect(r.touchpoints.chat!.supports_tools).toBe(false);
-    expect(r.touchpoints.chat!.supports_subagent_loop).toBe(false);
-  });
-
-  test('recipe ships minimaxCompatFetch via compat.fetch (no env-templated base URL)', () => {
-    const r = getRecipe('minimax')!;
-    expect(r.compat?.fetch).toBe(minimaxCompatFetch);
-    // base_urls config override must keep working: no resolveOpenAICompatConfig.
-    expect(r.resolveOpenAICompatConfig).toBeUndefined();
-  });
-});
-
-describe('minimaxCompatFetch (#1977)', () => {
-  const realFetch = globalThis.fetch;
-  afterEach(() => { globalThis.fetch = realFetch; });
-
-  function stubFetch(body: unknown, init?: { status?: number; contentType?: string }) {
-    const calls: { url: string; init?: RequestInit }[] = [];
-    globalThis.fetch = (async (input: any, i?: RequestInit) => {
-      calls.push({ url: String(input), init: i });
-      return new Response(typeof body === 'string' ? body : JSON.stringify(body), {
-        status: init?.status ?? 200,
-        headers: { 'content-type': init?.contentType ?? 'application/json' },
-      });
-    }) as unknown as typeof fetch;
-    return calls;
-  }
-
-  const EMBED_URL = 'https://api.minimaxi.com/v1/embeddings';
-  const CHAT_URL = 'https://api.minimaxi.com/v1/chat/completions';
-
-  test('embedding request: input → texts, type:db injected, encoding_format dropped', async () => {
-    const calls = stubFetch({ vectors: [[0.1, 0.2]] });
-    await minimaxCompatFetch(EMBED_URL, {
-      method: 'POST',
-      headers: { 'content-type': 'application/json', 'content-length': '99' },
-      body: JSON.stringify({ model: 'embo-01', input: ['hello', 'world'], encoding_format: 'float' }),
-    });
-    const wire = JSON.parse(calls[0]!.init!.body as string);
-    expect(wire.texts).toEqual(['hello', 'world']);
-    expect(wire.input).toBeUndefined();
-    expect(wire.encoding_format).toBeUndefined();
-    expect(wire.type).toBe('db');
-    expect(new Headers(calls[0]!.init!.headers).get('content-length')).toBeNull();
-  });
-
-  test('embedding response: {vectors} rewritten to OpenAI {data:[{embedding}]}', async () => {
-    stubFetch({ vectors: [[0.1, 0.2], [0.3, 0.4]], total_tokens: 7 });
-    const res = await minimaxCompatFetch(EMBED_URL, {
-      method: 'POST',
-      body: JSON.stringify({ model: 'embo-01', input: ['a', 'b'] }),
-    });
-    const json = await res.json();
-    expect(json.data).toEqual([
-      { object: 'embedding', embedding: [0.1, 0.2], index: 0 },
-      { object: 'embedding', embedding: [0.3, 0.4], index: 1 },
-    ]);
-    expect(json.usage).toEqual({ prompt_tokens: 7, total_tokens: 7 });
-  });
-
-  test('chat completion passes through with body UNREAD (consumed-body regression)', async () => {
-    stubFetch({ choices: [{ message: { role: 'assistant', content: 'hi' } }] });
-    const res = await minimaxCompatFetch(CHAT_URL, {
-      method: 'POST',
-      body: JSON.stringify({ model: 'MiniMax-M3', messages: [{ role: 'user', content: 'say hi' }] }),
-    });
-    expect(res.bodyUsed).toBe(false); // the broken PR #2882 wrapper consumed this
-    const json = await res.json(); // must NOT throw "Body already used"
-    expect(json.choices[0].message.content).toBe('hi');
-  });
-
-  test('chat request body is never rewritten (messages untouched, no type injected)', async () => {
-    const calls = stubFetch({ choices: [] });
-    const body = JSON.stringify({ model: 'MiniMax-M3', messages: [{ role: 'user', content: 'x' }] });
-    await minimaxCompatFetch(CHAT_URL, { method: 'POST', body });
-    expect(calls[0]!.init!.body).toBe(body);
-  });
-
-  test('embedding error response ({vectors:null, base_resp}) passes through re-readable', async () => {
-    stubFetch({ vectors: null, base_resp: { status_code: 2013, status_msg: 'invalid params' } });
-    const res = await minimaxCompatFetch(EMBED_URL, {
-      method: 'POST',
-      body: JSON.stringify({ model: 'embo-01', input: ['a'] }),
-    });
-    expect(res.bodyUsed).toBe(false);
-    const json = await res.json();
-    expect(json.base_resp.status_code).toBe(2013);
-  });
-
-  test('fail-open: non-JSON response body passes through untouched', async () => {
-    stubFetch('not json', { contentType: 'application/json' });
-    const res = await minimaxCompatFetch(EMBED_URL, {
-      method: 'POST',
-      body: JSON.stringify({ model: 'embo-01', input: ['a'] }),
-    });
-    expect(await res.text()).toBe('not json');
-  });
 });

test/cycle-dream-output-root.test.ts

@@ -26,6 +26,7 @@ const transcript: DiscoveredTranscript = {
   content: 'User: hello world',
   contentHash: 'abcdef0123456789',
   inferredDate: '2026-07-17',
+  transcriptSource: null,
 } as DiscoveredTranscript;
 
 describe('#2415: buildSynthesisPrompt output root', () => {

test/cycle-synthesize-slug-collection.test.ts

@@ -152,3 +152,94 @@ describe('#2569: stampDreamProvenance persists the marker into DB frontmatter',
     await stampDreamProvenance(engine as any, refs, '2026-07-17'); // idempotent
   });
 });
+
+describe('#2285: orchestrator-owned deterministic transcript frontmatter', () => {
+  const meta = {
+    idx: 1,
+    hash6: 'abc123',
+    chunkTotal: 3,
+    transcriptSource: 'claude-code',
+    transcriptId: 'session-uuid',
+    inferredDate: '2026-05-15',
+  };
+
+  test('collectChildPutPageSlugs pairs each ref back to the writing job', async () => {
+    const refs = await collectChildPutPageSlugs(
+      engine as any, [1001], new Map([[1001, { ...meta, chunkTotal: 1 }]]), 'mybrain',
+    );
+    expect(refs.length).toBeGreaterThan(0);
+    for (const r of refs) {
+      expect(r.jobId).toBe(1001);
+      expect(r.source_id).toBe('mybrain'); // #1586: cycle source, never hardcoded 'default'
+    }
+  });
+
+  test('slug collision across children attributes the LAST writer (matches surviving putPage)', async () => {
+    const db = (engine as any).db;
+    // Jobs 1001 then 1002 write the same slug; the pages row would hold
+    // 1002's content (last put_page wins), so the ref must carry jobId 1002.
+    await db.query(
+      `INSERT INTO subagent_tool_executions (job_id, message_idx, tool_use_id, tool_name, status, input)
+       VALUES (1001, 9, 'tool_dup_a', 'brain_put_page', 'complete', $1::jsonb)`,
+      [JSON.stringify({ slug: 'wiki/agents/test/collision', body: 'first' })],
+    );
+    await db.query(
+      `INSERT INTO subagent_tool_executions (job_id, message_idx, tool_use_id, tool_name, status, input)
+       VALUES (1002, 9, 'tool_dup_b', 'brain_put_page', 'complete', $1::jsonb)`,
+      [JSON.stringify({ slug: 'wiki/agents/test/collision', body: 'second' })],
+    );
+    const refs = await collectChildPutPageSlugs(engine as any, [1001, 1002], new Map());
+    const hit = refs.find((r: { slug: string }) => r.slug === 'wiki/agents/test/collision');
+    expect(hit?.jobId).toBe(1002);
+  });
+
+  test('stampDreamProvenance merges the transcript metadata into DB frontmatter', async () => {
+    const slug = 'wiki/originals/ideas/2026-07-17-transcript-meta-abc123';
+    await engine.putPage(slug, {
+      type: 'note',
+      title: 'Meta stamp',
+      compiled_truth: 'body',
+      timeline: '',
+      frontmatter: { keep_me: 'yes', transcript_id: 'subagent-drift' },
+    });
+    await stampDreamProvenance(
+      engine as any,
+      [{ slug, source_id: 'default', jobId: 42 }],
+      '2026-07-17',
+      new Map([[42, meta]]),
+    );
+    const rows = await engine.executeRaw<{ fm: Record<string, unknown> }>(
+      `SELECT frontmatter AS fm FROM pages WHERE slug = $1`, [slug],
+    );
+    const fm = rows[0].fm as Record<string, unknown>;
+    expect(fm.dream_generated).toBe(true);
+    expect(fm.dream_cycle_date).toBe('2026-07-17');
+    expect(fm.transcript_id).toBe('session-uuid'); // orchestrator wins over subagent drift
+    expect(fm.transcript_hash).toBe('abc123');
+    expect(fm.transcript_source).toBe('claude-code');
+    expect(fm.chunk).toBe('2/3');
+    expect(fm.date).toBe('2026-05-15');
+    expect(fm.keep_me).toBe('yes'); // subagent-owned keys survive
+  });
+
+  test('single-chunk children with no inferredDate stamp only the applicable fields', async () => {
+    const slug = 'wiki/originals/ideas/2026-07-17-minimal-meta-abc123';
+    await engine.putPage(slug, {
+      type: 'note', title: 'Minimal', compiled_truth: 'b', timeline: '', frontmatter: {},
+    });
+    await stampDreamProvenance(
+      engine as any,
+      [{ slug, source_id: 'default', jobId: 43 }],
+      '2026-07-17',
+      new Map([[43, { ...meta, chunkTotal: 1, transcriptSource: null, inferredDate: null }]]),
+    );
+    const rows = await engine.executeRaw<{ fm: Record<string, unknown> }>(
+      `SELECT frontmatter AS fm FROM pages WHERE slug = $1`, [slug],
+    );
+    const fm = rows[0].fm as Record<string, unknown>;
+    expect(fm.transcript_id).toBe('session-uuid');
+    expect(fm.chunk).toBeUndefined();
+    expect(fm.transcript_source).toBeUndefined();
+    expect(fm.date).toBeUndefined();
+  });
+});

test/cycle-synthesize.test.ts

@@ -302,6 +302,7 @@ describe('judgeSignificance', () => {
       content: 'A short conversation about something interesting.',
       basename: 'x',
       inferredDate: null,
+      transcriptSource: null,
     };
   }
 
@@ -415,6 +416,7 @@ describe('judgeSignificance — UTF-16 safety (v0.41.13)', () => {
       content,
       basename: 'long',
       inferredDate: null,
+      transcriptSource: null,
     };
   }
 

test/cycle/synthesize-gateway-adapter.test.ts

@@ -45,6 +45,7 @@ const FIXTURE_TRANSCRIPT: DiscoveredTranscript = {
   content: 'Synthetic transcript content for gateway-adapter parity tests.',
   contentHash: 'sha-fixture-1',
   inferredDate: '2026-05-24',
+  transcriptSource: null,
 };
 
 describe('makeJudgeClient — construction-time provider probe', () => {

test/cycle/transcript-discovery-metadata.test.ts

@@ -0,0 +1,109 @@
+/**
+ * #2285 — transcript metadata discovery.
+ *
+ * Pins the two discovery-side additions:
+ *   1. `transcriptSource` — derived from the `<source>/<date>/<file>` path
+ *      layout; null for ad-hoc inputs that don't match.
+ *   2. Content-based date inference — the `| First message | <ISO> |` row in
+ *      the transcript's `## Metadata` table wins over the filename-regex
+ *      date (stable across mtime-restamping re-syncs); filename is the
+ *      fallback.
+ *
+ * Pure filesystem; no engine, no LLM.
+ */
+
+import { describe, test, expect, beforeEach, afterEach } from 'bun:test';
+import { mkdtempSync, rmSync, writeFileSync, mkdirSync } from 'node:fs';
+import { tmpdir } from 'node:os';
+import { join, dirname } from 'node:path';
+import {
+  discoverTranscripts,
+  readSingleTranscript,
+  deriveTranscriptSource,
+  inferContentDate,
+} from '../../src/core/cycle/transcript-discovery.ts';
+
+let tmpDir: string;
+
+beforeEach(() => {
+  tmpDir = mkdtempSync(join(tmpdir(), 'gbrain-transcript-meta-'));
+});
+
+afterEach(() => {
+  rmSync(tmpDir, { recursive: true, force: true });
+});
+
+function write(relPath: string, body: string): string {
+  const full = join(tmpDir, relPath);
+  mkdirSync(dirname(full), { recursive: true });
+  writeFileSync(full, body);
+  return full;
+}
+
+const FILLER = 'User: hello world. '.repeat(200);
+const METADATA_BLOCK =
+  '## Metadata\n\n| Key | Value |\n| --- | --- |\n| First message | 2026-05-15T03:51:11.584Z |\n\n';
+
+describe('deriveTranscriptSource', () => {
+  test('extracts the source slug from <source>/<date>/<file> layout', () => {
+    expect(deriveTranscriptSource('/corpus/claude-code/2026-06-12/abc.md')).toBe('claude-code');
+    expect(deriveTranscriptSource('/corpus/voice-notes/2026-06-12/xyz.md')).toBe('voice-notes');
+  });
+
+  test('null when the parent dir is not a date dir or grandparent is not a slug', () => {
+    expect(deriveTranscriptSource('/corpus/flat-file.md')).toBeNull();
+    expect(deriveTranscriptSource('/corpus/claude-code/not-a-date/abc.md')).toBeNull();
+    expect(deriveTranscriptSource('/corpus/Not A Slug/2026-06-12/abc.md')).toBeNull();
+  });
+});
+
+describe('inferContentDate', () => {
+  test('parses the | First message | row', () => {
+    expect(inferContentDate(METADATA_BLOCK)).toBe('2026-05-15');
+  });
+  test('null when absent', () => {
+    expect(inferContentDate(FILLER)).toBeNull();
+  });
+});
+
+describe('discoverTranscripts — transcriptSource + date cascade', () => {
+  test('populates transcriptSource per file; null for flat files', () => {
+    write('claude-code/2026-06-12/aaaa.md', FILLER);
+    write('2026-06-12-flat.md', FILLER);
+    const out = discoverTranscripts({ corpusDir: tmpDir, minChars: 100 });
+    const byBase = new Map(out.map(t => [t.basename, t.transcriptSource]));
+    expect(byBase.get('aaaa')).toBe('claude-code');
+    expect(byBase.get('2026-06-12-flat')).toBeNull();
+  });
+
+  test('content First-message date wins over the filename date', () => {
+    write('2026-01-01-named.md', METADATA_BLOCK + FILLER);
+    const out = discoverTranscripts({ corpusDir: tmpDir, minChars: 100 });
+    expect(out).toHaveLength(1);
+    expect(out[0].inferredDate).toBe('2026-05-15');
+  });
+
+  test('filename date remains the fallback when content has no metadata row', () => {
+    write('2026-01-01-named.md', FILLER);
+    const out = discoverTranscripts({ corpusDir: tmpDir, minChars: 100 });
+    expect(out[0].inferredDate).toBe('2026-01-01');
+  });
+
+  test('date filter matches on the content date for UUID-named transcripts', () => {
+    write('claude-code/2026-05-15/uuid-basename.md', METADATA_BLOCK + FILLER);
+    const hit = discoverTranscripts({ corpusDir: tmpDir, minChars: 100, date: '2026-05-15' });
+    expect(hit).toHaveLength(1);
+    const miss = discoverTranscripts({ corpusDir: tmpDir, minChars: 100, date: '2026-05-16' });
+    expect(miss).toHaveLength(0);
+  });
+});
+
+describe('readSingleTranscript — same metadata surface', () => {
+  test('carries transcriptSource and prefers the content date', () => {
+    const p = write('claude-code/2026-05-15/2026-01-01-single.md', METADATA_BLOCK + FILLER);
+    const t = readSingleTranscript(p, { minChars: 100 });
+    expect(t).not.toBeNull();
+    expect(t!.transcriptSource).toBe('claude-code');
+    expect(t!.inferredDate).toBe('2026-05-15');
+  });
+});

test/progress.test.ts

@@ -216,17 +216,21 @@ describe('progress reporter', () => {
   });
 
   test('only one process-level signal handler installed across many reporters', () => {
-    // Baseline: one handler already installed by prior tests in this file.
+    // Baseline: one handler already installed by prior tests in this file, and
+    // possibly live reporters from OTHER test files sharing this bun process
+    // (shard composition is not this test's invariant — assert the delta, not
+    // an absolute zero, or shard reshuffles make this fail spuriously).
     const installedBefore = __signalHandlerInstalledForTest();
+    const liveBefore = __liveReporterCountForTest();
     const { stream } = sink(false);
     for (let i = 0; i < 50; i++) {
       const p = createProgress({ mode: 'json', stream, minIntervalMs: 0, minItems: 1 });
       p.start(`phase_${i}`, 1);
       p.finish();
     }
-    // After 50 reporter lifecycles, still exactly one handler and zero leaked live entries.
+    // After 50 reporter lifecycles, still exactly one handler and zero NET leaked live entries.
     expect(__signalHandlerInstalledForTest()).toBe(installedBefore || true);
-    expect(__liveReporterCountForTest()).toBe(0);
+    expect(__liveReporterCountForTest()).toBe(liveBefore);
   });
 
   test('startHeartbeat() fires heartbeats and stop() clears', async () => {