docs: cross-model doc-review fixes — cache-key phrasing, counter scope, slug-fence allow-list, entry date

- KEY_FILES.md: the per-call reranker-override cache-skip rationale now says
  the rrm= key part is bundle/config-resolved (a reranker field IS in the
  key; it just never reflects a per-call override); cross-source drop prose
  reworded to current-state (clears the doc guard's history-marker warn).
- README.md: scope the skipped-candidate counting to the DB extract paths
  (--source db, --stale) — the default fs walk can't see other sources.
- brains-and-sources.md: the slug-fence bullet now matches
  CLIENT_FENCED_WRITE_OPS — a fenced allow-list of write ops is available to
  slug-bound clients and everything else non-read is refused (fail-closed),
  instead of the self-contradictory "non-read operations are refused".
- CHANGELOG.md: 0.42.78.0 entry date 2026-08-09 → 2026-08-08 (matches the
  release commit and adjacent entries; header only, content untouched).
- llms.txt / llms-full.txt regenerated.

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

CHANGELOG.md

@@ -2,7 +2,7 @@
 
 All notable changes to GBrain will be documented in this file.
 
-## [0.42.78.0] - 2026-08-09
+## [0.42.78.0] - 2026-08-08
 
 **Search stops collapsing rich result pools on weak matches, cross-source wikilinks stop vanishing silently, and the remote-access rules are written down in one place.**
 

README.md

@@ -264,7 +264,7 @@ The whole loop is described in [`docs/architecture/topologies.md`](docs/architec
 
 **Hybrid search.** Vector (HNSW on pgvector) + BM25 keyword + reciprocal-rank fusion + source-tier boost + intent-aware query rewriting. Three named search modes (`conservative`, `balanced`, `tokenmax`) bundle the cost/quality knobs into a single config key. Live cost/recall comparisons in [`docs/eval/SEARCH_MODE_METHODOLOGY.md`](docs/eval/SEARCH_MODE_METHODOLOGY.md). Default: `balanced` with ZeroEntropy reranker on. Per-query graph signals notice when a top result is a hub for THAT query (adjacency boost), is corroborated across team brains (cross-source boost), or is being crowded out by weak chunks from a chatty session (session demote). Run `gbrain search "<query>" --explain` to see per-stage attribution: base score, every boost that fired, what it multiplied. `gbrain doctor` ships a `graph_signals_coverage` check; `gbrain search stats` shows fire counts and failure breakdowns. Vector retrieval pools the best chunk per page, so a page surfaces on its strongest evidence instead of losing to a neighbor on one weak chunk. Queries that match a page's title phrase or a declared free-text alias (`gbrain reindex --aliases` backfills existing pages) get boosted to the page they name. Every result carries an `evidence` tag (why it matched) and a `create_safety` hint (`exists` / `probable` / `unknown`) so an agent decides whether a page already exists instead of guessing from a raw score. `gbrain search diagnose "<query>" --target <slug>` traces which retrieval layer surfaces (or misses) a page.
 
-**Self-wiring knowledge graph.** Every `put_page` extracts entity refs from markdown/wikilinks/typed-link syntax and writes edges with zero LLM calls. Typed edges (`attended`, `works_at`, `invested_in`, `founded`, `advises`, `mentions`, …). Multi-hop traversal via `gbrain graph-query`. The graph is what produces the +31.4 P@5 lift over vector-only RAG. **Obsidian-style vaults:** bare `[[note-name]]` wikilinks that point across folders — you wrote `[[struktura]]` but the page lives at `projects/struktura.md` — resolve by basename once you opt in with `gbrain config set link_resolution.global_basename true`. Off by default; `gbrain doctor` tells you how many edges you'd gain before you flip it. See [migrating an Obsidian vault](INSTALL_FOR_AGENTS.md#step-45-wire-the-knowledge-graph). **Multi-source brains:** a wikilink whose target page lives in a *different source* creates an edge once you opt in with `gbrain config set link_resolution.cross_source true` (then run `gbrain extract links --source db`); with it off (the default), the extract summary counts the skipped candidates — cross-source links never vanish silently.
+**Self-wiring knowledge graph.** Every `put_page` extracts entity refs from markdown/wikilinks/typed-link syntax and writes edges with zero LLM calls. Typed edges (`attended`, `works_at`, `invested_in`, `founded`, `advises`, `mentions`, …). Multi-hop traversal via `gbrain graph-query`. The graph is what produces the +31.4 P@5 lift over vector-only RAG. **Obsidian-style vaults:** bare `[[note-name]]` wikilinks that point across folders — you wrote `[[struktura]]` but the page lives at `projects/struktura.md` — resolve by basename once you opt in with `gbrain config set link_resolution.global_basename true`. Off by default; `gbrain doctor` tells you how many edges you'd gain before you flip it. See [migrating an Obsidian vault](INSTALL_FOR_AGENTS.md#step-45-wire-the-knowledge-graph). **Multi-source brains:** a wikilink whose target page lives in a *different source* creates an edge once you opt in with `gbrain config set link_resolution.cross_source true` (then run `gbrain extract links --source db`); with it off (the default), the DB extract paths (`extract links --source db`, `extract --stale`) count the skipped candidates in their summaries instead of dropping them silently.
 
 **Job queue (Minions).** BullMQ-shaped, Postgres-native job queue. Durable subagents (LLM tool loops that survive crashes via two-phase pending→done persistence), shell jobs with audit, child jobs with cascading timeouts, rate leases for outbound providers, attachments via S3/Supabase storage. Replaces "spawn subagent as fire-and-forget Promise" with something that recovers from anything.
 

docs/architecture/brains-and-sources.md

@@ -248,7 +248,10 @@ fail-closed and tested:
 - **Takes holders** — per-token allow-lists (`gbrain auth permissions <token>
   set-takes-holders ...`) scope which held takes a remote caller sees.
 - **Write-side slug fences** — a client bound to slug prefixes can only write
-  under them, and non-read operations are refused for slug-bound clients.
+  under them, and only a small fenced allow-list of write operations
+  (`put_page`, `add_link`, `add_timeline_entry`, …) is available to
+  slug-bound clients; every other non-read operation is refused (fail-closed:
+  a write op added later is denied until it is fenced and allow-listed).
 
 One known soft edge: the backlink-count ranking boost counts referrers
 without source filtering, so the *existence* of out-of-grant referrers can

llms-full.txt

@@ -1778,7 +1778,7 @@ The whole loop is described in [`docs/architecture/topologies.md`](docs/architec
 
 **Hybrid search.** Vector (HNSW on pgvector) + BM25 keyword + reciprocal-rank fusion + source-tier boost + intent-aware query rewriting. Three named search modes (`conservative`, `balanced`, `tokenmax`) bundle the cost/quality knobs into a single config key. Live cost/recall comparisons in [`docs/eval/SEARCH_MODE_METHODOLOGY.md`](docs/eval/SEARCH_MODE_METHODOLOGY.md). Default: `balanced` with ZeroEntropy reranker on. Per-query graph signals notice when a top result is a hub for THAT query (adjacency boost), is corroborated across team brains (cross-source boost), or is being crowded out by weak chunks from a chatty session (session demote). Run `gbrain search "<query>" --explain` to see per-stage attribution: base score, every boost that fired, what it multiplied. `gbrain doctor` ships a `graph_signals_coverage` check; `gbrain search stats` shows fire counts and failure breakdowns. Vector retrieval pools the best chunk per page, so a page surfaces on its strongest evidence instead of losing to a neighbor on one weak chunk. Queries that match a page's title phrase or a declared free-text alias (`gbrain reindex --aliases` backfills existing pages) get boosted to the page they name. Every result carries an `evidence` tag (why it matched) and a `create_safety` hint (`exists` / `probable` / `unknown`) so an agent decides whether a page already exists instead of guessing from a raw score. `gbrain search diagnose "<query>" --target <slug>` traces which retrieval layer surfaces (or misses) a page.
 
-**Self-wiring knowledge graph.** Every `put_page` extracts entity refs from markdown/wikilinks/typed-link syntax and writes edges with zero LLM calls. Typed edges (`attended`, `works_at`, `invested_in`, `founded`, `advises`, `mentions`, …). Multi-hop traversal via `gbrain graph-query`. The graph is what produces the +31.4 P@5 lift over vector-only RAG. **Obsidian-style vaults:** bare `[[note-name]]` wikilinks that point across folders — you wrote `[[struktura]]` but the page lives at `projects/struktura.md` — resolve by basename once you opt in with `gbrain config set link_resolution.global_basename true`. Off by default; `gbrain doctor` tells you how many edges you'd gain before you flip it. See [migrating an Obsidian vault](INSTALL_FOR_AGENTS.md#step-45-wire-the-knowledge-graph). **Multi-source brains:** a wikilink whose target page lives in a *different source* creates an edge once you opt in with `gbrain config set link_resolution.cross_source true` (then run `gbrain extract links --source db`); with it off (the default), the extract summary counts the skipped candidates — cross-source links never vanish silently.
+**Self-wiring knowledge graph.** Every `put_page` extracts entity refs from markdown/wikilinks/typed-link syntax and writes edges with zero LLM calls. Typed edges (`attended`, `works_at`, `invested_in`, `founded`, `advises`, `mentions`, …). Multi-hop traversal via `gbrain graph-query`. The graph is what produces the +31.4 P@5 lift over vector-only RAG. **Obsidian-style vaults:** bare `[[note-name]]` wikilinks that point across folders — you wrote `[[struktura]]` but the page lives at `projects/struktura.md` — resolve by basename once you opt in with `gbrain config set link_resolution.global_basename true`. Off by default; `gbrain doctor` tells you how many edges you'd gain before you flip it. See [migrating an Obsidian vault](INSTALL_FOR_AGENTS.md#step-45-wire-the-knowledge-graph). **Multi-source brains:** a wikilink whose target page lives in a *different source* creates an edge once you opt in with `gbrain config set link_resolution.cross_source true` (then run `gbrain extract links --source db`); with it off (the default), the DB extract paths (`extract links --source db`, `extract --stale`) count the skipped candidates in their summaries instead of dropping them silently.
 
 **Job queue (Minions).** BullMQ-shaped, Postgres-native job queue. Durable subagents (LLM tool loops that survive crashes via two-phase pending→done persistence), shell jobs with audit, child jobs with cascading timeouts, rate leases for outbound providers, attachments via S3/Supabase storage. Replaces "spawn subagent as fire-and-forget Promise" with something that recovers from anything.