feat: OCR scans with tesseract, drop docling from the default path.

pdftotext still wins on born-digital PDFs. Empty text layers go through
pdftoppm + tesseract eng+deu. Optional OCR_ENGINE=paddle. Gitea #6.

.dockerignore

@@ -3,6 +3,7 @@
 var
 .git
 .github
+lib-ladybug
 __pycache__
 *.pyc
 *.lbug

.github/workflows/ci.yml

@@ -37,6 +37,9 @@ jobs:
           bash -n bin/db/ssh-tunnel
           bash -n bin/docker-entrypoint
           bash -n bin/kb/search
+          bash -n bin/cgo/zig
+          sh -n bin/cgo/zcc
+          sh -n bin/cgo/zc++
 
       - name: Python unit tests (offline, vendored tools)
         run: |
@@ -51,17 +54,44 @@ jobs:
         run: go test ./internal/brain/rank -count=1
 
       - name: facts/audit self (lexicon consistency, no network)
-        run: |
+        run: ./bin/facts/audit self
-          ./bin/facts/audit self 2>/dev/null || echo "audit: not yet implemented; gate skipped"
 
-      - name: kb/eval recall gate
+      - name: CGO via Zig (compile brain/search + eval)
         run: |
-          ./bin/kb/eval 2>/dev/null || echo "eval: not yet implemented; gate skipped"
+          chmod +x bin/cgo/zig bin/cgo/zcc bin/cgo/zc++
+          bin/cgo/zig go build -tags system_ladybug -o /tmp/brain-search ./bin/brain/search.go
+          bin/cgo/zig go build -tags 'system_ladybug,brain_eval' -o /tmp/brain-eval ./bin/brain/eval.go
+
+      - uses: actions/cache@v4
+        with:
+          path: ~/.cache/huggingface
+          key: ${{ runner.os }}-hf-potion-multilingual-128M
+
+      - name: recall@5 SoT (Zig bin/brain/eval.go)
+        run: |
+          uv run python bin/kb/index --rebuild --json
+          KB_ROOT="$PWD" /tmp/brain-eval --json
+
+  ocr:
+    name: OCR (tesseract fixture)
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+      - uses: actions/setup-go@v5
+        with:
+          go-version-file: go.mod
+      - name: Install tesseract + poppler
+        run: |
+          sudo apt-get update
+          sudo apt-get install -y --no-install-recommends \
+            tesseract-ocr tesseract-ocr-eng tesseract-ocr-deu poppler-utils
+      - name: Go OCR tests (synthetic HELLO PNG)
+        run: go test ./internal/ocr -count=1
 
   release:
     name: Release (semver)
     if: github.event_name == 'push' && github.ref == 'refs/heads/main'
-    needs: test
+    needs: [test, ocr]
     runs-on: ubuntu-latest
     permissions:
       contents: write

.gitignore

@@ -11,3 +11,4 @@ __pycache__/
 .secrets/
 lib-ladybug/
 go.work.local
+models/

AGENTS.md

@@ -3,7 +3,8 @@
 Evidence-first brain over the ops/eSlider stack. Facts need proof or they are
 `(not confirmed)`.
 
-Read first: [PLAN](PLAN.md) → [docs](docs/).
+Read first: [PLAN](PLAN.md) → [docs](docs/) → [roadmap](docs/roadmap.md)
+(epic [#16](https://git.produktor.io/eSlider/2dph/issues/16)).
 
 ## Method (detective, no fork)
 
@@ -15,6 +16,9 @@ Read first: [PLAN](PLAN.md) → [docs](docs/).
 - `info` root = descriptive/narrative leafs, searchable, never asserted as fact.
 - Search is deduction: `facts` → `info` → `web-search` (second independent
   source). An answer is `confirmed` only if it comes off the facts root.
+- Fact-check every *claim* (facts → info → live → web), not every edit or
+  syntax tweak. PicoClaw: `search` then `get` then `audit` before a factual
+  reply (`skills/picoclaw/SKILL.md`). `throttled` is not a negative finding.
 
 ## Hard rules
 
@@ -36,19 +40,21 @@ PLAN.md       decisions + execution + open questions
 docs/         published docs
 skills/       in-project agent skills (vendored, no external links)
 bin/          self-describing tools bin/{subject}/{method}.go (shebang)
-bin/brain/    search.go serve.go index.go get.go stats.go eval.go watch.go
+bin/brain/    search.go serve.go index.go add.go get.go stats.go eval.go watch.go
 bin/chats/    sync.go import.go facts.go apply.go; libs in internal/chats
-bin/mail/     sync.go import.go (index_mail → brain/index.go)
+bin/mail/     sync.go import.go ocr.go (index_mail → brain/index.go)
-bin/markdown/ import.go (mistune leafs)
+bin/markdown/ import.go (H2 leaf split; Python bin/md/import fallback)
 bin/postgres/ query.go (read-only YAML)
 bin/git/      import.go (go-git history; Python shim execs it)
 bin/web/      search.go (SearXNG; Python shim execs it)
-internal/     shared Go (brain/rank is cgo-free; chats parsers; gitlog; websearch)
+bin/reasoner/ bakeoff.go (D18 CPU OpenAI tool-call bake-off)
+internal/     shared Go (brain/rank is cgo-free; chats parsers; gitlog; websearch; reasoner)
 bin/watch/    corpus watcher (used by bin/brain/watch.go)
 bin/tools/    vendored python libs behind bin/* (kblib, yamlout, websearch)
-bin/docker-entrypoint  container entrypoint (brain index|search|serve|watch)
+bin/cgo/      zig zcc zc++ (CGO via zig cc, not gcc)
+bin/docker-entrypoint  container entrypoint (api: serve|search|watch; index: python)
 compose.yaml  docker composition (root level, not docker/)
-Dockerfile    multi-stage: python deps + static Go binaries
+Dockerfile    api (Zig CGO, no Python) + index (Python write)
 var/          kb.lbug, var/mail/*, caches (gitignored)
 .venv/        ladybug + model2vec + mistune
 ```
@@ -59,18 +65,18 @@ var/          kb.lbug, var/mail/*, caches (gitignored)
 bin/mail/sync.go --source onlyoffice,gmail --workers 8 --out var/mail   # raw message.json + attachments
 bin/mail/sync.go --source gmail --query 'from:example.com' --out var/mail  # Gmail search (default in:inbox)
 bin/mail/import.go --from-raw var/mail                                  # message.json → message.md (convert only)
-bin/brain/index.go --rebuild                                            # rebuild brain incl. all mail (fresh DB)
+bin/brain/index.go --rebuild --with-facts --with-chats
 ```
 
 - `sync` (Go) downloads messages + attachments; Gmail uses paginated list +
   `body.attachmentId` (not partId) for attachments.
 - `import` converts body + attachments to markdown. PDFs use poppler
-  `pdftotext -layout` fast path (~15ms); textless/scanned PDFs fall back to
+  `pdftotext -layout` fast path (~15ms); textless/scanned PDFs use
-  docling (isolated subprocess — its native onnx can segfault the parent).
+  `pdftoppm` + tesseract `eng+deu` (`bin/mail/ocr.go`). Optional
-  Conversion never touches the brain DB (crash safety).
+  `OCR_ENGINE=paddle`. Conversion never touches the brain DB (crash safety).
-- `index_mail` is a deprecation shim for `bin/brain/index.go --rebuild`. Ladybug
+- `index_mail` is a deprecation shim for `bin/brain/index.go --rebuild`. Bulk
-  corrupts its WAL when brand-new leafs are bulk-inserted while FTS/vector
+  rebuild still deletes `var/kb.lbug` and creates FTS/HNSW last. Single-leaf
-  indexes exist; a fresh DB with indexes created last is the only safe path.
+  write is `bin/brain/add.go` (safe while indexes exist; do not DROP INDEX).
   Keep conversion + indexing separate so a conversion crash can't leave the
   DB mid-transaction.
 
@@ -82,14 +88,20 @@ bin/facts/crm.go [--dry-run]                       # proof person↔company/comp
 bin/kb/search "query" [--repo X]                  # deprecated wrapper → bin/brain/search.go
 bin/brain/search.go "query" [--root facts|info]   # deduction search → YAML
 bin/brain/search.go "query" --no-web              # local graph only
+eval "$(bin/cgo/zig env)"                         # Zig cc + liblbug (not gcc)
+bin/brain/index.go --rebuild [--with-mail] [--with-facts] [--with-chats]
+bin/brain/add.go --text T --root facts --source "a.md x b.md"  # incremental write
+bin/brain/add.go --json                                      # stdin leaf or {leafs:[...]}
 bin/brain/get.go <id> [--body] [--json]          # Go read; Python bin/kb/get CI fallback
 bin/brain/stats.go [--json]
 bin/brain/eval.go [--json]                       # recall@5; questions in internal/brain/rank
 bin/brain/serve.go                               # HTTP :8630; GET /openapi.json POST /mcp
-bin/markdown/import.go [dir]                      # mistune leaves → YAML
+bin/markdown/import.go [dir]                      # H2 leafs → YAML; Python bin/md/import fallback
 bin/git/import.go [REPO] [--json] [--limit N]     # go-git history → commit leafs
 bin/web/search.go "query" [--json]                # SearXNG; throttled ≠ absence
+bin/reasoner/bakeoff.go [--model ID] [--json]     # D18 CPU tool-call bake-off
 bin/postgres/query.go --profile onlyoffice -c 'SELECT 1'
+bin/mail/ocr.go <image|pdf>                      # tesseract eng+deu (scans)
 bin/md/tables                                     # what the graph holds → YAML
 bin/brain/deduce "question"                       # thinking wrapper
 ```

Dockerfile

@@ -1,5 +1,13 @@
 # syntax=docker/dockerfile:1
-FROM python:3.12-slim AS base
+#
+#   docker build --target api -t 2dph:api .
+#   docker build --target index -t 2dph:index .
+#
+# API: Go + ladybug via Zig CGO (no CPython).
+# Index: Python write path (profile `index` until brain/add is v2).
+
+# --- Python sidecar (Ladybug write / rebuild) ---
+FROM python:3.12-slim AS index
 
 ENV PYTHONUNBUFFERED=1 \
     PYTHONDONTWRITEBYTECODE=1 \
@@ -8,26 +16,16 @@ ENV PYTHONUNBUFFERED=1 \
 
 WORKDIR /app
 RUN id -u 2dph 2>/dev/null || useradd --create-home --uid 1001 2dph
+RUN apt-get update \
+    && apt-get install -y --no-install-recommends \
+         poppler-utils tesseract-ocr tesseract-ocr-eng tesseract-ocr-deu \
+    && rm -rf /var/lib/apt/lists/*
 
-# deps layer-first: rebuild only on dependency change
 COPY requirements.lock.txt /tmp/requirements.lock.txt
 RUN python -m pip install --no-cache-dir -r /tmp/requirements.lock.txt \
     && rm /tmp/requirements.lock.txt
 
-# Go services: static binaries, no interpreter at runtime
-FROM golang:1.25 AS go-build
-WORKDIR /src
-COPY go.mod ./
-COPY bin/server ./bin/server
-COPY bin/watch ./bin/watch
-RUN CGO_ENABLED=0 go build -o /serve ./bin/server \
-    && CGO_ENABLED=0 go build -o /watch ./bin/watch
-
-# runtime: python toolchain + Go services
-FROM base
 COPY . .
-COPY --from=go-build /serve /app/bin/serve
-COPY --from=go-build /watch /app/bin/watch
 RUN chmod +x /app/bin/docker-entrypoint \
     && chown -R 2dph:2dph /app
 USER 2dph
@@ -37,5 +35,51 @@ ENV PATH="/app/bin:${PATH}" \
     KB_ROOT=/app
 HEALTHCHECK --interval=30s --timeout=5s --start-period=10s --retries=3 \
     CMD python -c "import model2vec, ladybug, mistune; print('ok')" || exit 1
-
 ENTRYPOINT ["/app/bin/docker-entrypoint"]
+
+# --- Go API: CGO with Zig, not gcc ---
+FROM golang:1.26-bookworm AS api-build
+WORKDIR /src
+RUN apt-get update \
+    && apt-get install -y --no-install-recommends curl xz-utils ca-certificates \
+    && rm -rf /var/lib/apt/lists/*
+
+COPY bin/cgo ./bin/cgo
+RUN chmod +x bin/cgo/zig bin/cgo/zcc bin/cgo/zc++ \
+    && ./bin/cgo/zig env >/dev/null
+
+COPY go.mod go.sum ./
+RUN go mod download
+
+COPY . .
+ENV CGO_RPATH=/usr/local/lib
+RUN eval "$(./bin/cgo/zig env)" \
+    && go build -tags brain_serve,system_ladybug -o /out/brain-serve ./bin/brain/serve.go \
+    && go build -tags system_ladybug -o /out/brain-search ./bin/brain/search.go \
+    && CGO_ENABLED=0 go build -tags brain_watch -o /out/brain-watch ./bin/brain/watch.go
+
+FROM debian:bookworm-slim AS api
+RUN apt-get update \
+    && apt-get install -y --no-install-recommends libssl3 ca-certificates wget \
+    && rm -rf /var/lib/apt/lists/* \
+    && useradd --create-home --uid 1001 2dph
+COPY --from=api-build /out/brain-serve /usr/local/bin/brain-serve
+COPY --from=api-build /out/brain-search /usr/local/bin/brain-search
+COPY --from=api-build /out/brain-watch /usr/local/bin/brain-watch
+COPY --from=api-build /src/lib-ladybug/liblbug.so.0.19.1 /usr/local/lib/liblbug.so.0.19.1
+COPY bin/docker-entrypoint /usr/local/bin/docker-entrypoint
+RUN chmod +x /usr/local/bin/docker-entrypoint \
+    && ln -s liblbug.so.0.19.1 /usr/local/lib/liblbug.so.0 \
+    && ln -s liblbug.so.0 /usr/local/lib/liblbug.so \
+    && ldconfig
+USER 2dph
+ENV KB_ROOT=/data \
+    KB_PORT=8630 \
+    LD_LIBRARY_PATH=/usr/local/lib \
+    HF_HOME=/data/hf
+WORKDIR /data
+EXPOSE 8630
+HEALTHCHECK --interval=30s --timeout=5s --start-period=10s --retries=3 \
+    CMD wget -qO- http://127.0.0.1:8630/health || exit 1
+ENTRYPOINT ["/usr/local/bin/docker-entrypoint"]
+CMD ["serve"]

PLAN.md

@@ -4,7 +4,10 @@ A brain that loves facts and deduction. Evidence-first knowledge graph + hybrid
 RAG over the operational Brain/ops/eSlider stack. Built like Sherlock
 Holmes: nothing is asserted unless it has proof.
 
-Status: **in progress** — this file is the plan and the record of decisions.
+Status: **v1 in** (epic [#16](https://git.produktor.io/eSlider/2dph/issues/16) closed).
+v2 board: milestone [v2](https://git.produktor.io/eSlider/2dph/milestone/13) — OCR [#6](https://git.produktor.io/eSlider/2dph/issues/6),
+[#29](https://git.produktor.io/eSlider/2dph/issues/29) OQ1, [#30](https://git.produktor.io/eSlider/2dph/issues/30) OQ3.
+Gap: [docs/roadmap.md](docs/roadmap.md).
 
 ## What
 
@@ -29,9 +32,9 @@ detective method: **a fact needs ≥2 independent sources or it is
 | D3 | web search | Go client `bin/web/search.go` (`internal/websearch`). SearXNG URL is config (`BRAIN_SEARCH_URL`). Optional Compose profile `searxng` (sanitized settings). Do not run a second copy on a host that already has one. Empty/`throttled` ≠ “nothing exists”. |
 | D4 | embeddings | **model2vec** `minishlab/potion-multilingual-128M` instead of embeddinggemma. |
 | D5 | parser | **mistune** for MD → leaf extraction (duckdb-md documented as future optional SQL/export layer, not v1). |
-| D6 | graph engine | **LadybugDB**. Go is the service (`bin/brain/search.go`, `bin/brain/serve.go` in-process, `internal/brain`). Read path (`get.go` / `stats.go` / `eval.go`) is Go + cgo. Python `bin/kb/{get,stats,eval}` is the CI fallback (GitHub runners have no ladybug cgo). Index/write stays Python until the Go write path is safe. |
+| D6 | graph engine | **LadybugDB**. Go is the service (`bin/brain/search.go`, `bin/brain/serve.go` in-process, `internal/brain`). Read path is Go + Zig CGO (D21). Python `bin/kb/{get,stats,eval}` is the CI fallback when Zig/libs are not fetched. Incremental write is Python `bin/kb/add` (`bin/brain/add.go`). Bulk rebuild stays `compose --profile index` until the Go write path is safe. |
 | D7 | db access | `db-yaml`/`psql-yq`-style, read-only, YAML out. OnlyOffice Postgres via SSH tunnel (`127.0.0.1:5433`). |
-| D8 | evidence | detective method: ≥2 independent sources or `(not confirmed)`. Auto-pair docker ps × compose × ssh-config × docs. |
+| D8 | evidence | detective method: ≥2 independent sources or `(not confirmed)`. 2-source auto-pair docker ps × compose × ssh-config × docs. |
 | D9 | facts/goal model | Who / What / How / Where / When + evidence + confidence on every edge. |
 | D10 | versioning | everything is a leaf with `sha256 + observed_at + source_rev`; `File-[:HAS_VERSION]->Commit-[:AUTHORED]->Person`. Stale = `source_rev` < git HEAD. |
 | D11 | strong/weak | `root` column: `facts` (strong) vs `info` (weak). Answer is `confirmed` only from facts root. |
@@ -41,9 +44,10 @@ detective method: **a fact needs ≥2 independent sources or it is
 | D15 | repo | Gitea [`eSlider/2dph`](https://git.produktor.io/eSlider/2dph) is origin + [issues](https://git.produktor.io/eSlider/2dph/issues). GitHub `eSlider/2dph` is the public clone (PRs + Actions CI). No direct `main` pushes. TDD → PR → CI green → merge. |
 | D16 | contradictions | ≥2 yes vs ≥2 no → unrelated sources conflict → hypothesis → `(not confirmed)`. Resolution (authority, staleness adjudication) = **v2**, tracked as open question. |
 | D17 | assertion gate | Fact-check every *claim* (facts → info → live → web), not every edit. `bin/brain/search.go` adds a `web` block when there is no facts hit (`throttled`/`skipped`/`refused` ≠ absence). `--root` and `--no-web` stay local. Missing graph ≠ “does not exist”. |
-| D18 | reasoner | Pluggable OpenAI-compatible URL. RAM: Qwen3.5-9B. Quality: Bonsai-27B or Qwen3.6-27B. No official Qwen3.6-9B. |
+| D18 | reasoner | Pluggable OpenAI-compatible URL (`REASONER_BASE_URL`). RAM: `Qwen/Qwen3.5-9B`. Quality: `prism-ml/Bonsai-27B-gguf` or `Qwen/Qwen3.6-27B`. No official Qwen3.6-9B. CPU bake-off: `bin/reasoner/bakeoff.go` + compose profile `reasoner` (`OLLAMA_NUM_GPU=0`, `:11435`). PicoClaw is compose profile `picoclaw`; tools are `search`/`get`/`audit`. Weights are not copied into the 2dph image. Agent lever/loop: [#15](https://git.produktor.io/eSlider/2dph/issues/15). |
 | D19 | git history | [go-git](https://github.com/go-git/go-git) via `bin/git/import.go`. No subprocess of the git binary. Conversion prints commit leafs; brain write is `bin/brain/index.go`. |
-| D20 | agent API | OpenAPI + MCP are generated from the same `internal/httpapi.Ops` table as `bin/brain/serve.go` handlers. `GET /openapi.json`, `POST /mcp` (JSON-RPC tools/list + tools/call). Tool names match OpenAPI paths (`search`/`get`/`stats`/`audit`). |
+| D20 | agent API | OpenAPI + MCP are generated from the same `internal/httpapi.Ops` table as `bin/brain/serve.go` handlers. `GET /openapi.json`, `POST /mcp` (JSON-RPC tools/list + tools/call). Tool names match OpenAPI paths (`search`/`get`/`stats`/`audit`/`ingest`). |
+| D21 | CGO | Ladybug/tokenizers CGO is compiled with **Zig** (`bin/cgo/zcc` → `zig cc -target …-linux-gnu`), not gcc. `bin/cgo/zig` pins Zig 0.14.1 + liblbug 0.19.1 + libtokenizers 1.27.0. Compose `target: api` has no CPython; write/rebuild is profile `index`. |
 
 ## Architecture
 
@@ -51,22 +55,27 @@ detective method: **a fact needs ≥2 independent sources or it is
 2dph/
   PLAN.md / AGENTS.md
   docs/                     published docs (this conversation → docs/ as md)
-  skills/                   in-project skills (web-search, db-yaml, brain, diataxis-docs)
+  skills/                   in-project skills (web-search, postgres, brain, picoclaw, diataxis-docs)
   bin/
     facts/extract.go audit.go crm.go  # D14 shebang; Python implementation
-    kb/index                Python write path (called by bin/brain/index.go)
+    kb/index                Python bulk write (called by bin/brain/index.go)
+    kb/add                  Python incremental write (called by bin/brain/add.go)
     brain/index.go          rebuild FTS + HNSW (incl. --with-mail)
+    brain/add.go            incremental leaf write (no rebuild)
     brain/get.go stats.go eval.go  # Go read (cgo); Python bin/kb/* CI fallback
     brain/watch.go
     brain/search.go         deduction: facts → info → web-search
-    brain/serve.go          HTTP API in-process + OpenAPI/MCP (D20)
+    brain/serve.go          HTTP API in-process + OpenAPI/MCP (D20); Zig CGO (D21)
+    cgo/zig zcc zc++        CGO toolchain (zig cc, not gcc)
     mail/import.go          JSON → markdown (no brain write)
-    markdown/import.go      mistune leaves
+    markdown/import.go      H2 leaf split (Go); Python bin/md/import fallback
     postgres/query.go       read-only YAML (wraps bin/db/psql-yq)
     git/import.go           go-git history (no git binary; conversion only)
     web/search.go           SearXNG client (throttled ≠ absence)
+    reasoner/bakeoff.go     CPU tool-call bake-off (D18; OpenAI tools)
     chats/sync.go import.go facts.go apply.go
                             (libs in internal/chats; no chats index)
+    mail/ocr.go             tesseract eng+deu (pdftoppm scans)
     md/import               (deprecated; bin/markdown/import.go)
     brain/extract  brain/audit   brain/deduce    (thinking wrapper)
     web/search               (deprecated shim → web/search.go)
@@ -80,7 +89,10 @@ detective method: **a fact needs ≥2 independent sources or it is
 
 Node tables: `Person, Service, Host, Container, Repo, File, Commit, Leaf`.
 `Leaf(embedding FLOAT[N])` — FTS on `text`, HNSW vector index on `embedding`.
-Edges: `RUNS / USES / HAS_VERSION / AUTHORED / ABOUT / ASSOCIATED / SIMILAR_0.85`.
+Edges: `RUNS / USES / FROM_FILE / HAS_VERSION / AUTHORED / ABOUT / ASSOCIATED / SIMILAR_0.85`.
+`FROM_FILE` / `HAS_VERSION` / `AUTHORED`: `bin/brain/search.go --hop N` walks
+them from each hit (1=File, 2=Commit, 3=Person). Rebuild writes
+`Leaf-[:FROM_FILE]->File`; git import writes the rest.
 
 Common props on every node/edge: `root`, `confidence`, `evidence[]`, `how`,
 `where`, `when`, `source_rev`.
@@ -105,10 +117,13 @@ Common props on every node/edge: `root`, `confidence`, `evidence[]`, `how`,
 ## Open questions (v2)
 
 - OQ1: mutually-contradicting evidence — how to resolve (authority weighting,
-  temporal freshness, audit adjudication).
+  temporal freshness, audit adjudication). **v2**; [#29](https://git.produktor.io/eSlider/2dph/issues/29).
-- OQ2: OCR pipeline for pdfs/images/docs — mostly solved: poppler pdftotext
+- OQ2: OCR — **in**. `pdftotext -layout` first; scans `pdftoppm` + tesseract
-  fast-path for born-digital PDFs, docling fallback for the ~5% textless ones.
+  `eng+deu` (`bin/mail/ocr.go`, `internal/ocr`). No gocv, no gosseract CGO
+  (D21 Zig owns Ladybug CGO). Optional `OCR_ENGINE=paddle` / compose profile
+  `ocr-paddle`. Docling left the default path. [#6](https://git.produktor.io/eSlider/2dph/issues/6).
 - OQ3: optional duckdb-md layer for `SELECT … FORMAT MARKDOWN` export/write-back.
+  [#30](https://git.produktor.io/eSlider/2dph/issues/30).
 - OQ4: YAML-first storage for leafs — deferred: JSON is ~10x faster to
   serialize and unambiguous; YAML only where humans edit files.
 
@@ -117,7 +132,8 @@ Common props on every node/edge: `root`, `confidence`, `evidence[]`, `how`,
 1. `bin/mail/sync.go` (Go, 8 workers) — paginated Gmail/OnlyOffice download.
    Gmail attachments key off `body.attachmentId`, not MIME `partId`.
 2. `bin/mail/import.go --from-raw` — message.json → message.md; PDFs via
-   `pdftotext -layout` (~15ms) with docling subprocess fallback; ICS sidecars
+   `pdftotext -layout` (~15ms); textless/scanned PDFs `pdftoppm` + tesseract
+   `eng+deu`. ICS sidecars
    Latin-1→UTF-8 normalized.
 3. `bin/brain/index.go --rebuild` — fresh rebuild (repo corpus + mail) because ladybug
    corrupts its WAL on bulk-insert into an already-indexed DB. Conversion and
@@ -134,10 +150,9 @@ Common props on every node/edge: `root`, `confidence`, `evidence[]`, `how`,
 2. `go test ./internal/brain/rank` (cgo-free ranking + flag parser)
 3. python -m unittest discover -s bin/tools (includes published-docs SoT)
 4. `bin/facts/audit self` (lexicon internal consistency; `bin/facts/audit.go` is the D14 wrapper)
-5. `bin/kb/eval` (recall@5 ≥ 0.95). Local SoT is `bin/brain/eval.go`; CI uses
+5. `bin/brain/eval.go` via Zig (recall@5 ≥ 0.95). Python `bin/kb/eval` is an
-   the Python twin until the runner has ladybug cgo. Questions live in
+   explicit fallback, not the CI SoT.
-   `internal/brain/rank`.
+6. `bin/cgo/zig go build -tags system_ladybug` (compile search with zig cc; fetches pinned zig+libs).
-6. md-docs build/lint if docs tooling arrives.
 
 Feedback loop: every commit → PR → CI → green/gate → merge. Same discipline as
 `db/tech-poc`: contract first where there is an OpenAPI/message shape.
@@ -146,9 +161,26 @@ Feedback loop: every commit → PR → CI → green/gate → merge. Same discipl
 
 1. scaffold repo (:done after this file + AGENTS.md + .gitignore + ci)
 2. gh repo create eSlider/2dph --private + initial commit + CI
-3. vendored skill integration (web-search, db-yaml, brain, diataxis-docs) — no remote links
+3. vendored skill integration (web-search, postgres, brain, diataxis-docs) — no remote links
 4. .venv: ladybug + model2vec + mistune
 5. schema + tools with TDD (kb + md + facts + brain)
 6. ~/.config/brain config
-7. corpus extraction (facts/info)
+7. corpus extraction (facts/info) — **in**: [#18](https://git.produktor.io/eSlider/2dph/issues/18)
 8. verify: web-search smoke, onlyoffice pg, md-db round-trip, eval, audit
+
+## Gap to v1 (epic #16)
+
+Remaining: none for epic #16 (v1). Board:
+[epic #16](https://git.produktor.io/eSlider/2dph/issues/16),
+milestone [v1 detective brain](https://git.produktor.io/eSlider/2dph/milestone/12).
+Narrative: [docs/roadmap.md](docs/roadmap.md).
+
+| Order | Issue | Gap |
+|-------|-------|-----|
+| 1 | [#14](https://git.produktor.io/eSlider/2dph/issues/14) | **in** — `bin/brain/add.go` / `POST /ingest` write facts+info without deleting `kb.lbug`. Bulk corpus still `--rebuild`. Leftover Python (mail/facts) is not the living-graph blocker. |
+| 2 | [#17](https://git.produktor.io/eSlider/2dph/issues/17) | **in** — `--hop N` walks `FROM_FILE` → `HAS_VERSION` → `AUTHORED` (max 3). |
+| 3 | [#18](https://git.produktor.io/eSlider/2dph/issues/18) | **in** — `--with-facts` / `--facts-json` land `root=facts`; `--with-chats` indexes `var/chats/md`. WhatsApp sync is out of v1. |
+| 4 | [#15](https://git.produktor.io/eSlider/2dph/issues/15) | **in** — lever/loop documented (`search` → `get` → `audit`). |
+| 5 | [#19](https://git.produktor.io/eSlider/2dph/issues/19) | **in** — CI recall SoT is `bin/brain/eval.go` via Zig. Python `bin/kb/eval` stays as an explicit fallback. |
+
+Does **not** block epic close: OQ1 [#29](https://git.produktor.io/eSlider/2dph/issues/29), OQ3 [#30](https://git.produktor.io/eSlider/2dph/issues/30), OQ4. OCR [#6](https://git.produktor.io/eSlider/2dph/issues/6) is **in**.

README.md

@@ -7,14 +7,16 @@
 [![Latest Release](https://img.shields.io/github/v/tag/eSlider/2dph?sort=semver&label=release)](https://github.com/eSlider/2dph/releases)
 [![GitHub Stars](https://img.shields.io/github/stars/eSlider/2dph?style=social)](https://github.com/eSlider/2dph/stargazers)
 
-An evidence-first brain over the operational eSlider stack. **Facts need two
+An evidence-first brain. **Facts need two independent sources, or they are
-independent sources, or they are `(not confirmed)`.**
+`(not confirmed)`.** Cursor is not the runtime.
 
-`2dph` is a single embedded knowledge graph (LadybugDB = Kuzu successor) with
+`2dph` is a single embedded knowledge graph (LadybugDB) with native **HNSW
-native **HNSW vector** + **BM25 full-text** indexes, built from markdown,
+vector** + **BM25 full-text** indexes. Search is *deduction*: confirmed facts
-compose files, ssh config, docker state, and git history. Search is
+first, supporting info second, `web-search` as the independent second source
-*deduction*: confirmed facts first, supporting info second, `web-search` as
+when the local graph cannot confirm.
-the independent second source when the local graph cannot confirm.
+
+Run it: [docs/runbook.md](docs/runbook.md). Design: [docs/design.md](docs/design.md).
+Docs index: [docs/README.md](docs/README.md).
 
 ## Architecture
 
@@ -31,7 +33,7 @@ graph TB
         EX["bin/facts/extract.go<br/>2-source pairing"]
         AU["bin/facts/audit.go<br/>confidence + staleness"]
         IDX["bin/brain/index.go<br/>chunk + embed"]
-        MD["bin/markdown/import.go<br/>mistune leaves"]
+        MD["bin/markdown/import.go<br/>H2 leaf split"]
         SR["bin/brain/search.go<br/>deduction"]
     end
 
@@ -74,7 +76,7 @@ graph TB
 ## The method
 
 Every assertion is `Who / What / How / Where / When + evidence + confidence`,
-mirroring the detective detective skill: **≥2 independent sources confirm a
+mirroring the detective method: **≥2 independent sources confirm a
 fact; conflicting sources or a single source → `hypothesis` → `(not confirmed)`.**
 
 | root | meaning | used for answers |
@@ -94,7 +96,7 @@ bin/brain/stats.go                                   # index health
 bin/brain/eval.go                                    # recall@5 gate
 ```
 
-`--hop` is not implemented (needs File/FROM_FILE edges); the flag errors instead of walking. `bin/kb/search` is a deprecated wrapper around `bin/brain/search.go`.
+`--hop N` walks File/Commit/Person from each hit (max 3). `bin/kb/search` is a deprecated wrapper around `bin/brain/search.go`.
 
 Git history is read with [go-git](https://github.com/go-git/go-git) (no git binary):
 
@@ -118,23 +120,24 @@ Mail is a first-class corpus (retrievable through the same search):
 ```bash
 bin/mail/sync.go --source onlyoffice,gmail --workers 8 --out var/mail  # raw sync (Go)
 bin/mail/import.go --from-raw var/mail                                  # JSON → markdown
+bin/brain/add.go --text T --root facts --source "a.md x b.md"
+bin/brain/index.go --rebuild --with-facts --with-chats                  # facts extract + chats md
 bin/brain/index.go --rebuild                                            # rebuild brain (incl. mail)
 bin/brain/search.go "invoice from last week"                            # same search over mail leafs
 ```
 
 ## Storage
 
-- **LadybugDB** — single `var/kb.lbug`, Cypher property graph, HNSW + BM25
+- **LadybugDB** — single `var/kb.lbug`, Cypher + HNSW + BM25, embedded.
-  in one engine, embedded (no server), ACID, read-only-safe for concurrent
+  Read tools (`get` / `stats` / `eval`) are Go + Zig CGO (`bin/cgo/zcc`).
-  readers. Read tools (`get` / `stats` / `eval`) are Go + cgo; Python
+  Python fallbacks stay for CI until the runner fetches Zig. Incremental
-  `bin/kb/{get,stats,eval}` is the CI fallback. **Never `DROP INDEX` FTS/VECTOR** on Ladybug 0.19: DROP leaves
+  write is `bin/brain/add.go` (Python `kblib.add_leafs`). Bulk rebuild is
-  ghost catalog tables (`_0_Leaf_vec_UPPER`) so recreate fails while
+  Compose profile `index` (`bin/brain/index.go --rebuild`).
-  `SHOW_INDEXES` omits HNSW. Fresh indexes = delete `var/kb.lbug` +
+- **model2vec** — `potion-multilingual-128M` (256-dim), CPU, no Ollama
-  `bin/brain/index.go --rebuild`. Use `ensure_indexes()` after upserts.
+  runtime dependency.
-- **model2vec** — `potion-multilingual-128M` static embeddings (256-dim),
+- facts and info split by `root` but written in the same transaction.
-  CPU-fast, deterministic, no Ollama runtime dependency.
+
-- facts and info split semantically by `root` column but written inside the
+Ladybug 0.19 DROP INDEX warning: [docs/runbook.md](docs/runbook.md).
-  same transaction.
 
 ## Tooling conventions
 
@@ -145,31 +148,39 @@ machines. Tests gate every commit. HTTP: `bin/brain/serve.go` calls
 
 ## Development
 
+See the portable runbook: [docs/runbook.md](docs/runbook.md).
+
 ```bash
-uv venv .venv                                  # Python 3.12, uv-managed
+uv venv .venv
-uv pip install -r requirements.lock.txt        # pinned toolchain
+uv pip install -r requirements.lock.txt
-bin/facts/audit.go self                        # lexicon consistency gate
+bin/facts/audit.go self
-go test ./... && python -m unittest discover -s bin/tools -t .
+go test ./... && uv run python -m unittest discover -s bin/tools -t .
 ```
 
 Docker (optional, cached model + var volumes):
 
 ```bash
-docker compose run --rm brain index            # (re)index corpus
+docker compose up -d brain                     # API (Zig CGO serve :8630)
-docker compose run --rm brain search "query"   # one-shot query
+docker compose --profile index run --rm index  # Python Ladybug rebuild
-docker compose run --rm brain serve            # bin/brain/serve.go
+docker compose --profile picoclaw up brain-mcp # MCP on 127.0.0.1:8630
+docker compose --profile reasoner up -d reasoner  # CPU Ollama 127.0.0.1:11435
 docker compose up brain-watch                  # auto re-index on change
 ```
 
 ## Related
 
+eSlider DevOps engineer practice: ops, OnlyOffice, and mail feed the facts
+root through `bin/facts/extract` (two-source pairing).
+
 - [go-second-brain](https://github.com/eSlider/go-second-brain) — the earlier
   Neo4j + Qdrant + Matrix RAG brain
 - [agent-skills](https://github.com/eSlider/agent-skills) — upstream
-  skills (`web-search`, `db-yaml`, …) that 2dph integrates
+  skills (`web-search`, `postgres`, …) that 2dph integrates
 - detective method — the two-source method
 
-Work board (issues): [git.produktor.io/eSlider/2dph/issues](https://git.produktor.io/eSlider/2dph/issues).
+Work board (issues): [epic #16](https://git.produktor.io/eSlider/2dph/issues/16)
+on [git.produktor.io/eSlider/2dph/issues](https://git.produktor.io/eSlider/2dph/issues).
 PRs and CI: GitHub [`eSlider/2dph`](https://github.com/eSlider/2dph).
 
-See [PLAN.md](PLAN.md) for decisions, execution status, and v2 open questions.
+See [PLAN.md](PLAN.md) for decisions, [docs/roadmap.md](docs/roadmap.md) for
+the gap to v1, and v2 open questions.

bin/brain/add.go

@@ -0,0 +1,21 @@
+//usr/bin/env go run -tags=brain_add "$0" "$@"; exit
+//go:build brain_add
+//
+// bin/brain/add.go - incremental leaf write (Python kblib, no rebuild).
+//
+//	./bin/brain/add.go --text T --root facts --source "a.md x b.md"
+//	./bin/brain/add.go --json
+//
+// D6: write stays Python. Does not delete var/kb.lbug.
+// NOTE: never run `gofmt -w` on this file — it breaks the shebang.
+package main
+
+import (
+	"os"
+
+	"github.com/eSlider/2dph/internal/cmdbin"
+)
+
+func main() {
+	os.Exit(cmdbin.ExecFile("bin/kb/add", os.Args[1:]))
+}

bin/brain/get.go

@@ -8,6 +8,7 @@
 //	./bin/brain/get.go <id> --json
 //
 // Needs CGO + libladybug. Python bin/kb/get is the CI fallback (no cgo).
+// CGO compiler is Zig (`eval "$(bin/cgo/zig env)"`), not gcc.
 // NOTE: never run `gofmt -w` on this file — it breaks the shebang.
 package main
 

bin/brain/index.go

@@ -3,12 +3,12 @@
 //
 // bin/brain/index.go - rebuild the Ladybug graph (Python write path).
 //
-//	./bin/brain/index.go --rebuild
+//	./bin/brain/index.go --rebuild --with-facts --with-chats
 //	./bin/brain/index.go --rebuild --with-mail
 //	./bin/brain/index.go --dry-run --with-mail
 //
-// v1 write is always a rebuild when mail is included (live FTS/HNSW + bulk
+// v1 write: bin/brain/add.go for one/few leafs (indexes may already exist).
-// insert corrupts Ladybug 0.19 WAL). `add` is v2.
+// Bulk mail/corpus still --rebuild (fresh file, indexes last).
 // NOTE: never run `gofmt -w` on this file — it breaks the shebang.
 package main
 

bin/brain/search.go

@@ -3,11 +3,11 @@
 //
 // bin/brain/search.go - deduction search over the 2dph brain.
 //
-//	./bin/brain/search.go "query" [--root facts|info] [--repo P] [-n N] [--json] [--no-web]
+//	./bin/brain/search.go "query" [--root facts|info] [--repo P] [-n N] [--hop N] [--json] [--no-web]
 //	./bin/brain/search.go serve [port]
 //	./bin/brain/search.go --list-model
 //
-// Needs CGO + libladybug (CGO_CFLAGS/CGO_LDFLAGS). Prefer the wrapper
+// Needs CGO + libladybug via Zig (`eval "$(bin/cgo/zig env)"`), not gcc.
 // bin/kb/search which sets those and builds a binary for the embed daemon.
 // NOTE: never run `gofmt -w` on this file — it breaks the shebang.
 package main

bin/cgo/zc++

@@ -0,0 +1,23 @@
+#!/bin/sh
+# bin/cgo/zc++ — CGO CXX. Zig, not g++.
+set -eu
+ROOT="$(CDPATH= cd -- "$(dirname "$0")/../.." && pwd)"
+case "$(uname -m)" in
+  x86_64|amd64) TARGET=x86_64-linux-gnu ;;
+  aarch64|arm64) TARGET=aarch64-linux-gnu ;;
+  *)
+    echo "zc++: unsupported arch $(uname -m)" >&2
+    exit 2
+    ;;
+esac
+if [ -n "${ZIG:-}" ] && [ -x "$ZIG" ]; then
+  :
+elif [ -x "$ROOT/var/zig/zig" ]; then
+  ZIG="$ROOT/var/zig/zig"
+elif command -v zig >/dev/null 2>&1; then
+  ZIG="$(command -v zig)"
+else
+  echo "zc++: zig missing; run bin/cgo/zig first" >&2
+  exit 127
+fi
+exec "$ZIG" c++ -target "$TARGET" "$@"

bin/cgo/zcc

@@ -0,0 +1,24 @@
+#!/bin/sh
+# bin/cgo/zcc — CGO CC. Zig, not gcc.
+# Go invokes CC with many args; a wrapper avoids spaces in $CC.
+set -eu
+ROOT="$(CDPATH= cd -- "$(dirname "$0")/../.." && pwd)"
+case "$(uname -m)" in
+  x86_64|amd64) TARGET=x86_64-linux-gnu ;;
+  aarch64|arm64) TARGET=aarch64-linux-gnu ;;
+  *)
+    echo "zcc: unsupported arch $(uname -m)" >&2
+    exit 2
+    ;;
+esac
+if [ -n "${ZIG:-}" ] && [ -x "$ZIG" ]; then
+  :
+elif [ -x "$ROOT/var/zig/zig" ]; then
+  ZIG="$ROOT/var/zig/zig"
+elif command -v zig >/dev/null 2>&1; then
+  ZIG="$(command -v zig)"
+else
+  echo "zcc: zig missing; run bin/cgo/zig first" >&2
+  exit 127
+fi
+exec "$ZIG" cc -target "$TARGET" "$@"

bin/cgo/zig

@@ -0,0 +1,130 @@
+#!/usr/bin/env bash
+# bin/cgo/zig — CGO toolchain: zig cc (not gcc) + pinned liblbug + libtokenizers.
+#
+#   eval "$(bin/cgo/zig env)"     # export CC/CXX/CGO_*
+#   bin/cgo/zig go build ...      # ensure, then exec with env
+#   bin/cgo/zig ./bin/brain/search.go "query"
+#
+# Pins live in this file. Downloads land in var/ (gitignored).
+set -euo pipefail
+
+ROOT="$(CDPATH= cd -- "$(dirname "$0")/../.." && pwd)"
+ZIG_VERSION=0.14.1
+LBUG_VERSION=0.19.1
+TOKENIZERS_VERSION=1.27.0
+
+arch="$(uname -m)"
+case "$arch" in
+  x86_64|amd64)
+    ZIG_ARCH=x86_64
+    LBUG_ARCH=x86_64
+    TOK_ARCH=x86_64
+    ZIG_SHA=24aeeec8af16c381934a6cd7d95c807a8cb2cf7df9fa40d359aa884195c4716c
+    LBUG_SHA=ed263ae913f68cb0ddba0b98548b58edaac49929766d03bdaaa83be46c68847d
+    TOK_SHA=72556cdca798dd4ea7cdaba308e5f0d68a8cb93b67c96edf485b7a0edd7b07f4
+    ;;
+  aarch64|arm64)
+    ZIG_ARCH=aarch64
+    LBUG_ARCH=aarch64
+    TOK_ARCH=aarch64
+    ZIG_SHA=f7a654acc967864f7a050ddacfaa778c7504a0eca8d2b678839c21eea47c992b
+    LBUG_SHA=b07df2cd533c3976a2a3025866d6420a5f35514d0a822ecc4b2902d55b4725b7
+    TOK_SHA=e96545ad05930c26f51f63d932ee6d3bbd32bbed149e102c5290d587a2293067
+    ;;
+  *)
+    echo "bin/cgo/zig: unsupported arch $arch" >&2
+    exit 2
+    ;;
+esac
+
+CACHE="$ROOT/var/cache"
+LIB="$ROOT/lib-ladybug"
+ZIG_DIR="$ROOT/var/zig-dist"
+ZIG_BIN="$ROOT/var/zig/zig"
+
+sha256of() {
+  if command -v sha256sum >/dev/null 2>&1; then
+    sha256sum "$1" | awk '{print $1}'
+  else
+    shasum -a 256 "$1" | awk '{print $1}'
+  fi
+}
+
+fetch() {
+  local url="$1" dest="$2" expect="$3"
+  if [ -f "$dest" ] && [ "$(sha256of "$dest")" = "$expect" ]; then
+    return 0
+  fi
+  mkdir -p "$(dirname "$dest")"
+  echo "fetch $url" >&2
+  curl -fsSL "$url" -o "$dest"
+  local got
+  got="$(sha256of "$dest")"
+  if [ "$got" != "$expect" ]; then
+    echo "checksum mismatch $dest: got $got want $expect" >&2
+    rm -f "$dest"
+    exit 1
+  fi
+}
+
+ensure_zig() {
+  if [ -n "${ZIG:-}" ] && [ -x "$ZIG" ]; then
+    return 0
+  fi
+  if [ -x "$ZIG_BIN" ]; then
+    export ZIG="$ZIG_BIN"
+    return 0
+  fi
+  if command -v zig >/dev/null 2>&1; then
+    export ZIG
+    ZIG="$(command -v zig)"
+    return 0
+  fi
+  local tar="$CACHE/zig-${ZIG_ARCH}-linux-${ZIG_VERSION}.tar.xz"
+  fetch "https://ziglang.org/download/${ZIG_VERSION}/zig-${ZIG_ARCH}-linux-${ZIG_VERSION}.tar.xz" \
+    "$tar" "$ZIG_SHA"
+  mkdir -p "$CACHE"
+  rm -rf "$ZIG_DIR"
+  tar -xJf "$tar" -C "$CACHE"
+  mv "$CACHE/zig-${ZIG_ARCH}-linux-${ZIG_VERSION}" "$ZIG_DIR"
+  mkdir -p "$ROOT/var/zig"
+  ln -sfn "$ZIG_DIR/zig" "$ZIG_BIN"
+  export ZIG="$ZIG_BIN"
+}
+
+ensure_libs() {
+  mkdir -p "$LIB"
+  if [ ! -f "$LIB/liblbug.so" ]; then
+    local tar="$CACHE/liblbug-linux-${LBUG_ARCH}.tar.gz"
+    fetch "https://github.com/LadybugDB/ladybug/releases/download/v${LBUG_VERSION}/liblbug-linux-${LBUG_ARCH}.tar.gz" \
+      "$tar" "$LBUG_SHA"
+    tar -xzf "$tar" -C "$LIB"
+  fi
+  if [ ! -f "$LIB/libtokenizers.a" ]; then
+    local tar="$CACHE/libtokenizers.linux-${TOK_ARCH}.tar.gz"
+    fetch "https://github.com/daulet/tokenizers/releases/download/v${TOKENIZERS_VERSION}/libtokenizers.linux-${TOK_ARCH}.tar.gz" \
+      "$tar" "$TOK_SHA"
+    tar -xzf "$tar" -C "$LIB"
+  fi
+}
+
+print_env() {
+  printf 'export ZIG=%q\n' "$ZIG"
+  printf 'export CC=%q\n' "$ROOT/bin/cgo/zcc"
+  printf 'export CXX=%q\n' "$ROOT/bin/cgo/zc++"
+  printf 'export CGO_ENABLED=1\n'
+  printf 'export CGO_CFLAGS=%q\n' "-I$LIB"
+  printf 'export CGO_LDFLAGS=%q\n' "-L$LIB -Wl,-rpath,${CGO_RPATH:-$LIB}"
+}
+
+ensure_zig
+ensure_libs
+
+cmd="${1:-env}"
+if [ "$cmd" = "env" ]; then
+  print_env
+  exit 0
+fi
+
+eval "$(print_env)"
+exec "$@"

bin/chats/sync.go

@@ -29,10 +29,11 @@ func main() {
 	case "linkedin":
 		os.Exit(chats.RunSyncLinkedIn(args))
 	case "whatsapp":
-		fmt.Fprintln(os.Stderr, "chats: WhatsApp not implemented yet")
+		fmt.Fprintln(os.Stderr, "chats: WhatsApp sync is out of v1")
 		os.Exit(1)
 	case "help", "-h", "--help":
-		fmt.Fprintln(os.Stderr, `usage: bin/chats/sync.go telegram|linkedin [flags]`)
+		fmt.Fprintln(os.Stderr, `usage: bin/chats/sync.go telegram|linkedin [flags]
+WhatsApp sync is out of v1.`)
 		return
 	default:
 		fmt.Fprintf(os.Stderr, "chats: unknown platform %q\n", platform)

bin/docker-entrypoint

@@ -1,13 +1,10 @@
 #!/usr/bin/env bash
 # bin/docker-entrypoint - run 2dph tools inside the container.
 #
-#   brain                 shell (default)
+# API image (Zig CGO binaries):
-#   brain search <q>      bin/brain/search.go
+#   serve | search | watch
-#   brain index           bin/kb/index --with-mail
+# Index image (Python write path, compose profile `index`):
-#   brain watch <dir>     compiled /app/bin/watch (bin/brain/watch.go)
+#   index | extract | audit | search (deprecated python wrapper)
-#   brain serve           compiled /app/bin/serve (bin/brain/serve.go)
-#   brain extract         bin/facts/extract (docker×compose pairing)
-#   brain audit           bin/facts/audit
 #
 # Usage comment starts at line 2 (self-describing convention).
 set -euo pipefail
@@ -15,6 +12,20 @@ set -euo pipefail
 CMD="${1:-shell}"
 shift || true
 
+if [ -x /usr/local/bin/brain-serve ]; then
+	case "$CMD" in
+		shell) exec bash ;;
+		serve) exec /usr/local/bin/brain-serve "$@" ;;
+		search) exec /usr/local/bin/brain-search "$@" ;;
+		watch) exec /usr/local/bin/brain-watch "$@" ;;
+		index)
+			echo "index is the Python sidecar: docker compose --profile index run --rm index" >&2
+			exit 2
+			;;
+		*) echo "unknown command: $CMD (api: serve|search|watch)" >&2; exit 2 ;;
+	esac
+fi
+
 case "$CMD" in
 	shell) exec bash ;;
 	search) exec "$KB_PY" /app/bin/kb/search "$@" ;;

bin/kb/add

@@ -0,0 +1,114 @@
+#!/usr/bin/env python3
+"""kb/add - incremental leaf write (no rebuild).
+
+  bin/kb/add --text T --root facts|info --source S
+  bin/kb/add --json                 # stdin: one object or {"leafs":[...]}
+  bin/kb/add --db PATH --json
+
+Writes facts+info in one Ladybug transaction. Does not delete kb.lbug.
+Embedding is used when provided; otherwise model2vec encodes the text.
+"""
+from __future__ import annotations
+
+import json
+import sys
+from pathlib import Path
+
+ROOT = Path(__file__).resolve().parents[2]
+sys.path.insert(0, str(ROOT / "bin" / "tools"))
+
+from kblib import (  # noqa: E402
+    EMBED_DIM,
+    add_leafs,
+    connect,
+    ensure_indexes,
+    init_schema,
+)
+
+
+def _as_leafs(payload: object) -> list[dict]:
+    if isinstance(payload, list):
+        return [dict(x) for x in payload]
+    if isinstance(payload, dict):
+        if "leafs" in payload:
+            return [dict(x) for x in payload["leafs"]]
+        return [dict(payload)]
+    raise ValueError("json must be an object, a list, or {leafs:[...]}")
+
+
+def _embed_missing(leafs: list[dict]) -> None:
+    missing = [lf for lf in leafs if not lf.get("embedding")]
+    if not missing:
+        return
+    from model2vec import StaticModel
+
+    model = StaticModel.from_pretrained("minishlab/potion-multilingual-128M")
+    for lf in missing:
+        text = str(lf.get("text") or "")
+        vec = model.encode([text])[0].astype(float).tolist()
+        if len(vec) != EMBED_DIM:
+            vec = (vec + [0.0] * EMBED_DIM)[:EMBED_DIM]
+        lf["embedding"] = vec
+
+
+def main(argv: list[str]) -> int:
+    import argparse
+
+    p = argparse.ArgumentParser(description="add leafs without rebuilding the brain")
+    p.add_argument("--db", default="", help="path to kb.lbug (default var/kb.lbug)")
+    p.add_argument("--json", action="store_true", help="read leaf JSON from stdin")
+    p.add_argument("--text", default="", help="leaf text")
+    p.add_argument("--root", default="info", choices=("facts", "info"))
+    p.add_argument("--source", default="")
+    p.add_argument("--confidence", default="confirmed")
+    p.add_argument("--source-rev", default="working-tree")
+    p.add_argument("--how", default="brain/add")
+    p.add_argument("--loc", default="")
+    p.add_argument("--type", default="reference", dest="type_")
+    args = p.parse_args(argv)
+
+    if args.json:
+        raw = sys.stdin.read()
+        if not raw.strip():
+            print("kb/add: empty stdin", file=sys.stderr)
+            return 2
+        leafs = _as_leafs(json.loads(raw))
+    else:
+        if not args.text or not args.source:
+            print("kb/add: --text and --source are required (or --json)", file=sys.stderr)
+            return 2
+        leafs = [{
+            "text": args.text,
+            "root": args.root,
+            "source": args.source,
+            "confidence": args.confidence,
+            "source_rev": args.source_rev,
+            "how": args.how,
+            "loc": args.loc or args.source,
+            "type": args.type_,
+        }]
+
+    for lf in leafs:
+        if not lf.get("text") or not lf.get("source"):
+            print("kb/add: each leaf needs text and source", file=sys.stderr)
+            return 2
+
+    _embed_missing(leafs)
+
+    from kblib import DB_PATH, VAR
+
+    dbpath = Path(args.db) if args.db else DB_PATH
+    dbpath.parent.mkdir(parents=True, exist_ok=True)
+    VAR.mkdir(exist_ok=True)
+    db, conn = connect(dbpath, read_only=False)
+    init_schema(conn)
+    ids = add_leafs(conn, leafs)
+    ensure_indexes(conn)
+    conn.close()
+    db.close()
+    print(json.dumps({"mode": "add", "ids": ids, "db": str(dbpath)}))
+    return 0
+
+
+if __name__ == "__main__":
+    sys.exit(main(sys.argv[1:]))

bin/kb/index

@@ -2,12 +2,15 @@
 """kb/index - build the 2dph brain from markdown + factual leafs.
 
   bin/kb/index [--corpus DIR] [--rebuild] [--limit N]
+  bin/kb/index --rebuild --with-facts --with-chats
   bin/kb/index --json                 # emit stats as JSON
 
 Reads every .md under the corpus (default: repo root docs, skills, READMEs)
 as `info` leafs, embeds them with model2vec (potion-multilingual-128M), and
 writes them into var/kb.lbug with FTS + HNSW indexes. `facts` leafs come
-from bin/facts/extract (docker x compose x ssh-config pairing).
+from bin/facts/extract (docker × compose × ssh-config pairing) when
+`--with-facts` is set. `--with-chats` indexes markdown under var/chats/md
+(or a given dir) as info. WhatsApp sync stays out of v1.
 
 --rebuild drops the database file and indexes from scratch. Without it a run
 is idempotent (MERGE by (source,text) id).
@@ -22,7 +25,7 @@ ROOT = Path(__file__).resolve().parents[2]
 sys.path.insert(0, str(ROOT / "bin" / "tools"))
 
 from kblib import (  # noqa: E402
-    connect, ensure_indexes, init_schema, upsert_leaf,
+    add_leafs, connect, ensure_indexes, init_schema, upsert_leaf, link_from_file,
     open_readonly, stats,
 )
 from mdleaves import read_markdown, to_all, walk_markdown  # noqa: E402
@@ -81,10 +84,11 @@ def index_leafs(conn, leafs: list[dict], embed_fn, limit: int) -> tuple[int, int
     for lf in leafs[:limit] if limit else leafs:
         query = f"{lf['heading']}\n\n{lf['text']}"
         emb = embed_fn(lf["text"]) if lf["text"] else None
-        upsert_leaf(conn, text=query, root="info", confidence="confirmed",
+        lid = upsert_leaf(conn, text=query, root="info", confidence="confirmed",
                     source=lf["source"], source_rev="working-tree",
                     how="kb/index", loc=lf["source"], type_=lf.get("type", "reference"),
                     embedding=emb)
+        link_from_file(conn, lid, lf["source"], repo=str(lf.get("repo") or ""))
         count += 1
     return count, len(leafs)
 
@@ -95,12 +99,65 @@ def embedder():
     return lambda text: model.encode([text])[0].astype(float).tolist()
 
 
+def index_fact_dicts(conn, facts: list[dict], embed_fn) -> int:
+    """Write extract-shaped dicts as root=facts leafs (2-source source field)."""
+    leafs = []
+    for f in facts:
+        text = str(f.get("text") or "")
+        source = str(f.get("source") or "")
+        if not text or not source:
+            continue
+        leafs.append({
+            "text": text,
+            "root": "facts",
+            "confidence": "confirmed",
+            "source": source,
+            "source_rev": f.get("source_rev") or "working-tree",
+            "how": f.get("how") or "facts/extract",
+            "loc": f.get("loc") or source,
+            "type": "fact",
+            "embedding": embed_fn(text) if text else None,
+        })
+    return len(add_leafs(conn, leafs))
+
+
+def facts_from_extract() -> list[dict]:
+    import subprocess
+    proc = subprocess.run(
+        [sys.executable, str(ROOT / "bin" / "facts" / "extract"), "--json", "--dry-run"],
+        cwd=ROOT,
+        capture_output=True,
+        text=True,
+        check=False,
+    )
+    if proc.returncode != 0:
+        print(f"kb/index: facts/extract failed: {proc.stderr}", file=sys.stderr)
+        return []
+    try:
+        payload = json.loads(proc.stdout)
+    except json.JSONDecodeError:
+        print("kb/index: facts/extract produced non-JSON", file=sys.stderr)
+        return []
+    return list(payload.get("facts") or [])
+
+
 def main(argv: list[str]) -> int:
     import argparse
     p = argparse.ArgumentParser(description="build the 2dph brain index")
     p.add_argument("--corpus", action="append", help="extra markdown dir/file to index (may repeat)")
     p.add_argument("--rebuild", action="store_true", help="fresh db + indexes")
+    p.add_argument("--db", default="", help="path to kb.lbug (default var/kb.lbug)")
+    p.add_argument("--no-defaults", action="store_true", help="do not index repo README/docs/skills")
     p.add_argument("--with-mail", action="store_true", help="include var/mail message.md leafs")
+    p.add_argument("--with-facts", action="store_true", help="run facts/extract into root=facts")
+    p.add_argument("--facts-json", default="", help="JSON list (or {facts:[...]}) of fact dicts")
+    p.add_argument(
+        "--with-chats",
+        nargs="?",
+        const=str(ROOT / "var" / "chats" / "md"),
+        default="",
+        help="index chat markdown as info (default var/chats/md)",
+    )
     p.add_argument("--since", default="", help="with --with-mail, only messages dated >= YYYY-MM-DD")
     p.add_argument("--dry-run", action="store_true", help="count leafs, write nothing")
     p.add_argument(
@@ -114,44 +171,71 @@ def main(argv: list[str]) -> int:
 
     from kblib import DB_PATH, VAR
 
-    leafs = load_corpus(ROOT)
+    dbpath = Path(a.db) if a.db else DB_PATH
+    leafs: list[dict] = [] if a.no_defaults else load_corpus(ROOT)
     if a.corpus:
         for source in a.corpus:
             leafs.extend(load_corpus_glob(source))
+    chat_n = 0
+    if a.with_chats:
+        chats = load_corpus_glob(a.with_chats)
+        chat_n = len(chats)
+        leafs.extend(chats)
     mail_n = 0
     if a.with_mail:
         mail = from_mail_root(ROOT / "var" / "mail", since=a.since)
         mail_n = len(mail)
         leafs.extend(mail)
 
+    facts: list[dict] = []
+    if a.facts_json:
+        raw = Path(a.facts_json).read_text(encoding="utf-8")
+        payload = json.loads(raw)
+        facts = list(payload.get("facts") if isinstance(payload, dict) else payload)
+    if a.with_facts:
+        facts.extend(facts_from_extract())
+
     if a.dry_run:
-        msg = {"indexed": 0, "corpus_total": len(leafs), "mail_leafs": mail_n, "dry_run": True}
+        msg = {
+            "indexed": 0,
+            "corpus_total": len(leafs),
+            "mail_leafs": mail_n,
+            "chat_leafs": chat_n,
+            "facts_leafs": len(facts),
+            "dry_run": True,
+        }
         print(json.dumps(msg, indent=2) if a.json else
-              f"brain/index: {len(leafs)} leafs would be indexed (mail={mail_n})")
+              f"brain/index: {len(leafs)} info + {len(facts)} facts would be indexed")
         return 0
 
     VAR.mkdir(exist_ok=True)
-    if a.rebuild and DB_PATH.exists():
+    dbpath.parent.mkdir(parents=True, exist_ok=True)
-        DB_PATH.unlink()
+    if a.rebuild and dbpath.exists():
+        dbpath.unlink()
 
-    db, conn = connect(DB_PATH, read_only=False)
+    db, conn = connect(dbpath, read_only=False)
     init_schema(conn)
 
-    # Never DROP FTS/VECTOR (ghost catalog). Write leafs, then ensure indexes
-    # unless --skip-indexes (seed facts first — MERGE under live FTS corrupts it).
-    # --rebuild already deleted kb.lbug above, so CREATE runs on a clean DB.
     embed = embedder()
     done, total = index_leafs(conn, leafs, embed, a.limit)
+    fact_n = index_fact_dicts(conn, facts, embed) if facts else 0
     if not a.skip_indexes:
         ensure_indexes(conn)
     s = stats(conn)
     conn.close()
     db.close()
 
-    result = {"indexed": done, "corpus_total": total, **{k: v for k, v in s.items() if k in ("total", "by_root")}}
+    result = {
+        "indexed": done,
+        "corpus_total": total,
+        "facts_leafs": fact_n,
+        "chat_leafs": chat_n,
+        **{k: v for k, v in s.items() if k in ("total", "by_root")},
+    }
     if a.skip_indexes:
         result["indexes"] = "skipped"
-    print(json.dumps(result, indent=2) if a.json else f"indexed {done}/{total} leafs; db total {s['total']}")
+    print(json.dumps(result, indent=2) if a.json else
+          f"indexed {done}/{total} info + {fact_n} facts; db total {s['total']}")
     return 0
 
 

bin/kb/search

@@ -1,10 +1,9 @@
 #!/usr/bin/env bash
 # bin/kb/search — deprecated wrapper. Use bin/brain/search.go.
-# Sets CGO for ladybug, builds a binary (embed daemon needs a real executable),
+# CGO via Zig (bin/cgo/zig), not gcc. Builds a binary then execs it.
-# then execs it. Prints one deprecation line.
 set -euo pipefail
 
-ROOT="$(cd "$(dirname "$0")/../.." && pwd)"
+ROOT="$(CDPATH= cd -- "$(dirname "$0")/../.." && pwd)"
 BIN="$ROOT/var/bin/brain-search"
 SRC="$ROOT/internal/brain"
 CMD="$ROOT/bin/brain"
@@ -24,11 +23,10 @@ else
 fi
 
 if [ "$need_build" -eq 1 ]; then
-	echo "Building brain/search..." >&2
+	echo "Building brain/search (zig cc)..." >&2
 	(
 		cd "$ROOT" &&
-			CGO_CFLAGS="-I$ROOT/lib-ladybug" \
+			eval "$("$ROOT/bin/cgo/zig" env)" &&
-			CGO_LDFLAGS="-L$ROOT/lib-ladybug -Wl,-rpath,$ROOT/lib-ladybug" \
 			go build -tags system_ladybug -o "$BIN" ./bin/brain
 	) || exit 1
 fi

bin/mail/import

@@ -7,7 +7,7 @@
   bin/mail/import --since 2026-01-01    only messages after a date
   bin/mail/import --limit 50            cap messages per run
   bin/mail/import --no-attachments      body only, skip attachment conversion
-  bin/mail/import --ocr                OCR scanned PDFs/images via docling
+  bin/mail/import --ocr                OCR images (PDFs OCR when textless)
   bin/mail/import --dry-run            list messages without writing anything
 
 Writes one directory per message: var/mail/{folder}/{message_id}/
@@ -16,10 +16,11 @@ Writes one directory per message: var/mail/{folder}/{message_id}/
   attachments/*.md          converted attachment content
 
 Indexing is a separate step (`bin/brain/index.go --rebuild`): conversion can
-crash in native docling and must not leave the brain DB mid-transaction.
+crash and must not leave the brain DB mid-transaction.
 
-Requires ONLYOFFICE_URL/USER/PASS in .env (or env). Idempotent: a message
+Requires ONLYOFFICE_URL/USER/PASS in .env (or env) except `--from-raw`.
-already present (message.md exists) is skipped unless --force.
+Idempotent: a message already present (message.md exists) is skipped unless
+--force.
 """
 from __future__ import annotations
 
@@ -41,10 +42,10 @@ from mailconv import (  # noqa: E402
     IMAGE_SUFFIXES,
     LEGACY_OFFICE_SUFFIXES,
     TEXT_SUFFIXES,
+    convert_pdf,
     html_to_markdown,
-    is_convertible,
     normalize_markdown,
-    subject_to_filename,
+    ocr_image,
     zip_extract_safe,
 )
 
@@ -147,9 +148,9 @@ def convert_file_to_md(path: Path, ocr: bool) -> str | None:
         except Exception as e:
             return f"\n<!-- conversion failed: {e} -->\n"
     if suffix == ".pdf":
-        return _convert_pdf(path, ocr)
+        return convert_pdf(path, ocr)
     if suffix in IMAGE_SUFFIXES and ocr:
-        return _convert_pdf(path, ocr)
+        return ocr_image(path) or "\n<!-- ocr unavailable -->\n"
     if suffix in LEGACY_OFFICE_SUFFIXES:
         return _convert_legacy(path)
     if suffix in ARCHIVE_SUFFIXES:
@@ -157,67 +158,6 @@ def convert_file_to_md(path: Path, ocr: bool) -> str | None:
     return None
 
 
-def _convert_pdf(path: Path, ocr: bool) -> str:
-    """Convert one PDF to markdown.
-
-    Fast path: poppler's pdftotext (-layout) extracts exact text from
-    born-digital PDFs in ~15ms vs docling's 1-3s. Only textless PDFs (scanned
-    pages, layout-heavy) fall back to docling, which runs isolated in a
-    subprocess because its native onnx/RT-DETR has segfaulted the main process.
-    """
-    text = _pdf_fast_text(path)
-    if ocr or text is None or not text.strip():
-        return _convert_pdf_docling(path, ocr)
-    return normalize_markdown(text)
-
-
-def _pdf_fast_text(path: Path) -> str | None:
-    """pdftotext -layout; None when poppler is unavailable (or the PDF has no text layer)."""
-    try:
-        proc = subprocess.run(
-            ["pdftotext", "-layout", str(path), "-"],
-            capture_output=True, timeout=60)
-    except (OSError, subprocess.TimeoutExpired):
-        return None
-    if proc.returncode != 0:
-        return None
-    return proc.stdout.decode("utf-8", errors="replace")
-
-
-def _convert_pdf_docling(path: Path, ocr: bool) -> str:
-    try:
-        proc = subprocess.run(
-            [sys.executable, os.path.abspath(__file__), "--pdf-worker", str(path),
-             "--ocr" if ocr else "--no-ocr"],
-            capture_output=True, text=True, timeout=600)
-    except subprocess.TimeoutExpired:
-        return "\n<!-- pdf conversion timed out -->\n"
-    if proc.returncode != 0:
-        tail = proc.stderr.strip().splitlines()[-3:]
-        return f"\n<!-- pdf conversion failed: {proc.returncode}: {' | '.join(tail)} -->\n"
-    return proc.stdout
-
-
-def _pdf_worker(path: Path, ocr: bool) -> None:
-    """docling worker entry: prints converted markdown on stdout, exits non-zero on error."""
-    try:
-        from docling.document_converter import DocumentConverter, PdfFormatOption
-        from docling.datamodel.pipeline_options import PdfPipelineOptions
-        opts = PdfPipelineOptions()
-        opts.do_ocr = bool(ocr)
-        opts.do_table_structure = True
-        conv = DocumentConverter(format_options={"pdf": PdfFormatOption(pipeline_options=opts)})
-        res = conv.convert(str(path))
-        sys.stdout.write(normalize_markdown(res.document.export_to_markdown()))
-        sys.exit(0)
-    except Exception as e:
-        # errors/stacktraces to stderr; the caller only reports a one-liner
-        print(f"pdf-worker: {e}", file=sys.stderr)
-        import traceback
-        traceback.print_exc(file=sys.stderr)
-        sys.exit(1)
-
-
 def _convert_legacy(path: Path) -> str:
     """Legacy .doc/.xls/.ppt -> md via pandoc (installed) or a stub."""
     try:
@@ -356,19 +296,12 @@ def main(argv: list[str]) -> int:
     p.add_argument("--from-raw", default="",
                    help="convert Go-synced dirs (var/mail/<folder>/<id>/message.json) to markdown")
     p.add_argument("--no-attachments", action="store_true", help="skip attachment download+convert")
-    p.add_argument("--ocr", action="store_true", help="OCR scanned PDFs/images via docling")
+    p.add_argument("--ocr", action="store_true", help="OCR images (PDFs OCR when textless)")
     p.add_argument("--force", action="store_true", help="re-import even if message.md exists")
     p.add_argument("--dry-run", action="store_true", help="list messages, write nothing")
     p.add_argument("--json", action="store_true")
-    p.add_argument("--pdf-worker", default="", help=argparse.SUPPRESS)
-    p.add_argument("--no-ocr", action="store_true", help=argparse.SUPPRESS)
     a = p.parse_args(argv)
 
-    if a.pdf_worker:
-        _pdf_worker(Path(a.pdf_worker), ocr=not a.no_ocr)
-        return 0
-
-    conf = load_env()
     fid = folder_id(a.folder)
     out_root = ROOT / "var" / "mail"
     summary: list[dict] = []
@@ -394,6 +327,7 @@ def main(argv: list[str]) -> int:
                                      target_dir=msg_dir.parent))
             summary.append(entry)
     else:
+        conf = load_env()
         OOCLIENT = OOClient(conf)
         if a.id:
             messages = [{"id": i} for i in a.id]

bin/mail/ocr.go

@@ -0,0 +1,48 @@
+//usr/bin/env go run -tags=mail_ocr "$0" "$@"; exit
+//go:build mail_ocr
+//
+// bin/mail/ocr.go - OCR an image or scanned PDF (tesseract eng+deu).
+//
+//	./bin/mail/ocr.go scan.png
+//	./bin/mail/ocr.go scan.pdf
+//	OCR_ENGINE=paddle ./bin/mail/ocr.go scan.png
+//
+// PDFs try pdftotext -layout first; empty text layer uses pdftoppm + tesseract.
+// No gocv. Tesseract CGO bindings are not used (D21 Zig owns Ladybug CGO).
+// NOTE: never run `gofmt -w` on this file — it breaks the shebang.
+package main
+
+import (
+	"fmt"
+	"os"
+	"strings"
+
+	"github.com/eSlider/2dph/internal/ocr"
+)
+
+func main() {
+	os.Exit(run(os.Args[1:]))
+}
+
+func run(args []string) int {
+	if len(args) != 1 || strings.HasPrefix(args[0], "-") {
+		fmt.Fprintln(os.Stderr, `usage: bin/mail/ocr.go <image|pdf>`)
+		return 2
+	}
+	path := args[0]
+	var (
+		text string
+		err  error
+	)
+	if strings.HasSuffix(strings.ToLower(path), ".pdf") {
+		text, err = ocr.PDFFile(path)
+	} else {
+		text, err = ocr.ImageFile(path)
+	}
+	if err != nil {
+		fmt.Fprintf(os.Stderr, "mail/ocr: %v\n", err)
+		return 1
+	}
+	fmt.Println(text)
+	return 0
+}

bin/markdown/import.go

@@ -1,20 +1,100 @@
-//usr/bin/env go run -tags=markdown_import "$0" "$@"; exit
+//usr/bin/env go run "$0" "$@"; exit
-//go:build markdown_import
 //
-// bin/markdown/import.go - split markdown into leafs (mistune).
+// bin/markdown/import.go - split markdown into leafs (H2 boundaries).
 //
 //	./bin/markdown/import.go [dir]
 //	./bin/markdown/import.go --files a.md,b.md --json
 //
+// Conversion only. Brain write is bin/brain/index.go.
+// Python bin/md/import remains as a fallback.
 // NOTE: never run `gofmt -w` on this file — it breaks the shebang.
 package main
 
 import (
+	"fmt"
 	"os"
+	"strings"
 
-	"github.com/eSlider/2dph/internal/cmdbin"
+	"github.com/eSlider/2dph/internal/mdleaves"
 )
 
 func main() {
-	os.Exit(cmdbin.ExecFile("bin/md/import", os.Args[1:]))
+	os.Exit(run(os.Args[1:]))
+}
+
+func run(args []string) int {
+	jsonOut := false
+	files := ""
+	root := "."
+	for i := 0; i < len(args); i++ {
+		a := args[i]
+		switch {
+		case a == "--json":
+			jsonOut = true
+		case a == "--files" && i+1 < len(args):
+			i++
+			files = args[i]
+		case strings.HasPrefix(a, "--files="):
+			files = strings.TrimPrefix(a, "--files=")
+		case a == "-h" || a == "--help":
+			fmt.Fprintln(os.Stderr, "bin/markdown/import.go [dir] [--files a.md,b.md] [--json]")
+			return 0
+		case strings.HasPrefix(a, "-"):
+			fmt.Fprintln(os.Stderr, "unknown arg:", a)
+			return 2
+		default:
+			root = a
+		}
+	}
+
+	var paths []string
+	if files != "" {
+		for _, f := range strings.Split(files, ",") {
+			f = strings.TrimSpace(f)
+			if f != "" {
+				paths = append(paths, f)
+			}
+		}
+	} else {
+		st, err := os.Stat(root)
+		if err != nil {
+			fmt.Fprintf(os.Stderr, "md/import: no such path %s\n", root)
+			return 2
+		}
+		if !st.IsDir() {
+			paths = []string{root}
+		} else {
+			var err error
+			paths, err = mdleaves.WalkMarkdown(root)
+			if err != nil {
+				fmt.Fprintf(os.Stderr, "md/import: %v\n", err)
+				return 1
+			}
+		}
+	}
+	if len(paths) == 0 {
+		fmt.Fprintln(os.Stderr, "md/import: no markdown files")
+		return 1
+	}
+
+	var all []mdleaves.Leaf
+	for _, p := range paths {
+		raw, err := os.ReadFile(p)
+		if err != nil {
+			fmt.Fprintf(os.Stderr, "md/import: %s: %v\n", p, err)
+			continue
+		}
+		all = append(all, mdleaves.ToAll(string(raw), p, "")...)
+	}
+	if jsonOut {
+		s, err := mdleaves.EncodeJSON(all)
+		if err != nil {
+			fmt.Fprintln(os.Stderr, err)
+			return 1
+		}
+		fmt.Print(s)
+		return 0
+	}
+	fmt.Print(mdleaves.EncodeYAML(all))
+	return 0
 }

bin/reasoner/bakeoff.go

@@ -0,0 +1,91 @@
+//usr/bin/env go run -tags=reasoner_bakeoff "$0" "$@"; exit
+//go:build reasoner_bakeoff
+//
+// bin/reasoner/bakeoff.go - CPU tool-call bake-off against an OpenAI-compatible URL (D18).
+//
+//	REASONER_BASE_URL=http://127.0.0.1:11435/v1 REASONER_MODEL=qwen3.5:9b ./bin/reasoner/bakeoff.go
+//	./bin/reasoner/bakeoff.go --model MichelRosselli/bonsai-27b:Q1_0 --json
+//
+// Measures OpenAI tool_calls (search/get/audit) and RSS from Ollama /api/ps, not VRAM.
+// PicoClaw is not in this repo; the tool names match internal/httpapi MCP ops.
+// NOTE: never run `gofmt -w` on this file — it breaks the shebang.
+package main
+
+import (
+	"encoding/json"
+	"fmt"
+	"os"
+	"strings"
+
+	"github.com/eSlider/2dph/internal/reasoner"
+)
+
+func main() {
+	os.Exit(run(os.Args[1:]))
+}
+
+func run(args []string) int {
+	base := os.Getenv("REASONER_BASE_URL")
+	if base == "" {
+		base = "http://127.0.0.1:11435/v1"
+	}
+	model := os.Getenv("REASONER_MODEL")
+	if model == "" {
+		model = reasoner.OllamaRAM
+	}
+	jsonOut := false
+	device := "cpu"
+	for i := 0; i < len(args); i++ {
+		a := args[i]
+		switch {
+		case a == "--json":
+			jsonOut = true
+		case a == "--model" && i+1 < len(args):
+			i++
+			model = args[i]
+		case strings.HasPrefix(a, "--model="):
+			model = strings.TrimPrefix(a, "--model=")
+		case a == "--base-url" && i+1 < len(args):
+			i++
+			base = args[i]
+		case a == "--device" && i+1 < len(args):
+			i++
+			device = args[i]
+		case a == "-h" || a == "--help":
+			fmt.Fprintln(os.Stderr, "bin/reasoner/bakeoff.go [--model ID] [--base-url URL] [--device cpu] [--json]")
+			return 0
+		default:
+			fmt.Fprintln(os.Stderr, "unknown arg:", a)
+			return 2
+		}
+	}
+	c := reasoner.Client{BaseURL: base, Model: model, Device: device}
+	rep := reasoner.Run(c)
+	raw, err := json.MarshalIndent(rep, "", "  ")
+	if err != nil {
+		fmt.Fprintln(os.Stderr, err)
+		return 1
+	}
+	if jsonOut {
+		fmt.Println(string(raw))
+	} else {
+		fmt.Printf("model: %s\n", rep.Model)
+		fmt.Printf("hf_id: %s\n", rep.HF)
+		fmt.Printf("device: %s\n", rep.Device)
+		fmt.Printf("tool_call: %d/%d\n", rep.ToolCallOK, rep.ToolCallN)
+		fmt.Printf("xml_leak: %d\n", rep.XMLLeak)
+		fmt.Printf("rss_mb: %d\n", rep.RSSMB)
+		fmt.Printf("vram_mb: %d\n", rep.VRAMMB)
+		for _, p := range rep.Prompts {
+			status := "fail"
+			if p.OK {
+				status = "ok"
+			}
+			fmt.Printf("  %s: %s wanted=%s got=%s xml=%v %dms %s\n", p.WantedTool, status, p.WantedTool, p.ToolName, p.XMLLeak, p.LatencyMS, p.Err)
+		}
+	}
+	if rep.ToolCallN == 0 {
+		return 1
+	}
+	return 0
+}

bin/reasoner/doc.go

@@ -0,0 +1,2 @@
+// Commands in this directory are shebang mains (bakeoff.go).
+package main

bin/tools/kblib.py

@@ -4,7 +4,7 @@ Single embedded graph `var/kb.lbug`. Two roots: facts (assertions backed by
 >=2 independent sources) and info (narrative leafs). Hybrid retrieval: BM25
 (FTS extension) + HNSW cosine (VECTOR extension) + Cypher graph hops.
 
-All access is read-only unless `--rebuild` is passed to kb/index.
+All access is read-only unless `--rebuild` (kb/index) or `kb/add`.
 """
 from __future__ import annotations
 
@@ -118,6 +118,97 @@ def upsert_leaf(conn: ladybug.Connection, *, text: str, root: str, confidence: s
     return lid
 
 
+def add_leafs(conn: ladybug.Connection, leafs: list[dict]) -> list[str]:
+    """Write facts+info leafs in one transaction. Safe while FTS/HNSW exist.
+
+    Each leaf dict: text, source, optional root/confidence/source_rev/how/loc/type/embedding.
+    Does not delete the database file. Measured on Ladybug 0.19: MERGE of new
+    ids (and updates) stays FTS+HNSW queryable; DROP INDEX is the fatal path.
+    """
+    if not leafs:
+        return []
+    started = False
+    try:
+        conn.execute("BEGIN TRANSACTION")
+        started = True
+    except Exception:
+        started = False
+    ids: list[str] = []
+    try:
+        for lf in leafs:
+            ids.append(
+                upsert_leaf(
+                    conn,
+                    text=str(lf["text"]),
+                    root=str(lf.get("root") or ROOT_INFO),
+                    confidence=str(lf.get("confidence") or CONF_CONFIRMED),
+                    source=str(lf["source"]),
+                    source_rev=str(lf.get("source_rev") or "working-tree"),
+                    how=str(lf.get("how") or "brain/add"),
+                    loc=str(lf.get("loc") or lf.get("source") or ""),
+                    type_=str(lf.get("type") or lf.get("type_") or "reference"),
+                    embedding=lf.get("embedding"),
+                )
+            )
+        if started:
+            conn.execute("COMMIT")
+    except Exception:
+        if started:
+            try:
+                conn.execute("ROLLBACK")
+            except Exception:
+                pass
+        raise
+    return ids
+
+
+def file_id(repo: str, path: str) -> str:
+    """Stable File.id matching gitimport (`repo:path`)."""
+    return f"{repo}:{path}" if repo else path
+
+
+def link_from_file(conn: ladybug.Connection, leaf_id: str, path: str,
+                   repo: str = "", mtime: str = "") -> str:
+    """MERGE File and Leaf-[:FROM_FILE]->File so --hop 1 can walk."""
+    fid = file_id(repo, path)
+    conn.execute(
+        "MERGE (f:File {id:$id}) SET f.path=$path, f.repo=$repo, f.mtime=$mtime",
+        parameters={"id": fid, "path": path, "repo": repo, "mtime": mtime},
+    )
+    conn.execute(
+        "MATCH (l:Leaf {id:$lid}), (f:File {id:$fid}) "
+        "MERGE (l)-[:FROM_FILE]->(f)",
+        parameters={"lid": leaf_id, "fid": fid},
+    )
+    return fid
+
+
+HOP_STMTS = {
+    1: "MATCH (l:Leaf {id:$id})-[:FROM_FILE]->(f:File) RETURN f.id, f.path, 1",
+    2: ("MATCH (l:Leaf {id:$id})-[:FROM_FILE]->(f:File)-[:HAS_VERSION]->(c:Commit) "
+        "RETURN c.id, c.subject, 2"),
+    3: ("MATCH (l:Leaf {id:$id})-[:FROM_FILE]->(f:File)-[:HAS_VERSION]->(c:Commit)"
+        "-[:AUTHORED]->(p:Person) RETURN p.id, p.name, 3"),
+}
+HOP_LABELS = {1: "File", 2: "Commit", 3: "Person"}
+
+
+def hop_walk(conn: ladybug.Connection, leaf_id: str, n: int) -> list[dict]:
+    """Walk Leaf → File → Commit → Person up to n hops (max 3)."""
+    depth = min(max(int(n), 0), 3)
+    out: list[dict] = []
+    for d in range(1, depth + 1):
+        rows = conn.execute(HOP_STMTS[d], parameters={"id": leaf_id}).get_all()
+        for row in rows:
+            out.append({
+                "id": row[0],
+                "label": HOP_LABELS[d],
+                "name": row[1],
+                "depth": int(row[2]),
+            })
+    return out
+
+
 def leaf_index_names(conn: ladybug.Connection) -> set[str]:
     """Return index names on the Leaf table (e.g. {'id', 'Leaf_vec', '_PK'})."""
     rows = conn.execute("CALL SHOW_INDEXES() RETURN *").get_all()

bin/tools/mailconv.py

@@ -7,7 +7,10 @@ offline against fixtures.
 from __future__ import annotations
 
 import html
+import os
 import re
+import subprocess
+import tempfile
 import zipfile
 from pathlib import Path
 
@@ -18,9 +21,10 @@ OFFICE_SUFFIXES = {".docx", ".pptx", ".xlsx", ".html", ".htm", ".epub", ".eml",
 PDF_SUFFIXES = {".pdf"}
 IMAGE_SUFFIXES = {".png", ".jpg", ".jpeg", ".gif", ".bmp", ".tiff", ".tif", ".webp"}
 ARCHIVE_SUFFIXES = {".zip"}
-# Legacy binary Office (doc/xls/ppt) — markitdown/docling skip them; we try
+# Legacy binary Office (doc/xls/ppt) — markitdown skip them; we try
 # pandoc first, else leave a stub.
 LEGACY_OFFICE_SUFFIXES = {".doc", ".xls", ".ppt"}
+TESS_LANG = "eng+deu"
 
 CONVERTIBLE_SUFFIXES = (
     TEXT_SUFFIXES | OFFICE_SUFFIXES | PDF_SUFFIXES | IMAGE_SUFFIXES | ARCHIVE_SUFFIXES | LEGACY_OFFICE_SUFFIXES
@@ -146,3 +150,82 @@ def zip_extract_safe(zip_path: Path, dest: Path) -> list[Path]:
 
 def is_convertible(suffix: str) -> bool:
     return suffix.lower() in CONVERTIBLE_SUFFIXES
+
+
+def convert_pdf(path: Path, ocr: bool = False) -> str:
+    """pdftotext -layout first; empty text layer → pdftoppm + tesseract.
+
+    `ocr` is unused for born-digital PDFs (text layer wins). Scans OCR
+    automatically. This path never execs an ONNX document converter.
+    """
+    del ocr  # scans OCR when the text layer is empty; flag is for images
+    text = pdf_fast_text(path)
+    if text and text.strip():
+        return normalize_markdown(text)
+    scanned = ocr_pdf(path)
+    if scanned and scanned.strip():
+        return normalize_markdown(scanned)
+    if text:
+        return normalize_markdown(text)
+    return "\n<!-- pdf has no text layer (ocr unavailable) -->\n"
+
+
+def pdf_fast_text(path: Path) -> str | None:
+    """pdftotext -layout; None when poppler is missing or the command fails."""
+    try:
+        proc = subprocess.run(
+            ["pdftotext", "-layout", str(path), "-"],
+            capture_output=True, timeout=60)
+    except (OSError, subprocess.TimeoutExpired):
+        return None
+    if proc.returncode != 0:
+        return None
+    return proc.stdout.decode("utf-8", errors="replace")
+
+
+def ocr_pdf(path: Path) -> str:
+    """Rasterize with pdftoppm and OCR each page (tesseract or paddle)."""
+    try:
+        with tempfile.TemporaryDirectory(prefix="2dph-ocr-") as tmp:
+            prefix = str(Path(tmp) / "page")
+            proc = subprocess.run(
+                ["pdftoppm", "-png", "-r", "200", str(path), prefix],
+                capture_output=True, timeout=120)
+            if proc.returncode != 0:
+                return ""
+            pages = sorted(Path(tmp).glob("page*.png"))
+            parts = [ocr_image(p) for p in pages]
+            return "\n\n".join(p for p in parts if p and p.strip())
+    except (OSError, subprocess.TimeoutExpired):
+        return ""
+
+
+def ocr_image(path: Path) -> str:
+    engine = os.environ.get("OCR_ENGINE", "tesseract")
+    if engine == "paddle":
+        return _ocr_paddle(path)
+    return _ocr_tesseract(path)
+
+
+def _ocr_tesseract(path: Path) -> str:
+    try:
+        proc = subprocess.run(
+            ["tesseract", str(path), "stdout", "-l", TESS_LANG, "--psm", "6"],
+            capture_output=True, timeout=120)
+    except (OSError, subprocess.TimeoutExpired):
+        return ""
+    if proc.returncode != 0:
+        return ""
+    return proc.stdout.decode("utf-8", errors="replace").strip()
+
+
+def _ocr_paddle(path: Path) -> str:
+    try:
+        proc = subprocess.run(
+            ["paddleocr", "ocr", "-i", str(path)],
+            capture_output=True, timeout=180)
+    except (OSError, subprocess.TimeoutExpired):
+        return ""
+    if proc.returncode != 0:
+        return ""
+    return proc.stdout.decode("utf-8", errors="replace").strip()

bin/tools/test_bin_layout.py

@@ -1,6 +1,7 @@
 """D14 layout: bin/{subject}/{method}.go, libs in internal/, one go.mod."""
 from __future__ import annotations
 
+import os
 import unittest
 from pathlib import Path
 
@@ -74,9 +75,20 @@ class BinLayoutTest(unittest.TestCase):
         )
 
     def test_brain_methods_are_shebangs(self) -> None:
-        for method in ("index.go", "get.go", "stats.go", "eval.go", "watch.go"):
+        for method in ("index.go", "add.go", "get.go", "stats.go", "eval.go", "watch.go"):
             self._assert_shebang(f"bin/brain/{method}")
 
+    def test_brain_add_is_python_write_not_rebuild(self) -> None:
+        self._assert_shebang("bin/brain/add.go")
+        text = (ROOT / "bin" / "brain" / "add.go").read_text()
+        self.assertIn("cmdbin.ExecFile", text)
+        self.assertIn("bin/kb/add", text)
+        self.assertNotIn("--rebuild", text)
+        py = (ROOT / "bin" / "kb" / "add").read_text()
+        self.assertIn("add_leafs", py)
+        self.assertIn("--json", py)
+        self.assertNotIn("unlink", py.lower())
+
     def test_brain_get_stats_eval_are_not_python_exec(self) -> None:
         for method in ("get.go", "stats.go", "eval.go"):
             text = (ROOT / "bin" / "brain" / method).read_text()
@@ -124,8 +136,55 @@ class BinLayoutTest(unittest.TestCase):
             "index_mail must point at bin/brain/index.go",
         )
 
-    def test_markdown_import_is_shebang(self) -> None:
+    def test_mail_ocr_is_tesseract_not_docling(self) -> None:
+        self._assert_shebang("bin/mail/ocr.go")
+        ocr = (ROOT / "bin" / "mail" / "ocr.go").read_text()
+        self.assertIn("internal/ocr", ocr)
+        self.assertIn("mail_ocr", ocr)
+        self.assertNotIn("github.com/otiai10/gosseract", ocr)
+        py = (ROOT / "bin" / "mail" / "import").read_text()
+        self.assertNotIn("from docling", py)
+        self.assertNotIn("import docling", py)
+        self.assertIn("convert_pdf", py)
+        conv = (ROOT / "bin" / "tools" / "mailconv.py").read_text()
+        self.assertIn("pdftotext", conv)
+        self.assertIn("pdftoppm", conv)
+        self.assertIn("tesseract", conv)
+        self.assertIn("eng+deu", conv)
+        self.assertNotIn("from docling", conv)
+        self.assertNotIn("import docling", conv)
+        self.assertNotIn("gocv", conv.lower())
+        proj = (ROOT / "pyproject.toml").read_text()
+        self.assertNotIn("docling", proj)
+        ci = (ROOT / ".github" / "workflows" / "ci.yml").read_text()
+        self.assertIn("tesseract-ocr", ci)
+        self.assertIn("./internal/ocr", ci)
+        compose = (ROOT / "compose.yaml").read_text()
+        self.assertIn("ocr-paddle", compose)
+        self.assertIn("OCR_ENGINE", compose)
+
+    def test_markdown_import_is_go_not_python_exec(self) -> None:
         self._assert_shebang("bin/markdown/import.go")
+        text = (ROOT / "bin" / "markdown" / "import.go").read_text()
+        self.assertNotIn("ExecFile", text)
+        self.assertNotIn("cmdbin", text)
+        self.assertIn("internal/mdleaves", text)
+        self.assertNotIn("kb.lbug", text)
+
+    def test_import_adapters_do_not_write_ladybug(self) -> None:
+        for rel in (
+            "bin/mail/import.go",
+            "bin/mail/import",
+            "bin/markdown/import.go",
+            "bin/chats/import.go",
+            "bin/git/import.go",
+        ):
+            text = (ROOT / rel).read_text()
+            self.assertNotIn("upsert_leaf", text, rel)
+            self.assertNotIn("kb.lbug", text, rel)
+            self.assertNotIn("var/brain.lbug", text, rel)
+        index = (ROOT / "bin" / "brain" / "index.go").read_text()
+        self.assertIn("bin/kb/index", index)
 
     def test_postgres_query_is_shebang(self) -> None:
         self._assert_shebang("bin/postgres/query.go")
@@ -157,3 +216,43 @@ class BinLayoutTest(unittest.TestCase):
         for line in first.splitlines():
             if "go-git/go-git" in line:
                 self.assertNotIn("indirect", line)
+
+    def test_cgo_uses_zig_not_gcc(self) -> None:
+        for rel in ("bin/cgo/zig", "bin/cgo/zcc", "bin/cgo/zc++"):
+            p = ROOT / rel
+            self.assertTrue(p.is_file(), f"missing {rel}")
+            self.assertTrue(
+                os.access(p, os.X_OK),
+                f"{rel} must be executable",
+            )
+        zig = (ROOT / "bin" / "cgo" / "zig").read_text()
+        self.assertIn("zig cc", zig)
+        self.assertIn("0.14.1", zig)
+        zcc = (ROOT / "bin" / "cgo" / "zcc").read_text()
+        self.assertIn('exec "$ZIG" cc', zcc)
+        self.assertNotIn("command -v gcc", zcc)
+        search = (ROOT / "bin" / "kb" / "search").read_text()
+        self.assertIn("bin/cgo/zig", search)
+        self.assertNotIn("command -v gcc", search)
+
+    def test_ci_recall_sot_is_zig_brain_eval(self) -> None:
+        ci = (ROOT / ".github" / "workflows" / "ci.yml").read_text()
+        self.assertIn("bin/brain/eval.go", ci)
+        self.assertIn("system_ladybug,brain_eval", ci)
+        self.assertIn("/tmp/brain-eval", ci)
+        self.assertIn("KB_ROOT", ci)
+        self.assertNotIn("bin/kb/eval", ci)
+        self.assertNotIn("gate skipped", ci)
+        self.assertIn("./bin/facts/audit self", ci)
+
+    def test_eval_fragments_live_in_default_corpus(self) -> None:
+        """CI --rebuild indexes README/PLAN/docs/skills; fragments must be there."""
+        corpus = []
+        for rel in ("README.md", "PLAN.md", "AGENTS.md"):
+            corpus.append((ROOT / rel).read_text())
+        for d in ("docs", "skills"):
+            for p in (ROOT / d).rglob("*.md"):
+                corpus.append(p.read_text())
+        blob = "\n".join(corpus)
+        for frag in ("BM25", "DevOps", "LadybugDB"):
+            self.assertIn(frag, blob, f"{frag} must appear in default index corpus")

bin/tools/test_index_adapters.py

@@ -0,0 +1,97 @@
+"""Import adapters write files only. Index rebuild is brain/index (D14 / Gitea #7)."""
+from __future__ import annotations
+
+import json
+import os
+import subprocess
+import sys
+import tempfile
+import unittest
+from pathlib import Path
+
+ROOT = Path(__file__).resolve().parents[2]
+
+
+class IndexAdapterTest(unittest.TestCase):
+    def test_dry_run_fixture_corpus_does_not_write_lbug(self) -> None:
+        tmp = Path(tempfile.mkdtemp())
+        (tmp / "note.md").write_text("# Fixture\n\n## Leaf\n\nhello corpus\n", encoding="utf-8")
+        lbug = tmp / "kb.lbug"
+        try:
+            import ladybug  # noqa: F401
+        except ImportError:
+            venv_py = ROOT / ".venv" / "bin" / "python"
+            if not venv_py.is_file():
+                self.skipTest("ladybug missing")
+            py = str(venv_py)
+        else:
+            py = sys.executable
+        proc = subprocess.run(
+            [py, str(ROOT / "bin" / "kb" / "index"), "--dry-run", "--json", "--corpus", str(tmp)],
+            cwd=ROOT,
+            capture_output=True,
+            text=True,
+            env=os.environ.copy(),
+            check=False,
+        )
+        self.assertEqual(proc.returncode, 0, proc.stderr)
+        msg = json.loads(proc.stdout)
+        self.assertTrue(msg.get("dry_run"))
+        self.assertGreaterEqual(msg.get("corpus_total", 0), 1)
+        self.assertFalse(lbug.exists(), "dry-run must not create a Ladybug file")
+
+    def test_facts_json_and_chats_land_on_rebuild(self) -> None:
+        """Gitea #18: facts (2-source) + chats markdown become leafs on rebuild."""
+        tmp = Path(tempfile.mkdtemp())
+        dbpath = tmp / "kb.lbug"
+        chats = tmp / "chats"
+        chats.mkdir()
+        (chats / "alice.md").write_text(
+            "# Chat\n\n## Alice and Bob\n\nhello from chats fixture unique-chat-token\n",
+            encoding="utf-8",
+        )
+        facts_path = tmp / "facts.json"
+        facts_path.write_text(json.dumps([{
+            "text": "container 'brain' unique-fact-token is running and declared in compose.yaml",
+            "source": "docker ps x compose.yaml",
+            "loc": "compose.yaml:brain",
+            "how": "facts/extract",
+        }]), encoding="utf-8")
+        venv_py = ROOT / ".venv" / "bin" / "python"
+        py = str(venv_py) if venv_py.is_file() else sys.executable
+        proc = subprocess.run(
+            [
+                py, str(ROOT / "bin" / "kb" / "index"),
+                "--rebuild", "--db", str(dbpath), "--no-defaults",
+                "--with-chats", str(chats),
+                "--facts-json", str(facts_path),
+                "--json",
+            ],
+            cwd=ROOT,
+            capture_output=True,
+            text=True,
+            env=os.environ.copy(),
+            check=False,
+        )
+        self.assertEqual(proc.returncode, 0, proc.stderr)
+        msg = json.loads(proc.stdout)
+        self.assertGreaterEqual(msg.get("facts_leafs", 0), 1)
+        self.assertGreaterEqual(msg.get("chat_leafs", 0), 1)
+        self.assertTrue(dbpath.exists())
+        sys.path.insert(0, str(ROOT / "bin" / "tools"))
+        import kblib
+        db, conn = kblib.connect(dbpath, read_only=True)
+        try:
+            stats = kblib.stats(conn)
+            self.assertGreaterEqual(stats["by_root"].get("facts", 0), 1)
+            fts = kblib.query_fts(conn, "unique-chat-token", 5)
+            self.assertTrue(fts, "chats markdown must be FTS-searchable")
+            fact_hits = kblib.query_fts(conn, "unique-fact-token", 5)
+            self.assertTrue(any(h.get("root") == "facts" for h in fact_hits))
+            src = conn.execute(
+                "MATCH (l:Leaf {root:'facts'}) RETURN l.source"
+            ).get_all()
+            self.assertTrue(any(" x " in str(r[0]) for r in src))
+        finally:
+            conn.close()
+            db.close()

bin/tools/test_kb_add.py

@@ -0,0 +1,69 @@
+"""Incremental add writes leafs without deleting kb.lbug."""
+from __future__ import annotations
+
+import json
+import os
+import subprocess
+import sys
+import tempfile
+import unittest
+from pathlib import Path
+
+ROOT = Path(__file__).resolve().parents[2]
+
+
+class KbAddCLITest(unittest.TestCase):
+    def test_json_add_does_not_delete_db(self) -> None:
+        tmp = Path(tempfile.mkdtemp())
+        dbpath = tmp / "kb.lbug"
+        py = sys.executable
+        venv_py = ROOT / ".venv" / "bin" / "python"
+        if venv_py.is_file():
+            py = str(venv_py)
+        payload = {
+            "text": "cli zebra leaf",
+            "root": "info",
+            "source": "cli-test",
+            "confidence": "confirmed",
+            "how": "test",
+            "loc": str(tmp),
+            "type": "reference",
+            "embedding": [0.0] * 256,
+        }
+        payload["embedding"][0] = 0.3
+        proc = subprocess.run(
+            [py, str(ROOT / "bin" / "kb" / "add"), "--db", str(dbpath), "--json"],
+            cwd=ROOT,
+            input=json.dumps(payload),
+            capture_output=True,
+            text=True,
+            env=os.environ.copy(),
+            check=False,
+        )
+        self.assertEqual(proc.returncode, 0, proc.stderr)
+        self.assertTrue(dbpath.exists(), "add must create the db, not skip write")
+        out = json.loads(proc.stdout)
+        self.assertEqual(out.get("mode"), "add")
+        self.assertEqual(len(out.get("ids") or []), 1)
+        again = subprocess.run(
+            [py, str(ROOT / "bin" / "kb" / "add"), "--db", str(dbpath), "--json"],
+            cwd=ROOT,
+            input=json.dumps({
+                **payload,
+                "text": "second moose leaf",
+                "source": "cli-test-2",
+            }),
+            capture_output=True,
+            text=True,
+            env=os.environ.copy(),
+            check=False,
+        )
+        self.assertEqual(again.returncode, 0, again.stderr)
+        self.assertTrue(dbpath.exists())
+        second = json.loads(again.stdout)
+        self.assertEqual(len(second.get("ids") or []), 1)
+        self.assertNotEqual(out["ids"][0], second["ids"][0])
+
+
+if __name__ == "__main__":
+    unittest.main()

bin/tools/test_kblib.py

@@ -71,6 +71,69 @@ class KblibTest(unittest.TestCase):
         self.assertTrue(hits)
         self.assertIn("Leaf_vec", kblib.leaf_index_names(self.conn))
 
+    def test_add_after_indexes_keeps_fts_queryable(self):
+        """Incremental add after FTS+HNSW must find the new leaf on both indexes."""
+        kblib.upsert_leaf(self.conn, text="seed fox leaf", root="info",
+                          confidence="confirmed", source="s", source_rev="r1",
+                          how="test", loc="/tmp", type_="reference",
+                          embedding=make_emb(0.1))
+        kblib.ensure_indexes(self.conn)
+        ids = kblib.add_leafs(self.conn, [{
+            "text": "added zebra after index",
+            "root": "facts",
+            "confidence": "confirmed",
+            "source": "a.md x b.md",
+            "source_rev": "r1",
+            "how": "test",
+            "loc": "/tmp",
+            "type": "fact",
+            "embedding": make_emb(0.9),
+        }])
+        self.assertEqual(len(ids), 1)
+        fts = kblib.query_fts(self.conn, "zebra", 5)
+        self.assertTrue(fts)
+        self.assertIn("zebra", fts[0]["text"])
+        self.assertEqual(fts[0]["root"], "facts")
+        vec = kblib.query_vector(self.conn, make_emb(0.9), 5)
+        self.assertTrue(any("zebra" in h["text"] for h in vec))
+        fox = kblib.query_fts(self.conn, "fox", 5)
+        self.assertTrue(fox)
+        self.assertIn("fox", fox[0]["text"])
+
+    def test_add_facts_and_info_one_transaction(self):
+        """D12: facts and info land in the same transaction."""
+        kblib.ensure_indexes(self.conn)
+        ids = kblib.add_leafs(self.conn, [
+            {
+                "text": "tx fact leaf two-source",
+                "root": "facts",
+                "confidence": "confirmed",
+                "source": "compose.yml x docker ps",
+                "source_rev": "r1",
+                "how": "test",
+                "loc": "/tmp",
+                "type": "fact",
+                "embedding": make_emb(0.4),
+            },
+            {
+                "text": "tx info narrative",
+                "root": "info",
+                "confidence": "confirmed",
+                "source": "note.md",
+                "source_rev": "r1",
+                "how": "test",
+                "loc": "/tmp",
+                "type": "reference",
+                "embedding": make_emb(0.5),
+            },
+        ])
+        self.assertEqual(len(ids), 2)
+        stats = kblib.stats(self.conn)
+        self.assertEqual(stats["by_root"].get("facts"), 1)
+        self.assertEqual(stats["by_root"].get("info"), 1)
+        self.assertTrue(kblib.query_fts(self.conn, "two-source", 5))
+        self.assertTrue(kblib.query_fts(self.conn, "narrative", 5))
+
     def test_drop_vector_then_create_raises_clear_error(self):
         """DROP INDEX leaves ghost catalog; create_fts_and_vector must raise."""
         kblib.upsert_leaf(self.conn, text="seed", root="info",
@@ -98,6 +161,39 @@ class KblibTest(unittest.TestCase):
         self.assertEqual(stats["total"], 2)
         self.assertEqual(stats["by_root"], {"facts": 1, "info": 1})
 
+    def test_hop_1_returns_file_hop_3_reaches_person(self):
+        """--hop walks FROM_FILE / HAS_VERSION / AUTHORED (Gitea #17)."""
+        import gitimport
+
+        lid = kblib.upsert_leaf(
+            self.conn, text="readme hop fixture", root="info",
+            confidence="confirmed", source="README.md", source_rev="r1",
+            how="test", loc="README.md", type_="reference",
+            embedding=make_emb(0.3),
+        )
+        kblib.link_from_file(self.conn, lid, "README.md", repo="sample-repo")
+        gitimport.index_commits(self.conn, [gitimport.Commit(
+            sha="a1b2c3d",
+            author="Ada Lovelace",
+            email="ada@example.com",
+            date="2026-08-10T12:00:00Z",
+            subject="feat: first commit",
+            files=["README.md"],
+        )], "sample-repo")
+        hop1 = kblib.hop_walk(self.conn, lid, 1)
+        self.assertEqual(len(hop1), 1)
+        self.assertEqual(hop1[0]["label"], "File")
+        self.assertEqual(hop1[0]["name"], "README.md")
+        self.assertEqual(hop1[0]["depth"], 1)
+        hop3 = kblib.hop_walk(self.conn, lid, 3)
+        labels = {n["label"] for n in hop3}
+        self.assertIn("File", labels)
+        self.assertIn("Commit", labels)
+        self.assertIn("Person", labels)
+        person = [n for n in hop3 if n["label"] == "Person"][0]
+        self.assertEqual(person["name"], "Ada Lovelace")
+        self.assertEqual(person["depth"], 3)
+
 
 if __name__ == "__main__":
     unittest.main()

bin/tools/test_mailconv.py

@@ -8,10 +8,13 @@ from pathlib import Path
 sys.path.insert(0, os.path.dirname(__file__))
 
 from mailconv import (  # noqa: E402
+    TESS_LANG,
     clean_email_address,
+    convert_pdf,
     html_to_markdown,
     is_convertible,
     normalize_markdown,
+    ocr_image,
     split_zip_members,
     subject_to_filename,
     zip_extract_safe,
@@ -100,6 +103,92 @@ class TestMailConv(unittest.TestCase):
         self.assertFalse(is_convertible(".exe"))
         self.assertFalse(is_convertible(".unknown"))
 
+    def test_convert_pdf_prefers_pdftotext(self):
+        import mailconv as mc
+
+        calls: list[list[str]] = []
+
+        def fake_run(cmd, **kwargs):
+            calls.append(list(cmd))
+
+            class P:
+                returncode = 0
+                stdout = b"Invoice BM25 layout"
+                stderr = b""
+
+            return P()
+
+        self._patch_run(mc, fake_run)
+        out = convert_pdf(Path(self._tmp("born.pdf")))
+        self.assertIn("BM25", out)
+        self.assertEqual(calls[0][:2], ["pdftotext", "-layout"])
+        self.assertFalse(any(c[0] == "tesseract" for c in calls))
+        self.assertFalse(any(c[0] == "pdftoppm" for c in calls))
+
+    def test_convert_pdf_empty_layer_uses_pdftoppm_tesseract(self):
+        import mailconv as mc
+
+        calls: list[list[str]] = []
+
+        def fake_run(cmd, **kwargs):
+            calls.append(list(cmd))
+
+            class P:
+                returncode = 0
+                stdout = b""
+                stderr = b""
+
+            if cmd[0] == "pdftotext":
+                P.stdout = b"  \n"
+                return P()
+            if cmd[0] == "pdftoppm":
+                prefix = Path(cmd[-1])
+                (prefix.parent / "page-1.png").write_bytes(b"fake")
+                return P()
+            if cmd[0] == "tesseract":
+                P.stdout = b"scanned HELLO"
+                return P()
+            return P()
+
+        self._patch_run(mc, fake_run)
+        out = convert_pdf(Path(self._tmp("scan.pdf")))
+        self.assertIn("HELLO", out)
+        bins = [c[0] for c in calls]
+        self.assertIn("pdftotext", bins)
+        self.assertIn("pdftoppm", bins)
+        self.assertIn("tesseract", bins)
+        tess = next(c for c in calls if c[0] == "tesseract")
+        self.assertIn(TESS_LANG, tess)
+        self.assertNotIn("docling", " ".join(bins))
+
+    def test_ocr_image_paddle_engine(self):
+        import mailconv as mc
+
+        calls: list[list[str]] = []
+
+        def fake_run(cmd, **kwargs):
+            calls.append(list(cmd))
+
+            class P:
+                returncode = 0
+                stdout = b"paddle text"
+                stderr = b""
+
+            return P()
+
+        self._patch_run(mc, fake_run)
+        os.environ["OCR_ENGINE"] = "paddle"
+        try:
+            out = ocr_image(Path(self._tmp("x.png")))
+        finally:
+            os.environ.pop("OCR_ENGINE", None)
+        self.assertEqual(out, "paddle text")
+        self.assertEqual(calls[0][:2], ["paddleocr", "ocr"])
+
+    def _patch_run(self, mod, fn) -> None:
+        self.addCleanup(setattr, mod.subprocess, "run", mod.subprocess.run)
+        mod.subprocess.run = fn
+
     def _mk_zip(self, members):
         zpath = Path(self._tmp("arc.zip"))
         with zipfile.ZipFile(zpath, "w") as zf:

bin/tools/test_published_docs.py

@@ -1,7 +1,6 @@
-"""Published docs must match live commands (Gitea SoT, brain/search, no fake --hop)."""
+"""Published docs must match live commands (Gitea SoT, brain/search)."""
 from __future__ import annotations
 
-import re
 import unittest
 from pathlib import Path
 
@@ -59,6 +58,18 @@ class PublishedDocsTest(unittest.TestCase):
         self.assertNotIn("password", settings.lower())
         self.assertIn("json", settings)
 
+    def test_picoclaw_compose_profile_has_mcp_example(self) -> None:
+        compose = (ROOT / "compose.yaml").read_text()
+        self.assertIn('profiles: ["picoclaw"]', compose)
+        self.assertIn("127.0.0.1:8630", compose)
+        example = (ROOT / "deploy" / "picoclaw" / "mcp.json.example").read_text()
+        self.assertIn("127.0.0.1:8630/mcp", example)
+        self.assertNotIn("password", example.lower())
+        self.assertNotIn("token", example.lower())
+        docs = (ROOT / "docs" / "picoclaw.md").read_text()
+        self.assertIn("search", docs)
+        self.assertIn("throttled", docs)
+
     def test_readme_read_path_is_go(self) -> None:
         plan = (ROOT / "PLAN.md").read_text()
         self.assertIn("get.go", plan)
@@ -74,6 +85,51 @@ class PublishedDocsTest(unittest.TestCase):
         self.assertIn("/mcp", (ROOT / "README.md").read_text())
         skill = (ROOT / "skills" / "brain" / "SKILL.md").read_text()
         self.assertIn("/mcp", skill)
+        self.assertFalse((ROOT / "skills" / "db-yaml").exists())
+        self.assertTrue((ROOT / "skills" / "postgres" / "SKILL.md").is_file())
+
+    def test_cgo_zig_and_index_profile(self) -> None:
+        plan = (ROOT / "PLAN.md").read_text()
+        self.assertIn("D21", plan)
+        self.assertIn("zig cc", plan)
+        dockerfile = (ROOT / "Dockerfile").read_text()
+        self.assertIn("bin/cgo/zcc", dockerfile)
+        self.assertIn("FROM debian:bookworm-slim AS api", dockerfile)
+        self.assertIn("FROM python:3.12-slim AS index", dockerfile)
+        api = dockerfile[dockerfile.index("FROM debian:bookworm-slim AS api") :]
+        self.assertNotIn("pip install", api)
+        compose = (ROOT / "compose.yaml").read_text()
+        self.assertIn('profiles: ["index"]', compose)
+        self.assertIn("target: api", compose)
+
+    def test_reasoner_docs_name_real_hf_ids_cpu_sidecar(self) -> None:
+        docs = (ROOT / "docs" / "reasoner.md").read_text()
+        for hf in (
+            "Qwen/Qwen3.5-9B",
+            "Qwen/Qwen3.6-27B",
+            "prism-ml/Bonsai-27B-gguf",
+        ):
+            self.assertIn(hf, docs)
+        self.assertIn("no official qwen3.6-9b", docs.lower())
+        self.assertIn("OLLAMA_NUM_GPU", docs)
+        self.assertIn("rss_mb", docs)
+        self.assertIn("vram_mb", docs)
+        self.assertIn("3/3", docs)
+        self.assertIn("Do not claim 9B is better at tools", docs)
+        self.assertNotIn("Qwen/Qwen3.6-9B", docs)
+        plan = (ROOT / "PLAN.md").read_text()
+        self.assertIn("D18", plan)
+        self.assertIn("Qwen/Qwen3.5-9B", plan)
+        compose = (ROOT / "compose.yaml").read_text()
+        self.assertIn('profiles: ["reasoner"]', compose)
+        self.assertIn("OLLAMA_NUM_GPU", compose)
+        self.assertIn("127.0.0.1:11435", compose)
+        dockerfile = (ROOT / "Dockerfile").read_text()
+        self.assertNotIn(".gguf", dockerfile.lower())
+        self.assertNotIn(".safetensors", dockerfile.lower())
+        api = dockerfile[dockerfile.index("FROM debian:bookworm-slim AS api") :]
+        self.assertNotIn("COPY models", api)
+        self.assertNotIn("qwen", api.lower())
 
     def test_readme_search_escalates_web(self) -> None:
         text = (ROOT / "README.md").read_text()
@@ -82,19 +138,51 @@ class PublishedDocsTest(unittest.TestCase):
         skill = (ROOT / "skills" / "brain" / "SKILL.md").read_text()
         self.assertIn("`web` block", skill)
 
-    def test_docs_do_not_claim_hop_walks(self) -> None:
+    def test_docs_say_hop_walks_from_file(self) -> None:
         paths = [
             ROOT / "README.md",
             ROOT / "docs" / "design.md",
             ROOT / "skills" / "brain" / "SKILL.md",
-            ROOT / "skills" / "diataxis-docs" / "SKILL.md",
+            ROOT / "docs" / "runbook.md",
+            ROOT / "docs" / "README.md",
         ]
-        # Command-style `--hop 1` / `--hop N` plus follow/walk = the old lie.
-        # Honest "not implemented" notes must not match.
-        lie = re.compile(r"--hop (?:N|1).*(?:follow|walk)", re.I | re.S)
         for path in paths:
             text = path.read_text()
-            self.assertIsNone(
+            self.assertIn("--hop", text, f"{path.relative_to(ROOT)} must document --hop")
-                lie.search(text),
+            self.assertNotIn(
-                f"{path.relative_to(ROOT)} still claims --hop walks the graph",
+                "not implemented",
+                text.lower(),
+                f"{path.relative_to(ROOT)} still says hop is not implemented",
             )
+
+    def test_docs_are_portable_diataxis(self) -> None:
+        index = (ROOT / "docs" / "README.md").read_text()
+        self.assertIn("type: reference", index)
+        for d in ("D3", "D6", "D14", "D15", "D17", "D18"):
+            self.assertIn(d, index)
+        runbook = (ROOT / "docs" / "runbook.md").read_text()
+        self.assertIn("type: howto", runbook)
+        self.assertIn("bin/brain/search.go", runbook)
+        self.assertIn("bin/brain/index.go", runbook)
+        self.assertNotIn("search.ops.io", runbook)
+        self.assertNotIn("/mnt/", runbook)
+        self.assertNotIn("/home/", runbook)
+        readme = (ROOT / "README.md").read_text()
+        self.assertIn("docs/runbook.md", readme)
+        self.assertNotIn("search.ops.io", readme)
+
+    def test_v1_epic_is_named_in_docs(self) -> None:
+        plan = (ROOT / "PLAN.md").read_text()
+        self.assertIn("Gap to v1", plan)
+        self.assertIn("eSlider/2dph/issues/16", plan)
+        self.assertIn("eSlider/2dph/issues/17", plan)
+        self.assertIn("eSlider/2dph/milestone/12", plan)
+        road = (ROOT / "docs" / "roadmap.md").read_text()
+        self.assertIn("type: explanation", road)
+        self.assertIn("issues/16", road)
+        self.assertIn("issues/14", road)
+        index = (ROOT / "docs" / "README.md").read_text()
+        self.assertIn("roadmap.md", index)
+        self.assertIn("epic #16", index)
+        agents = (ROOT / "AGENTS.md").read_text()
+        self.assertIn("roadmap.md", agents)

bin/tools/test_skills.py

@@ -0,0 +1,49 @@
+"""Skills must name live commands; every bin/ path in SKILL.md must exist."""
+from __future__ import annotations
+
+import re
+import unittest
+from pathlib import Path
+
+ROOT = Path(__file__).resolve().parents[2]
+BIN_PATH = re.compile(r"(bin/[A-Za-z0-9_./-]+)")
+
+
+class SkillsTest(unittest.TestCase):
+    def test_db_yaml_renamed_to_postgres(self) -> None:
+        self.assertFalse(
+            (ROOT / "skills" / "db-yaml").exists(),
+            "skills/db-yaml must be skills/postgres",
+        )
+        self.assertTrue((ROOT / "skills" / "postgres" / "SKILL.md").is_file())
+        text = (ROOT / "skills" / "postgres" / "SKILL.md").read_text()
+        self.assertIn("bin/postgres/query.go", text)
+        self.assertNotIn("search.ops.io", text)
+
+    def test_every_bin_path_in_skills_exists(self) -> None:
+        missing: list[str] = []
+        for path in (ROOT / "skills").rglob("SKILL.md"):
+            text = path.read_text()
+            for m in BIN_PATH.finditer(text):
+                rel = m.group(1).rstrip(")`.,;")
+                candidate = ROOT / rel
+                if not candidate.exists():
+                    missing.append(f"{path.relative_to(ROOT)}: {rel}")
+        self.assertEqual(missing, [], "skill bin paths must exist")
+
+    def test_brain_skill_lists_generated_tools(self) -> None:
+        tools = (ROOT / "skills" / "brain" / "tools.md").read_text()
+        skill = (ROOT / "skills" / "brain" / "SKILL.md").read_text()
+        self.assertIn("tools.md", skill)
+        for name in ("search", "get", "stats", "audit"):
+            self.assertIn(f"`{name}`", tools)
+
+    def test_picoclaw_lists_tool_order(self) -> None:
+        skill = (ROOT / "skills" / "picoclaw" / "SKILL.md").read_text()
+        agents = (ROOT / "AGENTS.md").read_text()
+        self.assertIn("**`search`**", skill)
+        self.assertIn("**`get`**", skill)
+        self.assertIn("**`audit`**", skill)
+        self.assertIn("throttled", skill.lower())
+        self.assertIn("not a negative finding", agents)
+        self.assertIn("Fact-check every", agents)

compose.yaml

@@ -1,66 +1,88 @@
 # 2dph — docker composition
 #
-#   docker compose run --rm brain index                     # rebuild graph
+#   docker compose up -d brain                  # API (Zig CGO serve)
-#   docker compose run --rm brain search "Matrix fed"       # one-shot query
+#   docker compose --profile index run --rm index   # Python rebuild
-#   docker compose run --rm brain serve                     # async Go server
+#   docker compose --profile picoclaw up brain-mcp
-#   docker compose up brain-watch                           # auto re-index
+#   docker compose --profile reasoner up -d reasoner   # CPU Ollama :11435
+#   docker compose --profile searxng up -d
+#   OCR_ENGINE=paddle docker compose --profile ocr-paddle run --rm ocr-paddle
 #
-# Caching: the 128M model (HF_HOME) and kb.lbug (VAR_DIR) live in named
+# Secrets never baked in: search.env + db-profiles.yml from ~/.config/brain.
-# volumes, so rebuilds never redownload the model or re-derive the graph.
-# Secrets are never baked into the image: search.env + db-profiles.yml mount
-# read-only from ~/.config/brain.
 
 name: 2dph
 
 services:
   brain:
-    image: ghcr.io/eslider/2dph:latest
+    image: ghcr.io/eslider/2dph:api
     build:
       context: .
       dockerfile: Dockerfile
+      target: api
       cache_from:
         - ghcr.io/eslider/2dph:cache
-    command: ["brain", "search", "help"]
+    command: ["serve"]
     environment: &env
       HF_HOME: /data/hf
       BRAIN_SEARCH_CACHE: /data/cache/web-search.sqlite
       BRAIN_DB_PROFILES: /secret/db-profiles.yml
       BRAIN_SEARCH_ENV: /secret/search.env
-      KB_SEARCH_CMD: /app/bin/kb/search
+      KB_ROOT: /data
       KB_WORKERS: "4"
+      KB_PORT: "8630"
     volumes:
       - kb-model:/data/hf
       - kb-var:/data
-      # corpus is read-only on the host, never written from the container
-      - ..:/corpus:ro
       - ~/.config/brain:/secret:ro
+    ports:
+      - "127.0.0.1:8630:8630"
     read_only: true
     tmpfs:
       - /tmp
     healthcheck:
-      test: ["CMD", "python3", "-c", "import ladybug, model2vec, mistune; print('ok')"]
+      test: ["CMD", "wget", "-qO-", "http://127.0.0.1:8630/health"]
       interval: 30s
       timeout: 5s
       retries: 3
     restart: unless-stopped
     stop_grace_period: 20s
 
-  # watcher: re-index on corpus file change (watchdog script)
   brain-watch:
-    image: ghcr.io/eslider/2dph:latest
+    image: ghcr.io/eslider/2dph:api
     environment: *env
     volumes:
       - kb-model:/data/hf
       - kb-var:/data
       - ..:/corpus:ro
       - ~/.config/brain:/secret:ro
-    command: ["brain", "watch", "/corpus"]
+    command: ["watch", "/corpus"]
     read_only: true
     tmpfs:
       - /tmp
     restart: unless-stopped
     stop_grace_period: 20s
 
+  # Python write path (Ladybug rebuild). Not in the API image.
+  #   docker compose --profile index run --rm index
+  index:
+    profiles: ["index"]
+    image: ghcr.io/eslider/2dph:index
+    build:
+      context: .
+      dockerfile: Dockerfile
+      target: index
+    environment:
+      HF_HOME: /data/hf
+      KB_PY: python3
+    volumes:
+      - kb-model:/data/hf
+      - kb-var:/app/var
+      - ..:/corpus:ro
+      - ~/.config/brain:/secret:ro
+    command: ["index"]
+    read_only: true
+    tmpfs:
+      - /tmp
+
   # Optional local SearXNG (D3). Skip if BRAIN_SEARCH_URL already points at a
   # live instance — do not run a second copy on that host.
   #   SEARXNG_SECRET=$(openssl rand -hex 32) docker compose --profile searxng up -d
@@ -76,6 +98,50 @@ services:
       - ./deploy/searxng/limiter.toml:/etc/searxng/limiter.toml:ro
     restart: unless-stopped
 
+  # MCP endpoint for an external agent (PicoClaw is not shipped here).
+  #   docker compose --profile picoclaw up brain-mcp
+  brain-mcp:
+    profiles: ["picoclaw"]
+    image: ghcr.io/eslider/2dph:api
+    environment: *env
+    volumes:
+      - kb-model:/data/hf
+      - kb-var:/data
+      - ~/.config/brain:/secret:ro
+    command: ["serve"]
+    ports:
+      - "127.0.0.1:8630:8630"
+    read_only: true
+    tmpfs:
+      - /tmp
+    restart: unless-stopped
+
+  # CPU OpenAI-compatible sidecar (D18). Weights are pulled at runtime, not
+  # baked into the 2dph image. Does not touch host Ollama on :11434.
+  #   docker compose --profile reasoner up -d reasoner
+  #   docker compose --profile reasoner exec reasoner ollama pull qwen3.5:9b
+  reasoner:
+    profiles: ["reasoner"]
+    image: docker.io/ollama/ollama:latest
+    environment:
+      OLLAMA_NUM_GPU: "0"
+      OLLAMA_HOST: "0.0.0.0:11434"
+    ports:
+      - "127.0.0.1:11435:11434"
+    volumes:
+      - reasoner-ollama:/root/.ollama
+    restart: unless-stopped
+
+  # Optional PP-OCRv5 (not default). Default OCR is tesseract eng+deu.
+  #   OCR_ENGINE=paddle docker compose --profile ocr-paddle run --rm ocr-paddle
+  ocr-paddle:
+    profiles: ["ocr-paddle"]
+    image: python:3.12-slim
+    environment:
+      OCR_ENGINE: paddle
+    command: ["python", "-c", "print('OCR_ENGINE=paddle; install paddleocr on PATH')"]
+
 volumes:
   kb-model:
   kb-var:
+  reasoner-ollama:

deploy/picoclaw/mcp.json.example

@@ -0,0 +1,8 @@
+{
+  "mcpServers": {
+    "2dph": {
+      "url": "http://127.0.0.1:8630/mcp",
+      "description": "2dph fact gate. Tool order: search → get → audit. throttled is not absence."
+    }
+  }
+}

docs/README.md

@@ -1,15 +1,38 @@
-# 2dph (deductionphile)
+---
+type: reference
+status: current
+related:
+  - docs/runbook.md
+  - docs/design.md
+  - PLAN.md
+  - docs/roadmap.md
+---
 
-Evidence-first knowledge graph + hybrid RAG over the operational
+# 2dph docs (Diataxis)
-Brain/ops/eSlider stack. Facts need proof or they are
+
+Evidence-first knowledge graph. Facts need proof or they are
 `(not confirmed)`.
 
-- [PLAN.md](../PLAN.md) — decisions, execution order, open questions (v2)
+| Type | Doc |
-- [design](design.md) — schema, deduction model, sources
+|------|-----|
-- [Gitea issues](https://git.produktor.io/eSlider/2dph/issues) — work board (origin)
+| tutorial / howto | [runbook](runbook.md) — run anywhere (uv, Go, Docker) |
+| explanation | [design](design.md) — two roots, deduction, D17/D20/D18 |
+| explanation | [roadmap](roadmap.md) — gap to v1 (epic #16) |
+| howto | [picoclaw](picoclaw.md) — MCP agent profile |
+| howto | [reasoner](reasoner.md) — CPU bake-off (D18) |
+| reference | [PLAN.md](../PLAN.md) — decisions D1–D21 |
+
+Decisions the public face must name: **D3** SearXNG compose, **D6** Go service /
+Python write sidecar, **D14** `bin/{subject}/{method}.go`, **D15** Gitea origin,
+**D17** assertion gate (facts → info → web), **D18** pluggable reasoner.
 
 Search: `bin/brain/search.go "query"` (HTTP: `bin/brain/serve.go` —
-`/health` `/search` `/get` `/stats` `/audit` `/ingest`). `--hop` is
+`/health` `/search` `/get` `/stats` `/audit` `/ingest`). `--hop N` walks
-not a walk; the flag errors until File/FROM_FILE edges exist.
+`FROM_FILE` → Commit → Person from each hit (max 3). Rebuild writes
+File edges ([#17](https://git.produktor.io/eSlider/2dph/issues/17)).
 
-Published docs live here and mirror the project state.
+Work board: [Gitea issues](https://git.produktor.io/eSlider/2dph/issues)
+([epic #16](https://git.produktor.io/eSlider/2dph/issues/16)).
+PRs and CI: GitHub [`eSlider/2dph`](https://github.com/eSlider/2dph).
+
+Published docs live here and match live commands.

docs/chat-import-plan.md

@@ -21,4 +21,5 @@ OO_CLI                     (default: $HOME/go/bin/oo)
 ./bin/chats/apply.go --dry-run
 ```
 
-JSONL → markdown only. Brain ingest is `bin/brain/index.go` (not a `chats index`).
+JSONL → markdown only. Brain ingest is `bin/brain/index.go --with-chats`
+(default `var/chats/md`). WhatsApp sync is out of v1.

docs/design.md

@@ -1,3 +1,12 @@
+---
+type: explanation
+status: current
+related:
+  - docs/README.md
+  - docs/runbook.md
+  - docs/roadmap.md
+---
+
 # Design — facts, info, deduction
 
 ## Two roots, one transaction
@@ -25,8 +34,9 @@ bin/brain/search.go "question"
      is not evidence of absence; `--no-web` / `--root` skip it)
 ```
 
-`--hop` is not implemented yet (needs File/FROM_FILE edges). The flag is an
+`--hop N` walks `Leaf-[:FROM_FILE]->File-[:HAS_VERSION]->Commit-[:AUTHORED]->Person`
-error; it is not a graph walk.
+from each hit (1=File, 2=Commit, 3=Person). Rebuild writes FROM_FILE;
+git import writes HAS_VERSION/AUTHORED ([#17](https://git.produktor.io/eSlider/2dph/issues/17)).
 
 ## Who / What / How / Where / When + evidence
 
@@ -66,14 +76,29 @@ Conflicting pairings (≥2 yes vs ≥2 no) = hypothesis (OQ1 → v2 resolution).
 ## Read path
 
 `bin/brain/get.go`, `stats.go`, and `eval.go` call `internal/brain` with cgo
-(`system_ladybug`). They do not exec Python. Control questions for recall@5
+(`system_ladybug`), compiled by **Zig** (`bin/cgo/zcc`, D21), not gcc.
-live in `internal/brain/rank` so CI can test the table without libladybug.
+They do not exec Python. Control questions for recall@5 live in
+`internal/brain/rank` so CI can test the table without libladybug.
 Python `bin/kb/{get,stats,eval}` remain for GitHub Actions until the runner
-has ladybug cgo. Index/write is still `bin/kb/index`.
+fetches Zig + libs (`bin/cgo/zig`). Incremental write is `bin/kb/add`
+(`bin/brain/add.go`). Bulk index/write is still `bin/kb/index`
+(`docker compose --profile index`).
 
 ## Agent API (D20)
 
 `bin/brain/serve.go` exposes the same `internal/httpapi.Ops` table as OpenAPI
 (`GET /openapi.json`) and MCP (`POST /mcp` JSON-RPC `tools/list` +
-`tools/call`). Tool names match paths: `search`, `get`, `stats`, `audit`.
+`tools/call`). Tool names match paths: `search`, `get`, `stats`, `audit`,
+`ingest` (add a leaf; omit body for the CLI hint).
 Agents should use these endpoints instead of shebang CLIs.
+
+## Reasoner (D18)
+
+Pluggable OpenAI-compatible URL. RAM: `Qwen/Qwen3.5-9B`. Quality:
+`prism-ml/Bonsai-27B-gguf` or `Qwen/Qwen3.6-27B`. No official Qwen3.6-9B.
+CPU sidecar: compose profile `reasoner` (`OLLAMA_NUM_GPU=0`,
+`127.0.0.1:11435`). Bake-off: `bin/reasoner/bakeoff.go`. Weights stay out
+of the 2dph image. See [docs/reasoner.md](reasoner.md).
+
+Gap to v1 (hops, corpus, CI eval): [roadmap](roadmap.md),
+[epic #16](https://git.produktor.io/eSlider/2dph/issues/16).

docs/picoclaw.md

@@ -0,0 +1,18 @@
+# PicoClaw profile (reference agent)
+
+2dph is the memory/fact gate. PicoClaw (or any MCP client) is the agent loop
+and is **not** shipped in this repo.
+
+```bash
+docker compose --profile picoclaw up brain-mcp
+```
+
+The API listens on `127.0.0.1:8630`. Point the agent at
+`http://127.0.0.1:8630/mcp` using [deploy/picoclaw/mcp.json.example](../deploy/picoclaw/mcp.json.example).
+
+OpenAPI: `GET http://127.0.0.1:8630/openapi.json`.
+
+Before a factual reply: `search` → `get` → `audit`. `throttled` is not a
+negative finding. See `skills/picoclaw/SKILL.md`.
+
+No Cursor required. A live PicoClaw binary/image is an operator choice.

docs/reasoner.md

@@ -0,0 +1,69 @@
+# Reasoner bake-off (D18)
+
+Pluggable OpenAI-compatible URL. 2dph does not ship weights. PicoClaw is
+not in this repo; the bake-off hits the same tool names PicoClaw would
+(`search` → `get` → `audit` from `internal/httpapi.Ops`).
+
+```bash
+docker compose --profile reasoner up -d reasoner
+docker compose --profile reasoner exec reasoner ollama pull qwen3.5:9b
+REASONER_BASE_URL=http://127.0.0.1:11435/v1 REASONER_MODEL=qwen3.5:9b \
+  ./bin/reasoner/bakeoff.go --json
+```
+
+Host Ollama on `:11434` is left alone. This sidecar binds `127.0.0.1:11435`
+with `OLLAMA_NUM_GPU=0` (CPU). Measure RSS (`/api/ps` `size`), not VRAM.
+
+If Compose cannot allocate a project network (Docker IPAM pool exhausted),
+the same sidecar is:
+
+```bash
+docker run -d --name 2dph-reasoner \
+  -e OLLAMA_NUM_GPU=0 \
+  -p 127.0.0.1:11435:11434 \
+  -v 2dph-reasoner-ollama:/root/.ollama \
+  ollama/ollama:latest
+```
+
+## Real Hugging Face ids
+
+| Role | HF id | Ollama tag (this bake-off) |
+|------|-------|----------------------------|
+| RAM / 9B | `Qwen/Qwen3.5-9B` | `qwen3.5:9b` |
+| Quality 27B (CPU) | `prism-ml/Bonsai-27B-gguf` (derived from Qwen3.6-27B) | `MichelRosselli/bonsai-27b:Q1_0` |
+| Quality 27B (full) | `Qwen/Qwen3.6-27B` | not pulled on this CPU box |
+
+There is **no official Qwen3.6-9B**. Do not invent that id.
+
+Qwen3.5-9B has documented upstream tool-call XML bugs. A 9B win on tools
+is only claimed if this bake-off records OpenAI `tool_calls` (not
+`<tool_call>` XML in `content`).
+
+The 2dph API image does not `COPY` GGUF/safetensors. Pull at runtime into
+the `reasoner-ollama` volume.
+
+## Live CPU run
+
+Host sidecar: Ollama **0.32.9**, `OLLAMA_NUM_GPU=0`, `127.0.0.1:11435`,
+`device: cpu`, `vram_mb: 0`. Date: 2026-08-13. Same three prompts
+(`search` / `get` / `audit`). PicoClaw binary was not used; the OpenAI
+tools payload is the surface it would send.
+
+| Model | HF id | tool_call | xml_leak | rss_mb | latency_ms (search/get/audit) |
+|-------|-------|-----------|----------|--------|-------------------------------|
+| `qwen3.5:9b` | `Qwen/Qwen3.5-9B` | 3/3 | 0 | 5790 | 50059 / 51678 / 31980 |
+| `MichelRosselli/bonsai-27b:Q1_0` | `prism-ml/Bonsai-27B-gguf` | 3/3 | 0 | 21951 | 327702 / 166830 / 119808 |
+| `Qwen/Qwen3.6-27B` | `Qwen/Qwen3.6-27B` | not loaded | — | — | too heavy for this CPU box |
+
+Both loaded models emitted OpenAI `tool_calls` (not `<tool_call>` XML) on
+this runtime. **Do not claim 9B is better at tools** — the score is tied
+at 3/3. 9B is smaller and faster. Bonsai RSS includes weights + KV
+(`size` from `/api/ps`); first Bonsai prompt includes cold load.
+
+Re-run:
+
+```bash
+REASONER_BASE_URL=http://127.0.0.1:11435/v1 REASONER_MODEL=qwen3.5:9b \
+  ./bin/reasoner/bakeoff.go --json
+REASONER_MODEL=MichelRosselli/bonsai-27b:Q1_0 ./bin/reasoner/bakeoff.go --json
+```

docs/roadmap.md

@@ -0,0 +1,66 @@
+---
+type: explanation
+status: current
+related:
+  - PLAN.md
+  - docs/design.md
+  - docs/runbook.md
+---
+
+# Gap to v1 — detective brain
+
+Goal: a brain that does not assert without proof. Search is deduction
+(`facts` ≥2 sources → `info` → `web`). `confirmed` only from the facts root.
+
+**v1 is a living graph the agent can write and walk**, not “more RAG”.
+
+Epic: [Gitea #16](https://git.produktor.io/eSlider/2dph/issues/16).
+Milestone: [v1 detective brain](https://git.produktor.io/eSlider/2dph/milestone/12).
+Decisions: [PLAN.md](../PLAN.md).
+
+## In (do not reopen)
+
+Read path Go + Zig CGO (D21). HTTP + OpenAPI + MCP (D20). PicoClaw compose
+profile + CPU reasoner (D18). Mail sync → import → rebuild. D14 shebangs.
+Compose `api` (no CPython) / `index` (Python write). Issues #1–#5, #7–#13.
+[#15](https://git.produktor.io/eSlider/2dph/issues/15) lever/loop.
+[#14](https://git.produktor.io/eSlider/2dph/issues/14) `bin/brain/add.go` /
+`POST /ingest` (Python `kblib.add_leafs`; no Go upsert port).
+[#17](https://git.produktor.io/eSlider/2dph/issues/17) `--hop N` walks
+FROM_FILE / HAS_VERSION / AUTHORED.
+[#18](https://git.produktor.io/eSlider/2dph/issues/18) `--with-facts` /
+`--with-chats` on rebuild (WhatsApp out of v1).
+[#19](https://git.produktor.io/eSlider/2dph/issues/19) CI recall SoT =
+`bin/brain/eval.go` via Zig.
+Epic [#16](https://git.produktor.io/eSlider/2dph/issues/16) closed.
+
+## v2
+
+[#6](https://git.produktor.io/eSlider/2dph/issues/6) OCR — `pdftotext` then
+`pdftoppm` + tesseract `eng+deu`. Optional `ocr-paddle`.
+[#29](https://git.produktor.io/eSlider/2dph/issues/29) OQ1 contradiction
+resolution. [#30](https://git.produktor.io/eSlider/2dph/issues/30) OQ3 duckdb-md.
+
+## Blockers
+
+None for epic #16 (closed). Remaining v2: OQ1, OQ3, OQ4.
+
+```
+question
+    │
+    ├─ FTS + HNSW          ← in
+    ├─ facts / info roots  ← in
+    ├─ web (D17)           ← in
+    ├─ brain/add ACID      ← in
+    ├─ Cypher hop          ← in
+    └─ facts+chats corpus  ← in
+```
+
+## Not v1
+
+OQ1 contradiction resolution, OQ3 duckdb-md export, OQ4 YAML-first leafs.
+OCR (OQ2) is in: tesseract, not docling.
+
+## Close epic #16 when
+
+Children #14, #15, #17, #18, #19 are closed. MCP tool order stays gated by tests.

docs/runbook.md

@@ -0,0 +1,85 @@
+---
+type: howto
+status: current
+related:
+  - docs/README.md
+  - PLAN.md
+---
+
+# Run 2dph (portable)
+
+No laptop-absolute paths. Config lives in env files under `$HOME/.config/brain/`
+(mode 0600), not in git.
+
+## Toolchain
+
+- Go (see `go.mod`)
+- Python 3.12 + [uv](https://docs.astral.sh/uv)
+- Optional: Docker, Zig CGO via `bin/cgo/zig` (not gcc)
+- Optional: poppler (`pdftotext`/`pdftoppm`) + tesseract `eng+deu` for mail OCR
+
+```bash
+uv venv .venv
+uv pip install -r requirements.lock.txt
+eval "$(bin/cgo/zig env)"   # when compiling Ladybug read tools
+go test ./...
+uv run python -m unittest discover -s bin/tools -t .
+```
+
+## Config
+
+| File / env | Purpose |
+|------------|---------|
+| `$BRAIN_SEARCH_ENV` (default `$HOME/.config/brain/search.env`) | `BRAIN_SEARCH_URL` (SearXNG). Optional Basic Auth. |
+| `$HOME/.config/brain/db-profiles.yml` | read-only Postgres profiles (OnlyOffice via tunnel) |
+
+If the host already runs SearXNG, point `BRAIN_SEARCH_URL` at it. Do not start
+a second copy (D3). Optional Compose instance:
+
+```bash
+SEARXNG_SECRET=$(openssl rand -hex 32) docker compose --profile searxng up -d
+```
+
+That binds `127.0.0.1:8888`. JSON format must stay enabled.
+
+## Index then search
+
+Write path is `bin/brain/add.go` for a leaf (or `POST /ingest`). Bulk
+corpus rebuild remains `bin/brain/index.go --rebuild` (Compose profile
+`index`). Do not DROP INDEX on Ladybug 0.19.
+
+```bash
+bin/brain/add.go --text "arc-1 runs Matrix" --root facts --source "compose.yml x docker ps"
+bin/brain/index.go --rebuild --with-facts --with-chats
+bin/brain/search.go "LadybugDB vector index"     # facts → info → web (D17)
+bin/brain/search.go "upstream flag" --no-web
+bin/brain/get.go <id> --body
+bin/brain/stats.go
+```
+
+`--hop N` walks File → Commit → Person from each hit. Empty web results are `throttled`, not absence.
+Gap to v1: [roadmap](roadmap.md) / [epic #16](https://git.produktor.io/eSlider/2dph/issues/16).
+
+Ladybug 0.19: never `DROP INDEX` FTS/VECTOR (ghost catalog). Fresh indexes =
+delete `var/kb.lbug` then `--rebuild`.
+
+## HTTP / MCP
+
+```bash
+docker compose up -d brain                       # :8630 Zig CGO serve
+docker compose --profile index run --rm index    # rebuild
+docker compose --profile picoclaw up brain-mcp   # MCP 127.0.0.1:8630
+```
+
+`GET /openapi.json`, `POST /mcp`. Agent tool order: `search` → `get` → `audit`.
+
+## Reasoner (optional, D18)
+
+CPU sidecar on `127.0.0.1:11435`. Weights are not in the 2dph image.
+
+```bash
+docker compose --profile reasoner up -d reasoner
+REASONER_BASE_URL=http://127.0.0.1:11435/v1 ./bin/reasoner/bakeoff.go --json
+```
+
+See [reasoner.md](reasoner.md).

internal/brain/http.go

@@ -7,6 +7,8 @@ import (
 	"context"
 	"encoding/json"
 	"fmt"
+	"os/exec"
+	"path/filepath"
 
 	"github.com/eSlider/2dph/internal/brain/rank"
 )
@@ -137,12 +139,22 @@ func (HTTP) Audit(context.Context) ([]byte, error) {
 	return json.Marshal(map[string]any{"status": "ok", "by_confidence": rows})
 }
 
-func (HTTP) Ingest(context.Context) ([]byte, error) {
+func (HTTP) Ingest(ctx context.Context, body []byte) ([]byte, error) {
+	if len(bytes.TrimSpace(body)) == 0 {
 		return json.Marshal(map[string]any{
-		"mode":    "rebuild",
+			"mode":    "add",
-		"command": "bin/brain/index.go --rebuild",
+			"command": "bin/brain/add.go",
-		"add":     "v2",
+			"rebuild": "bin/brain/index.go --rebuild",
 		})
+	}
+	cmd := exec.CommandContext(ctx, filepath.Join(repoRoot(), "bin", "kb", "add"), "--json")
+	cmd.Stdin = bytes.NewReader(body)
+	cmd.Dir = repoRoot()
+	out, err := cmd.Output()
+	if err != nil {
+		return nil, fmt.Errorf("add: %w", err)
+	}
+	return out, nil
 }
 
 func asInt(v any) int64 {

internal/brain/rank/args.go

@@ -6,7 +6,7 @@ import (
 	"strings"
 )
 
-const Usage = `usage: bin/brain/search.go "query" [--root facts|info] [--repo REPO] [-n N] [--json] [--no-web]
+const Usage = `usage: bin/brain/search.go "query" [--root facts|info] [--repo REPO] [-n N] [--hop N] [--json] [--no-web]
        bin/brain/search.go serve [port]
        bin/brain/search.go --list-model`
 
@@ -15,6 +15,7 @@ type Options struct {
 	Root      string
 	Repo      string
 	Limit     int
+	Hop       int
 	JSONOut   bool
 	ListModel bool
 	NoWeb     bool
@@ -22,8 +23,6 @@ type Options struct {
 
 // ParseArgs reads flags. Unknown flags are an error: silently dropping them
 // meant `--hop 1` vanished and its argument `1` was appended to the query.
-// --hop is recognised so it cannot be swallowed; it is not implemented until
-// File/FROM_FILE edges exist.
 func ParseArgs(args []string) (Options, error) {
 	opt := Options{Limit: 20}
 	var queryArgs []string
@@ -52,7 +51,15 @@ func ParseArgs(args []string) (Options, error) {
 			}
 			opt.Limit = n
 		case "--hop":
-			return opt, fmt.Errorf("--hop is not implemented yet (needs File/FROM_FILE edges)")
+			i++
+			n, err := strconv.Atoi(args[i])
+			if err != nil || n < 1 {
+				return opt, fmt.Errorf("--hop must be a positive integer, got %q", args[i])
+			}
+			if n > 3 {
+				return opt, fmt.Errorf("--hop max is 3 (File → Commit → Person)")
+			}
+			opt.Hop = n
 		case "--json":
 			opt.JSONOut = true
 		case "--no-web":

internal/brain/rank/query.go

@@ -7,3 +7,30 @@ const FTSStmt = "CALL QUERY_FTS_INDEX('Leaf', 'id', $q) " +
 
 const VecStmt = "CALL QUERY_VECTOR_INDEX('Leaf', 'Leaf_vec', $q, $n) " +
 	"RETURN node.id, node.text, node.root, node.source, distance ORDER BY distance LIMIT $n"
+
+// HopStmt is the Cypher walk from a search hit. Depth 1 = File, 2 = Commit, 3 = Person.
+func HopStmt(depth int) string {
+	switch depth {
+	case 1:
+		return "MATCH (l:Leaf {id:$id})-[:FROM_FILE]->(f:File) RETURN f.id, f.path, 1"
+	case 2:
+		return "MATCH (l:Leaf {id:$id})-[:FROM_FILE]->(f:File)-[:HAS_VERSION]->(c:Commit) RETURN c.id, c.subject, 2"
+	case 3:
+		return "MATCH (l:Leaf {id:$id})-[:FROM_FILE]->(f:File)-[:HAS_VERSION]->(c:Commit)-[:AUTHORED]->(p:Person) RETURN p.id, p.name, 3"
+	default:
+		return ""
+	}
+}
+
+func HopLabel(depth int) string {
+	switch depth {
+	case 1:
+		return "File"
+	case 2:
+		return "Commit"
+	case 3:
+		return "Person"
+	default:
+		return ""
+	}
+}

internal/brain/rank/rank.go

@@ -7,6 +7,13 @@ import (
 	"strings"
 )
 
+type HopNode struct {
+	ID    string `json:"id"`
+	Label string `json:"label"`
+	Name  string `json:"name"`
+	Depth int    `json:"depth"`
+}
+
 // Hit is one search result, mirroring the python script's dict shape.
 type Hit struct {
 	ID      string    `json:"id"`
@@ -15,6 +22,7 @@ type Hit struct {
 	Source  string    `json:"-"`
 	Score   float64   `json:"score"`
 	Snippet string    `json:"snippet,omitempty"`
+	Hops    []HopNode `json:"hops,omitempty"`
 }
 
 // rrfK dampens the contribution of low ranks; same constant as kblib.py.

internal/brain/rank/rank_test.go

@@ -91,15 +91,41 @@ func TestHybridKeepsVectorScoreForSharedHit(t *testing.T) {
 }
 
 // The old parser dropped unknown flags and appended their arguments to the
-// query, so `search "q" --hop 1` searched for "q 1". --hop is not implemented
+// query, so `search "q" --hop 1` searched for "q 1". --hop must stay a flag.
-// here (needs File edges); it must still fail closed instead of changing q.
 func TestParseHopIsNotSwallowedIntoTheQuery(t *testing.T) {
-	_, err := ParseArgs([]string{"what runs on arc-2", "--hop", "1"})
+	opt, err := ParseArgs([]string{"what runs on arc-2", "--hop", "1"})
-	if err == nil {
+	if err != nil {
-		t.Fatal("expected --hop to error (not implemented), not be swallowed")
+		t.Fatalf("unexpected error: %v", err)
 	}
-	if !strings.Contains(err.Error(), "--hop") {
+	if opt.Query != "what runs on arc-2" {
-		t.Fatalf("error should name --hop, got %v", err)
+		t.Fatalf("query swallowed hop arg: %q", opt.Query)
+	}
+	if opt.Hop != 1 {
+		t.Fatalf("hop = %d, want 1", opt.Hop)
+	}
+}
+
+func TestParseHopMaxIsThree(t *testing.T) {
+	if _, err := ParseArgs([]string{"q", "--hop", "4"}); err == nil {
+		t.Fatal("expected --hop 4 to error")
+	}
+	opt, err := ParseArgs([]string{"q", "--hop", "3"})
+	if err != nil || opt.Hop != 3 {
+		t.Fatalf("hop 3: %+v err=%v", opt, err)
+	}
+}
+
+func TestHopStmtWalksFromFile(t *testing.T) {
+	s := HopStmt(1)
+	if !strings.Contains(s, "FROM_FILE") || !strings.Contains(s, "File") {
+		t.Fatalf("hop 1 must walk FROM_FILE, got %q", s)
+	}
+	s3 := HopStmt(3)
+	if !strings.Contains(s3, "HAS_VERSION") || !strings.Contains(s3, "AUTHORED") || !strings.Contains(s3, "Person") {
+		t.Fatalf("hop 3 must reach Person, got %q", s3)
+	}
+	if HopLabel(1) != "File" || HopLabel(3) != "Person" {
+		t.Fatal("hop labels")
 	}
 }
 

internal/brain/search.go

@@ -56,6 +56,12 @@ func runSearch(args []string) int {
 		fmt.Fprintf(os.Stderr, "search: %v\n", err)
 		return 1
 	}
+	if opt.Hop > 0 {
+		if err := attachHops(hits, opt.Hop); err != nil {
+			fmt.Fprintf(os.Stderr, "hop: %v\n", err)
+			return 1
+		}
+	}
 
 	results := hits
 	for i := range results {
@@ -108,6 +114,44 @@ func searchHits(query, root, repo string, limit int) ([]Hit, error) {
 	return rank.RankAndFilter(fts, vec, root, repo, limit), nil
 }
 
+func attachHops(hits []Hit, n int) error {
+	if conn == nil {
+		return fmt.Errorf("brain not open")
+	}
+	for i := range hits {
+		var hops []rank.HopNode
+		for d := 1; d <= n; d++ {
+			stmt, err := conn.Prepare(rank.HopStmt(d))
+			if err != nil {
+				return err
+			}
+			res, err := conn.Execute(stmt, map[string]any{"id": hits[i].ID})
+			stmt.Close()
+			if err != nil {
+				return err
+			}
+			for res.HasNext() {
+				row, err := res.Next()
+				if err != nil {
+					return err
+				}
+				vals, err := row.GetAsSlice()
+				if err != nil || len(vals) < 3 {
+					continue
+				}
+				hops = append(hops, rank.HopNode{
+					ID:    fmt.Sprint(vals[0]),
+					Label: rank.HopLabel(d),
+					Name:  fmt.Sprint(vals[1]),
+					Depth: int(asInt(vals[2])),
+				})
+			}
+		}
+		hits[i].Hops = hops
+	}
+	return nil
+}
+
 func b2i(err error) int {
 	if err != nil {
 		return 1
@@ -188,6 +232,7 @@ type jsonHit struct {
 	Root    string         `json:"root"`
 	Score   float64        `json:"score"`
 	Snippet string         `json:"snippet,omitempty"`
+	Hops    []rank.HopNode `json:"hops,omitempty"`
 }
 
 func toJSONOut(hits []Hit, query, rootFilter string, web *rank.SecondSource) *jsonOut {
@@ -199,6 +244,7 @@ func toJSONOut(hits []Hit, query, rootFilter string, web *rank.SecondSource) *js
 			Root:    h.Root,
 			Score:   h.Score,
 			Snippet: h.Snippet,
+			Hops:    h.Hops,
 		}
 	}
 	return &jsonOut{
@@ -222,6 +268,18 @@ func resultsToDicts(hits []Hit) []any {
 		if h.Snippet != "" {
 			d = append(d, KV{"snippet", h.Snippet})
 		}
+		if len(h.Hops) > 0 {
+			nodes := make([]any, len(h.Hops))
+			for j, n := range h.Hops {
+				nodes[j] = Dict{
+					{"id", n.ID},
+					{"label", n.Label},
+					{"name", n.Name},
+					{"depth", n.Depth},
+				}
+			}
+			d = append(d, KV{"hops", nodes})
+		}
 		out[i] = d
 	}
 	return out

internal/httpapi/mcp.go

@@ -144,7 +144,15 @@ func (s *Server) mcpCall(r *http.Request, params json.RawMessage) (any, error) {
 			return nil, fmt.Errorf("cancelled")
 		}
 		defer s.release()
-		body, err = s.api.Ingest(r.Context())
+		var payload []byte
+		text := strings.TrimSpace(fmt.Sprint(p.Arguments["text"]))
+		if text != "" && text != "<nil>" {
+			payload, err = json.Marshal(p.Arguments)
+			if err != nil {
+				return nil, err
+			}
+		}
+		body, err = s.api.Ingest(r.Context(), payload)
 	default:
 		return nil, fmt.Errorf("unknown tool %s", p.Name)
 	}

internal/httpapi/server.go

@@ -11,6 +11,7 @@ import (
 	"context"
 	"encoding/json"
 	"errors"
+	"io"
 	"log"
 	"net/http"
 	"os"
@@ -27,7 +28,7 @@ type API interface {
 	Get(ctx context.Context, id string, body bool) ([]byte, error)
 	Stats(ctx context.Context) ([]byte, error)
 	Audit(ctx context.Context) ([]byte, error)
-	Ingest(ctx context.Context) ([]byte, error)
+	Ingest(ctx context.Context, body []byte) ([]byte, error)
 }
 
 type Server struct {
@@ -59,7 +60,7 @@ func (s *Server) ServeHTTP(w http.ResponseWriter, r *http.Request) {
 	case PathAudit:
 		s.handleJSON(w, r, s.api.Audit)
 	case PathIngest:
-		s.handleJSON(w, r, s.api.Ingest)
+		s.handleIngest(w, r)
 	case PathOpenAPI:
 		s.handleOpenAPI(w, r)
 	case PathMCP:
@@ -116,6 +117,24 @@ func (s *Server) handleJSON(w http.ResponseWriter, r *http.Request, fn func(cont
 	writeAPI(w, body, err)
 }
 
+func (s *Server) handleIngest(w http.ResponseWriter, r *http.Request) {
+	var raw []byte
+	if r.Method == http.MethodPost {
+		b, err := io.ReadAll(io.LimitReader(r.Body, 1<<20))
+		if err != nil {
+			writeJSON(w, http.StatusBadRequest, map[string]any{"error": "read body"})
+			return
+		}
+		raw = b
+	}
+	if !s.acquire(w, r) {
+		return
+	}
+	defer s.release()
+	body, err := s.api.Ingest(r.Context(), raw)
+	writeAPI(w, body, err)
+}
+
 func (s *Server) tryAcquire(r *http.Request) bool {
 	return s.acquire(nopWriter{}, r)
 }
@@ -191,11 +210,30 @@ func (ExecSearcher) Get(context.Context, string, bool) ([]byte, error) {
 }
 func (ExecSearcher) Stats(context.Context) ([]byte, error) { return nil, errUnimplemented }
 func (ExecSearcher) Audit(context.Context) ([]byte, error) { return nil, errUnimplemented }
-func (ExecSearcher) Ingest(context.Context) ([]byte, error) {
+func (b ExecSearcher) Ingest(ctx context.Context, body []byte) ([]byte, error) {
+	if len(strings.TrimSpace(string(body))) == 0 {
 		return json.Marshal(map[string]any{
-		"mode":    "rebuild",
+			"mode":    "add",
-		"command": "bin/brain/index.go --rebuild",
+			"command": "bin/brain/add.go",
+			"rebuild": "bin/brain/index.go --rebuild",
 		})
+	}
+	root := os.Getenv("KB_ROOT")
+	if root == "" {
+		root = "."
+	}
+	cmd := exec.CommandContext(ctx, filepath.Join(root, "bin/kb/add"), "--json")
+	cmd.Stdin = strings.NewReader(string(body))
+	cmd.Dir = root
+	out, err := cmd.Output()
+	if err != nil {
+		var exitErr *exec.ExitError
+		if errors.As(err, &exitErr) {
+			return nil, errors.New("add failed: " + strings.TrimSpace(string(exitErr.Stderr)))
+		}
+		return nil, err
+	}
+	return out, nil
 }
 
 func defaultSearchCmd(root string) string {

internal/httpapi/server_test.go

@@ -1,6 +1,7 @@
 package httpapi
 
 import (
+	"bytes"
 	"context"
 	"encoding/json"
 	"net/http"
@@ -64,8 +65,11 @@ func (f *fakeSearcher) Audit(context.Context) ([]byte, error) {
 	return []byte(`{"status":"ok"}`), nil
 }
 
-func (f *fakeSearcher) Ingest(context.Context) ([]byte, error) {
+func (f *fakeSearcher) Ingest(_ context.Context, body []byte) ([]byte, error) {
-	return []byte(`{"mode":"rebuild","command":"bin/brain/index.go --rebuild"}`), nil
+	if len(bytes.TrimSpace(body)) == 0 {
+		return []byte(`{"mode":"add","command":"bin/brain/add.go"}`), nil
+	}
+	return []byte(`{"mode":"add","ids":["fake-leaf"]}`), nil
 }
 
 func (f *fakeSearcher) count() int {
@@ -193,6 +197,27 @@ func TestStatsAuditIngest(t *testing.T) {
 	}
 }
 
+func TestIngestIsAddNotRebuildHint(t *testing.T) {
+	h := NewServer(&fakeSearcher{}, 1)
+	code, body := get(t, h, "/ingest")
+	if code != http.StatusOK {
+		t.Fatalf("GET /ingest code = %d body=%s", code, body)
+	}
+	if strings.Contains(string(body), `"add":"v2"`) || strings.Contains(string(body), "write is v2") {
+		t.Fatalf("GET /ingest still a v2 hint: %s", body)
+	}
+	if !strings.Contains(string(body), "bin/brain/add.go") {
+		t.Fatalf("GET /ingest should name add.go: %s", body)
+	}
+	code, body = postJSON(t, h, "/ingest", `{"text":"hello","root":"info","source":"t"}`)
+	if code != http.StatusOK {
+		t.Fatalf("POST /ingest code = %d body=%s", code, body)
+	}
+	if !strings.Contains(string(body), "fake-leaf") {
+		t.Fatalf("POST /ingest should add: %s", body)
+	}
+}
+
 func TestHTTPPackageDoesNotExecPython(t *testing.T) {
 	raw, err := os.ReadFile("server.go")
 	if err != nil {

internal/httpapi/spec.go

@@ -1,5 +1,7 @@
 package httpapi
 
+import "strings"
+
 // Shared HTTP surface: OpenAPI paths and MCP tools are generated from Ops.
 // ServeHTTP must keep the same path strings.
 
@@ -45,7 +47,15 @@ var Ops = []Op{
 	},
 	{Path: PathStats, Method: "get", ID: "stats", Summary: "index health", MCP: true},
 	{Path: PathAudit, Method: "get", ID: "audit", Summary: "facts confidence histogram", MCP: true},
-	{Path: PathIngest, Method: "get", ID: "ingest", Summary: "rebuild hint (write is v2)", MCP: true},
+	{
+		Path: PathIngest, Method: "post", ID: "ingest", Summary: "add a leaf without rebuild",
+		MCP: true,
+		Params: []Param{
+			{Name: "text", In: "query", Type: "string", Description: "leaf text (omit for CLI hint)"},
+			{Name: "root", In: "query", Type: "string", Description: "facts or info (default info)"},
+			{Name: "source", In: "query", Type: "string", Description: "evidence pointer; facts need two sources"},
+		},
+	},
 	{Path: PathOpenAPI, Method: "get", ID: "openapi", Summary: "OpenAPI 3 document for this server"},
 }
 
@@ -124,3 +134,19 @@ func MCPTools() []MCPTool {
 	}
 	return out
 }
+
+// SkillMarkdown is the Cursor skill fragment generated from Ops/MCPTools.
+func SkillMarkdown() string {
+	var b strings.Builder
+	b.WriteString("# brain HTTP / MCP tools\n\n")
+	b.WriteString("Generated from `internal/httpapi.Ops`. Do not edit by hand.\n\n")
+	b.WriteString("Serve: `bin/brain/serve.go` (`GET /openapi.json`, `POST /mcp`).\n\n")
+	for _, t := range MCPTools() {
+		b.WriteString("- `")
+		b.WriteString(t.Name)
+		b.WriteString("` — ")
+		b.WriteString(t.Description)
+		b.WriteString("\n")
+	}
+	return b.String()
+}

internal/httpapi/spec_test.go

@@ -4,6 +4,8 @@ import (
 	"encoding/json"
 	"net/http"
 	"net/http/httptest"
+	"os"
+	"path/filepath"
 	"strings"
 	"testing"
 )
@@ -36,11 +38,24 @@ func TestMCPToolsMatchOpenAPIPaths(t *testing.T) {
 			t.Fatalf("MCP tool %s has no OpenAPI path %s", tool.Name, path)
 		}
 	}
-	for _, need := range []string{"search", "get", "stats", "audit"} {
+	for _, need := range []string{"search", "get", "stats", "audit", "ingest"} {
 		if !names[need] {
 			t.Fatalf("MCP tools missing %s: %v", need, names)
 		}
 	}
+	var ingest MCPTool
+	for _, tool := range tools {
+		if tool.Name == "ingest" {
+			ingest = tool
+			break
+		}
+	}
+	if strings.Contains(ingest.Description, "v2") {
+		t.Fatalf("ingest still a v2 hint: %s", ingest.Description)
+	}
+	if !strings.Contains(ingest.Description, "add") {
+		t.Fatalf("ingest should describe add: %s", ingest.Description)
+	}
 }
 
 func TestOpenAPIHTTP(t *testing.T) {
@@ -76,6 +91,17 @@ func TestMCPToolsListAndCall(t *testing.T) {
 	}
 }
 
+func TestSkillMarkdownMatchesCommittedFile(t *testing.T) {
+	want, err := os.ReadFile(filepath.Join("..", "..", "skills", "brain", "tools.md"))
+	if err != nil {
+		t.Fatal(err)
+	}
+	got := SkillMarkdown()
+	if got != string(want) {
+		t.Fatalf("skills/brain/tools.md stale; regenerate from SkillMarkdown()\n--- got ---\n%s\n--- want ---\n%s", got, want)
+	}
+}
+
 func postJSON(t *testing.T, h http.Handler, path, raw string) (int, []byte) {
 	t.Helper()
 	req := httptest.NewRequest(http.MethodPost, path, strings.NewReader(raw))

internal/mdleaves/leaf.go

@@ -0,0 +1,146 @@
+package mdleaves
+
+import (
+	"encoding/json"
+	"os"
+	"path/filepath"
+	"regexp"
+	"strings"
+)
+
+type Leaf struct {
+	Source  string `json:"source"`
+	Repo    string `json:"repo"`
+	Heading string `json:"heading"`
+	Text    string `json:"text"`
+	Type    string `json:"type"`
+	Status  string `json:"status"`
+	Related string `json:"related"`
+}
+
+type chunk struct {
+	Heading string
+	Text    string
+}
+
+var (
+	h1 = regexp.MustCompile(`^# \S`)
+	h2 = regexp.MustCompile(`^## \S`)
+)
+
+func ExtractFrontmatter(text string) (map[string]string, string) {
+	if !strings.HasPrefix(text, "---") {
+		return map[string]string{}, text
+	}
+	end := strings.Index(text[3:], "\n---")
+	if end == -1 {
+		return map[string]string{}, text
+	}
+	fm := strings.TrimSpace(text[3 : 3+end])
+	body := text[3+end+4:]
+	meta := map[string]string{}
+	for _, line := range strings.Split(fm, "\n") {
+		key, value, ok := strings.Cut(line, ":")
+		if !ok {
+			continue
+		}
+		meta[strings.TrimSpace(key)] = strings.Trim(strings.TrimSpace(value), `"'`)
+	}
+	return meta, body
+}
+
+func SplitLeafs(meta map[string]string, body string) []chunk {
+	lines := strings.Split(body, "\n")
+	title := ""
+	type hdr struct {
+		heading string
+		start   int
+	}
+	var headers []hdr
+	for i, line := range lines {
+		switch {
+		case h1.MatchString(line):
+			title = strings.TrimSpace(strings.TrimLeft(line, "#"))
+		case h2.MatchString(line):
+			headers = append(headers, hdr{strings.TrimSpace(strings.TrimLeft(line, "#")), i})
+		}
+	}
+	if len(headers) == 0 {
+		var kept []string
+		for _, l := range lines {
+			if strings.TrimSpace(l) != "" {
+				kept = append(kept, l)
+			}
+		}
+		return []chunk{{Heading: title, Text: strings.TrimSpace(strings.Join(kept, "\n"))}}
+	}
+	out := make([]chunk, 0, len(headers))
+	for idx, h := range headers {
+		end := len(lines)
+		if idx+1 < len(headers) {
+			end = headers[idx+1].start
+		}
+		var kept []string
+		for _, l := range lines[h.start:end] {
+			if strings.TrimSpace(l) != "" {
+				kept = append(kept, l)
+			}
+		}
+		text := strings.Join(kept, "\n")
+		if idx == 0 && title != "" {
+			text = title + "\n\n" + text
+		}
+		out = append(out, chunk{Heading: h.heading, Text: strings.TrimSpace(text)})
+	}
+	return out
+}
+
+func ToAll(text, path, repo string) []Leaf {
+	meta, body := ExtractFrontmatter(text)
+	if meta["type"] == "" {
+		meta["type"] = "reference"
+	}
+	if meta["status"] == "" {
+		meta["status"] = "current"
+	}
+	chunks := SplitLeafs(meta, body)
+	out := make([]Leaf, 0, len(chunks))
+	for _, c := range chunks {
+		out = append(out, Leaf{
+			Source:  path,
+			Repo:    repo,
+			Heading: c.Heading,
+			Text:    c.Text,
+			Type:    meta["type"],
+			Status:  meta["status"],
+			Related: meta["related"],
+		})
+	}
+	return out
+}
+
+func WalkMarkdown(root string) ([]string, error) {
+	var out []string
+	err := filepath.Walk(root, func(p string, info os.FileInfo, err error) error {
+		if err != nil {
+			return err
+		}
+		if info.IsDir() {
+			return nil
+		}
+		ext := strings.ToLower(filepath.Ext(p))
+		if ext == ".md" || ext == ".markdown" {
+			out = append(out, p)
+		}
+		return nil
+	})
+	return out, err
+}
+
+func EncodeJSON(leafs []Leaf) (string, error) {
+	raw, err := json.MarshalIndent(leafs, "", "  ")
+	if err != nil {
+		return "", err
+	}
+	return string(raw) + "\n", nil
+}

internal/mdleaves/leaf_test.go

@@ -0,0 +1,73 @@
+package mdleaves
+
+import (
+	"strings"
+	"testing"
+)
+
+func TestSplitLeafsOnH2(t *testing.T) {
+	body := "# Title\n\n## One\n\nalpha\n\n## Two\n\nbeta\n"
+	leafs := SplitLeafs(map[string]string{}, body)
+	if len(leafs) != 2 {
+		t.Fatalf("n=%d", len(leafs))
+	}
+	if leafs[0].Heading != "One" || !strings.Contains(leafs[0].Text, "Title") {
+		t.Fatalf("first=%+v", leafs[0])
+	}
+	if leafs[1].Heading != "Two" || strings.Contains(leafs[1].Text, "Title") {
+		t.Fatalf("second=%+v", leafs[1])
+	}
+}
+
+func TestSplitLeafsNoH2IsWholeDoc(t *testing.T) {
+	body := "# Title\n\njust a paragraph\n"
+	leafs := SplitLeafs(nil, body)
+	if len(leafs) != 1 {
+		t.Fatalf("n=%d", len(leafs))
+	}
+	if leafs[0].Heading != "Title" {
+		t.Fatalf("heading=%q", leafs[0].Heading)
+	}
+	if !strings.Contains(leafs[0].Text, "just a paragraph") {
+		t.Fatalf("text=%q", leafs[0].Text)
+	}
+}
+
+func TestFrontmatterAndToAll(t *testing.T) {
+	raw := "---\ntype: howto\nstatus: current\nrelated: docs/design.md\n---\n# Doc\n\n## Step\n\ndo it\n"
+	got := ToAll(raw, "docs/x.md", "eSlider/2dph")
+	if len(got) != 1 {
+		t.Fatalf("n=%d", len(got))
+	}
+	if got[0].Type != "howto" || got[0].Status != "current" {
+		t.Fatalf("%+v", got[0])
+	}
+	if got[0].Source != "docs/x.md" || got[0].Repo != "eSlider/2dph" {
+		t.Fatalf("%+v", got[0])
+	}
+	if got[0].Related != "docs/design.md" {
+		t.Fatalf("related=%q", got[0].Related)
+	}
+}
+
+func TestEncodeJSONAndYAML(t *testing.T) {
+	leafs := ToAll("# Hi\n\nbody\n", "a.md", "")
+	js, err := EncodeJSON(leafs)
+	if err != nil {
+		t.Fatal(err)
+	}
+	if !strings.Contains(js, `"heading": "Hi"`) {
+		t.Fatalf("json=%s", js)
+	}
+	y := EncodeYAML(leafs)
+	if !strings.Contains(y, "heading: Hi") {
+		t.Fatalf("yaml=%s", y)
+	}
+}
+
+func TestDefaultsTypeAndStatus(t *testing.T) {
+	got := ToAll("# Hi\n\nbody\n", "a.md", "")
+	if got[0].Type != "reference" || got[0].Status != "current" {
+		t.Fatalf("%+v", got[0])
+	}
+}

internal/mdleaves/yaml.go

@@ -0,0 +1,36 @@
+package mdleaves
+
+import (
+	"fmt"
+	"strconv"
+	"strings"
+)
+
+func EncodeYAML(leafs []Leaf) string {
+	if len(leafs) == 0 {
+		return "[]\n"
+	}
+	var b strings.Builder
+	for _, lf := range leafs {
+		b.WriteString("-\n")
+		writeKV(&b, "source", lf.Source)
+		writeKV(&b, "repo", lf.Repo)
+		writeKV(&b, "heading", lf.Heading)
+		writeKV(&b, "text", lf.Text)
+		writeKV(&b, "type", lf.Type)
+		writeKV(&b, "status", lf.Status)
+		writeKV(&b, "related", lf.Related)
+	}
+	return b.String()
+}
+
+func writeKV(b *strings.Builder, k, v string) {
+	fmt.Fprintf(b, "  %s: %s\n", k, yamlScalar(v))
+}
+
+func yamlScalar(s string) string {
+	if strings.Contains(s, "\n") || s == "" || strings.ContainsAny(s, ":#'\"[]{}&*!|>%@`") || s != strings.TrimSpace(s) {
+		return strconv.Quote(s)
+	}
+	return s
+}

internal/ocr/ocr.go

@@ -0,0 +1,162 @@
+// Package ocr runs Tesseract (eng+deu) on images and scanned PDFs.
+//
+// Default engine is the tesseract CLI, not gosseract CGO: Ladybug CGO stays
+// Zig-only (D21). Same engine, no gocv. OCR_ENGINE=paddle selects paddleocr
+// when that binary is on PATH (compose profile ocr-paddle).
+package ocr
+
+import (
+	"fmt"
+	"image"
+	"image/color"
+	"image/png"
+	"os"
+	"os/exec"
+	"path/filepath"
+	"strings"
+)
+
+const TessLang = "eng+deu"
+
+func ImageFile(path string) (string, error) {
+	engine := os.Getenv("OCR_ENGINE")
+	if engine == "paddle" {
+		return runPaddle(path)
+	}
+	return runTesseract(path)
+}
+
+func PDFFile(path string) (string, error) {
+	text, err := pdfToText(path)
+	if err == nil && strings.TrimSpace(text) != "" {
+		return strings.TrimSpace(text), nil
+	}
+	ocr, oerr := pdfPages(path)
+	if oerr != nil {
+		if err != nil {
+			return "", err
+		}
+		return "", oerr
+	}
+	if strings.TrimSpace(ocr) != "" {
+		return strings.TrimSpace(ocr), nil
+	}
+	if text != "" {
+		return strings.TrimSpace(text), nil
+	}
+	return "", fmt.Errorf("pdf has no text layer (ocr unavailable)")
+}
+
+func pdfToText(path string) (string, error) {
+	cmd := exec.Command("pdftotext", "-layout", path, "-")
+	out, err := cmd.Output()
+	if err != nil {
+		return "", err
+	}
+	return string(out), nil
+}
+
+func pdfPages(path string) (string, error) {
+	dir, err := os.MkdirTemp("", "2dph-ocr-")
+	if err != nil {
+		return "", err
+	}
+	defer os.RemoveAll(dir)
+	prefix := filepath.Join(dir, "page")
+	cmd := exec.Command("pdftoppm", "-png", "-r", "200", path, prefix)
+	if err := cmd.Run(); err != nil {
+		return "", err
+	}
+	matches, err := filepath.Glob(prefix + "*.png")
+	if err != nil {
+		return "", err
+	}
+	var parts []string
+	for _, img := range matches {
+		t, err := ImageFile(img)
+		if err != nil {
+			continue
+		}
+		if s := strings.TrimSpace(t); s != "" {
+			parts = append(parts, s)
+		}
+	}
+	return strings.Join(parts, "\n\n"), nil
+}
+
+func runTesseract(path string) (string, error) {
+	pre, err := preprocessFile(path)
+	if err != nil {
+		pre = path
+	} else {
+		defer os.Remove(pre)
+	}
+	cmd := exec.Command("tesseract", pre, "stdout", "-l", TessLang, "--psm", "6")
+	out, err := cmd.Output()
+	if err != nil {
+		return "", err
+	}
+	return strings.TrimSpace(string(out)), nil
+}
+
+func runPaddle(path string) (string, error) {
+	cmd := exec.Command("paddleocr", "ocr", "-i", path)
+	out, err := cmd.Output()
+	if err != nil {
+		return "", err
+	}
+	return strings.TrimSpace(string(out)), nil
+}
+
+func preprocessFile(path string) (string, error) {
+	f, err := os.Open(path)
+	if err != nil {
+		return "", err
+	}
+	defer f.Close()
+	img, err := png.Decode(f)
+	if err != nil {
+		return "", err
+	}
+	out := filepath.Join(os.TempDir(), filepath.Base(path)+".gray.png")
+	w, err := os.Create(out)
+	if err != nil {
+		return "", err
+	}
+	defer w.Close()
+	if err := png.Encode(w, GrayContrast(img)); err != nil {
+		os.Remove(out)
+		return "", err
+	}
+	return out, nil
+}
+
+// GrayContrast is a stdlib preprocess (no gocv): grayscale + stretch.
+func GrayContrast(src image.Image) image.Image {
+	b := src.Bounds()
+	dst := image.NewGray(b)
+	var minL, maxL uint8 = 255, 0
+	for y := b.Min.Y; y < b.Max.Y; y++ {
+		for x := b.Min.X; x < b.Max.X; x++ {
+			g := color.GrayModel.Convert(src.At(x, y)).(color.Gray)
+			if g.Y < minL {
+				minL = g.Y
+			}
+			if g.Y > maxL {
+				maxL = g.Y
+			}
+		}
+	}
+	span := int(maxL) - int(minL)
+	if span < 1 {
+		span = 1
+	}
+	for y := b.Min.Y; y < b.Max.Y; y++ {
+		for x := b.Min.X; x < b.Max.X; x++ {
+			g := color.GrayModel.Convert(src.At(x, y)).(color.Gray)
+			v := uint8((int(g.Y) - int(minL)) * 255 / span)
+			dst.SetGray(x, y, color.Gray{Y: v})
+		}
+	}
+	return dst
+}

internal/ocr/ocr_test.go

@@ -0,0 +1,54 @@
+package ocr
+
+import (
+	"image"
+	"image/color"
+	"os/exec"
+	"path/filepath"
+	"strings"
+	"testing"
+)
+
+func TestGrayContrastStretches(t *testing.T) {
+	img := image.NewGray(image.Rect(0, 0, 2, 2))
+	img.SetGray(0, 0, color.Gray{Y: 64})
+	img.SetGray(0, 1, color.Gray{Y: 64})
+	img.SetGray(1, 0, color.Gray{Y: 64})
+	img.SetGray(1, 1, color.Gray{Y: 192})
+	out := GrayContrast(img).(*image.Gray)
+	if out.GrayAt(0, 0).Y != 0 {
+		t.Fatalf("min should map to 0, got %d", out.GrayAt(0, 0).Y)
+	}
+	if out.GrayAt(1, 1).Y != 255 {
+		t.Fatalf("max should map to 255, got %d", out.GrayAt(1, 1).Y)
+	}
+}
+
+func TestHelloPNGFixtureOCR(t *testing.T) {
+	if _, err := exec.LookPath("tesseract"); err != nil {
+		t.Skip("tesseract not installed")
+	}
+	path := filepath.Join("testdata", "hello.png")
+	got, err := ImageFile(path)
+	if err != nil {
+		t.Fatal(err)
+	}
+	up := strings.ToUpper(got)
+	if !strings.Contains(up, "HELLO") {
+		t.Fatalf("ocr %q missing HELLO", got)
+	}
+}
+
+func TestPaddleEngineUsesPaddleocrBinary(t *testing.T) {
+	t.Setenv("OCR_ENGINE", "paddle")
+	_, err := ImageFile(filepath.Join("testdata", "hello.png"))
+	if _, look := exec.LookPath("paddleocr"); look != nil {
+		if err == nil {
+			t.Fatal("expected error when paddleocr is missing")
+		}
+		return
+	}
+	if err != nil {
+		t.Fatal(err)
+	}
+}

internal/reasoner/client.go

@@ -0,0 +1,325 @@
+package reasoner
+
+import (
+	"bytes"
+	"encoding/json"
+	"fmt"
+	"io"
+	"net/http"
+	"strings"
+	"time"
+)
+
+// HF IDs named in docs. No Qwen3.6-9B exists.
+const (
+	HFQwen35_9B   = "Qwen/Qwen3.5-9B"
+	HFQwen36_27B  = "Qwen/Qwen3.6-27B"
+	HFBonsai27B   = "prism-ml/Bonsai-27B-gguf"
+	OllamaRAM     = "qwen3.5:9b"
+	OllamaQuality = "MichelRosselli/bonsai-27b:Q1_0"
+)
+
+type ToolCall struct {
+	Name      string
+	Arguments string
+}
+
+type Result struct {
+	Model      string `json:"model"`
+	OK         bool   `json:"ok"`
+	ToolName   string `json:"tool_name,omitempty"`
+	XMLLeak    bool   `json:"xml_leak"`
+	Err        string `json:"error,omitempty"`
+	LatencyMS  int64  `json:"latency_ms"`
+	RSSMB      int    `json:"rss_mb,omitempty"`
+	Device     string `json:"device"`
+	WantedTool string `json:"wanted_tool"`
+}
+
+type Prompt struct {
+	Name string
+	Want string
+	User string
+}
+
+var BakePrompts = []Prompt{
+	{
+		Name: "search-before-claim",
+		Want: "search",
+		User: "Use tools. Search the 2dph brain for LadybugDB before you answer. Call search.",
+	},
+	{
+		Name: "get-leaf",
+		Want: "get",
+		User: "Use tools. Fetch leaf id leaf-demo with get. Do not invent the body.",
+	},
+	{
+		Name: "audit-index",
+		Want: "audit",
+		User: "Use tools. Call audit on the brain index health.",
+	},
+}
+
+func MCPTools() []map[string]any {
+	return []map[string]any{
+		openaiTool("search", "deduction search (facts → info → web)", map[string]any{
+			"type": "object",
+			"properties": map[string]any{
+				"q": map[string]any{"type": "string", "description": "search query"},
+				"n": map[string]any{"type": "integer"},
+			},
+			"required": []string{"q"},
+		}),
+		openaiTool("get", "read one leaf by id", map[string]any{
+			"type": "object",
+			"properties": map[string]any{
+				"id":   map[string]any{"type": "string"},
+				"body": map[string]any{"type": "boolean"},
+			},
+			"required": []string{"id"},
+		}),
+		openaiTool("audit", "facts confidence histogram", map[string]any{
+			"type":       "object",
+			"properties": map[string]any{},
+		}),
+	}
+}
+
+func openaiTool(name, desc string, schema map[string]any) map[string]any {
+	return map[string]any{
+		"type": "function",
+		"function": map[string]any{
+			"name":        name,
+			"description": desc,
+			"parameters":  schema,
+		},
+	}
+}
+
+type Client struct {
+	BaseURL    string
+	Model      string
+	HTTP       *http.Client
+	Device     string
+	ToolChoice string
+}
+
+type Report struct {
+	Model      string   `json:"model"`
+	HF         string   `json:"hf_id"`
+	Device     string   `json:"device"`
+	ToolCallOK int      `json:"tool_call_ok"`
+	ToolCallN  int      `json:"tool_call_n"`
+	XMLLeak    int      `json:"xml_leak"`
+	RSSMB      int      `json:"rss_mb"`
+	VRAMMB     int      `json:"vram_mb"`
+	Prompts    []Result `json:"prompts"`
+}
+
+func HFFor(model string) string {
+	switch model {
+	case OllamaRAM:
+		return HFQwen35_9B
+	case OllamaQuality:
+		return HFBonsai27B
+	default:
+		if strings.Contains(model, "qwen3.6") || strings.Contains(model, "Qwen3.6") {
+			return HFQwen36_27B
+		}
+		return ""
+	}
+}
+
+func (c *Client) httpc() *http.Client {
+	if c.HTTP == nil {
+		c.HTTP = &http.Client{Timeout: 10 * time.Minute}
+	}
+	return c.HTTP
+}
+
+func Origin(base string) string {
+	s := strings.TrimRight(base, "/")
+	return strings.TrimSuffix(s, "/v1")
+}
+
+func (c Client) ChatTools(user string) (ToolCall, string, error) {
+	choice := c.ToolChoice
+	if choice == "" {
+		choice = "required"
+	}
+	body, _ := json.Marshal(map[string]any{
+		"model": c.Model,
+		"messages": []map[string]string{
+			{"role": "system", "content": "You are PicoClaw talking to 2dph MCP. Always call a tool before a factual claim. search then get then audit."},
+			{"role": "user", "content": user},
+		},
+		"tools":       MCPTools(),
+		"tool_choice": choice,
+	})
+	base := strings.TrimRight(c.BaseURL, "/")
+	if !strings.HasSuffix(base, "/v1") {
+		base += "/v1"
+	}
+	url := base + "/chat/completions"
+	req, err := http.NewRequest(http.MethodPost, url, bytes.NewReader(body))
+	if err != nil {
+		return ToolCall{}, "", err
+	}
+	req.Header.Set("Content-Type", "application/json")
+	res, err := c.httpc().Do(req)
+	if err != nil {
+		return ToolCall{}, "", err
+	}
+	defer res.Body.Close()
+	raw, _ := io.ReadAll(res.Body)
+	if res.StatusCode >= 300 {
+		return ToolCall{}, string(raw), fmt.Errorf("http %d", res.StatusCode)
+	}
+	return ParseToolResponse(raw)
+}
+
+func ParseToolResponse(raw []byte) (ToolCall, string, error) {
+	var wrap struct {
+		Choices []struct {
+			Message struct {
+				Content   string `json:"content"`
+				ToolCalls []struct {
+					Function struct {
+						Name      string          `json:"name"`
+						Arguments json.RawMessage `json:"arguments"`
+					} `json:"function"`
+				} `json:"tool_calls"`
+			} `json:"message"`
+		} `json:"choices"`
+	}
+	if err := json.Unmarshal(raw, &wrap); err != nil {
+		return ToolCall{}, "", err
+	}
+	content := ""
+	if len(wrap.Choices) > 0 {
+		content = wrap.Choices[0].Message.Content
+		if n := len(wrap.Choices[0].Message.ToolCalls); n > 0 {
+			fn := wrap.Choices[0].Message.ToolCalls[0].Function
+			return ToolCall{Name: fn.Name, Arguments: rawArgs(fn.Arguments)}, content, nil
+		}
+	}
+	return ToolCall{}, content, fmt.Errorf("no tool_calls")
+}
+
+func rawArgs(raw json.RawMessage) string {
+	if len(raw) == 0 {
+		return ""
+	}
+	var s string
+	if err := json.Unmarshal(raw, &s); err == nil {
+		return s
+	}
+	return string(raw)
+}
+
+func XMLLeak(content string) bool {
+	s := strings.ToLower(content)
+	return strings.Contains(s, "<tool_call>") ||
+		strings.Contains(s, "<function=") ||
+		strings.Contains(s, "<parameter")
+}
+
+func RunPrompt(c Client, p Prompt) Result {
+	start := time.Now()
+	tc, content, err := c.ChatTools(p.User)
+	out := Result{
+		Model:      c.Model,
+		WantedTool: p.Want,
+		LatencyMS:  time.Since(start).Milliseconds(),
+		Device:     c.Device,
+		XMLLeak:    XMLLeak(content),
+	}
+	if err != nil {
+		out.Err = err.Error()
+		if content != "" && out.XMLLeak {
+			out.Err = "xml tool call instead of openai tool_calls"
+		}
+		return out
+	}
+	out.ToolName = tc.Name
+	out.OK = tc.Name == p.Want
+	if !out.OK {
+		out.Err = "wanted " + p.Want + " got " + tc.Name
+	}
+	return out
+}
+
+type ProcMem struct {
+	Name   string
+	SizeMB int
+	VRAMMB int
+}
+
+func ParsePS(raw []byte) []ProcMem {
+	var wrap struct {
+		Models []struct {
+			Name     string `json:"name"`
+			Size     int64  `json:"size"`
+			SizeVRAM int64  `json:"size_vram"`
+		} `json:"models"`
+	}
+	if err := json.Unmarshal(raw, &wrap); err != nil {
+		return nil
+	}
+	out := make([]ProcMem, 0, len(wrap.Models))
+	for _, m := range wrap.Models {
+		out = append(out, ProcMem{
+			Name:   m.Name,
+			SizeMB: int(m.Size / (1024 * 1024)),
+			VRAMMB: int(m.SizeVRAM / (1024 * 1024)),
+		})
+	}
+	return out
+}
+
+func (c Client) FetchPS() []ProcMem {
+	url := Origin(c.BaseURL) + "/api/ps"
+	res, err := c.httpc().Get(url)
+	if err != nil {
+		return nil
+	}
+	defer res.Body.Close()
+	raw, _ := io.ReadAll(res.Body)
+	if res.StatusCode >= 300 {
+		return nil
+	}
+	return ParsePS(raw)
+}
+
+func Run(c Client) Report {
+	if c.Device == "" {
+		c.Device = "cpu"
+	}
+	rep := Report{
+		Model:  c.Model,
+		HF:     HFFor(c.Model),
+		Device: c.Device,
+	}
+	for _, p := range BakePrompts {
+		r := RunPrompt(c, p)
+		rep.Prompts = append(rep.Prompts, r)
+		rep.ToolCallN++
+		if r.OK {
+			rep.ToolCallOK++
+		}
+		if r.XMLLeak {
+			rep.XMLLeak++
+		}
+	}
+	if mems := c.FetchPS(); len(mems) > 0 {
+		rep.RSSMB = mems[0].SizeMB
+		rep.VRAMMB = mems[0].VRAMMB
+		if c.Device == "cpu" && mems[0].VRAMMB > 0 {
+			rep.Device = "gpu"
+		}
+		for i := range rep.Prompts {
+			rep.Prompts[i].RSSMB = rep.RSSMB
+		}
+	}
+	return rep
+}

internal/reasoner/client_test.go

@@ -0,0 +1,166 @@
+package reasoner
+
+import (
+	"encoding/json"
+	"io"
+	"net/http"
+	"net/http/httptest"
+	"strings"
+	"testing"
+
+	"github.com/eSlider/2dph/internal/httpapi"
+)
+
+func TestHFIdsAreRealAndNoQwen36Nine(t *testing.T) {
+	if HFQwen35_9B != "Qwen/Qwen3.5-9B" {
+		t.Fatalf("9B id = %s", HFQwen35_9B)
+	}
+	if HFQwen36_27B != "Qwen/Qwen3.6-27B" {
+		t.Fatalf("27B id = %s", HFQwen36_27B)
+	}
+	if HFBonsai27B != "prism-ml/Bonsai-27B-gguf" {
+		t.Fatalf("bonsai id = %s", HFBonsai27B)
+	}
+}
+
+func TestParseToolResponseOpenAI(t *testing.T) {
+	raw := []byte(`{"choices":[{"message":{"tool_calls":[{"function":{"name":"search","arguments":"{\"q\":\"LadybugDB\"}"}}]}}]}`)
+	tc, _, err := ParseToolResponse(raw)
+	if err != nil {
+		t.Fatal(err)
+	}
+	if tc.Name != "search" {
+		t.Fatalf("name=%s", tc.Name)
+	}
+}
+
+func TestParseToolResponseXMLIsFailure(t *testing.T) {
+	raw := []byte(`{"choices":[{"message":{"content":"<tool_call>search</tool_call>"}}]}`)
+	_, content, err := ParseToolResponse(raw)
+	if err == nil {
+		t.Fatal("expected no tool_calls")
+	}
+	if !XMLLeak(content) {
+		t.Fatal("xml leak not detected")
+	}
+}
+
+func TestBakePromptsWantMCPTools(t *testing.T) {
+	if len(BakePrompts) != 3 {
+		t.Fatalf("prompts=%d", len(BakePrompts))
+	}
+	names := map[string]bool{}
+	for _, p := range BakePrompts {
+		names[p.Want] = true
+	}
+	for _, n := range []string{"search", "get", "audit"} {
+		if !names[n] {
+			t.Fatalf("missing want %s", n)
+		}
+	}
+}
+
+func TestParseToolResponseObjectArgs(t *testing.T) {
+	raw := []byte(`{"choices":[{"message":{"tool_calls":[{"function":{"name":"get","arguments":{"id":"leaf-demo"}}}]}}]}`)
+	tc, _, err := ParseToolResponse(raw)
+	if err != nil {
+		t.Fatal(err)
+	}
+	if tc.Name != "get" {
+		t.Fatalf("name=%s", tc.Name)
+	}
+	if !strings.Contains(tc.Arguments, "leaf-demo") {
+		t.Fatalf("args=%s", tc.Arguments)
+	}
+}
+
+func TestParsePSCPUNotVRAM(t *testing.T) {
+	raw := []byte(`{"models":[{"name":"qwen3.5:9b","size":6900000000,"size_vram":0}]}`)
+	got := ParsePS(raw)
+	if len(got) != 1 {
+		t.Fatalf("n=%d", len(got))
+	}
+	if got[0].VRAMMB != 0 {
+		t.Fatalf("vram=%d", got[0].VRAMMB)
+	}
+	if got[0].SizeMB < 6000 {
+		t.Fatalf("rss=%d", got[0].SizeMB)
+	}
+}
+
+func TestMCPToolsArePicoClawSubset(t *testing.T) {
+	mcp := map[string]bool{}
+	for _, op := range httpapi.Ops {
+		if op.MCP {
+			mcp[op.ID] = true
+		}
+	}
+	seen := map[string]bool{}
+	for _, tool := range MCPTools() {
+		fn, _ := tool["function"].(map[string]any)
+		name, _ := fn["name"].(string)
+		if !mcp[name] {
+			t.Fatalf("%s is not an MCP op", name)
+		}
+		seen[name] = true
+	}
+	for _, n := range []string{"search", "get", "audit"} {
+		if !seen[n] {
+			t.Fatalf("bake-off missing %s", n)
+		}
+	}
+}
+
+func TestChatToolsHitsOpenAIPath(t *testing.T) {
+	var gotTools bool
+	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		if r.URL.Path != "/v1/chat/completions" {
+			t.Errorf("path=%s", r.URL.Path)
+		}
+		body, _ := io.ReadAll(r.Body)
+		var payload map[string]any
+		_ = json.Unmarshal(body, &payload)
+		if _, ok := payload["tools"]; ok {
+			gotTools = true
+		}
+		w.Header().Set("Content-Type", "application/json")
+		_, _ = w.Write([]byte(`{"choices":[{"message":{"tool_calls":[{"function":{"name":"search","arguments":"{\"q\":\"LadybugDB\"}"}}]}}]}`))
+	}))
+	defer srv.Close()
+	c := Client{BaseURL: srv.URL + "/v1", Model: "stub", HTTP: srv.Client(), Device: "cpu"}
+	r := RunPrompt(c, BakePrompts[0])
+	if !gotTools {
+		t.Fatal("tools not sent")
+	}
+	if !r.OK {
+		t.Fatalf("result=%+v", r)
+	}
+}
+
+func TestRunSamplesPSAfterPrompts(t *testing.T) {
+	srv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		w.Header().Set("Content-Type", "application/json")
+		if r.URL.Path == "/api/ps" {
+			_, _ = w.Write([]byte(`{"models":[{"name":"stub","size":6500000000,"size_vram":0}]}`))
+			return
+		}
+		_, _ = w.Write([]byte(`{"choices":[{"message":{"tool_calls":[{"function":{"name":"search","arguments":"{}"}}]}}]}`))
+	}))
+	defer srv.Close()
+	c := Client{BaseURL: srv.URL + "/v1", Model: "stub", HTTP: srv.Client(), Device: "cpu"}
+	if mems := c.FetchPS(); len(mems) != 1 || mems[0].VRAMMB != 0 {
+		t.Fatalf("%+v", mems)
+	}
+	if mems := c.FetchPS(); mems[0].SizeMB < 6000 {
+		t.Fatalf("rss=%d", mems[0].SizeMB)
+	}
+}
+
+func TestHFForKnownOllamaTags(t *testing.T) {
+	if HFFor(OllamaRAM) != HFQwen35_9B {
+		t.Fatal(HFFor(OllamaRAM))
+	}
+	if HFFor(OllamaQuality) != HFBonsai27B {
+		t.Fatal(HFFor(OllamaQuality))
+	}
+}

pyproject.toml

@@ -6,7 +6,6 @@ readme = "README.md"
 requires-python = ">=3.12"
 license = { text = "MIT" }
 dependencies = [
-    "docling>=2.119.0",
     "ladybug==0.19.1",
     "markitdown[docx,epub,html,image-exif,pdf,pptx,xlsx,zip]>=0.1.7",
     "mistune==3.3.4",

skills/brain/SKILL.md

@@ -26,13 +26,14 @@ second independent source when local roots cannot confirm. An answer is
 bin/brain/search.go "Matrix federation"                # pointers + snippets, YAML
 bin/brain/search.go "onlyoffice postgres" --root facts # restrict to confirmed
 bin/brain/search.go "where is cs-lexicon" --json | yq '.[].ref'
+bin/brain/add.go --text T --root facts --source "a.md x b.md"
 bin/brain/get.go <id> --body                           # full chunk only when needed
 bin/brain/stats.go                                     # index health
 bin/brain/eval.go                                      # recall@5 >= 0.95 gate (Go; Python bin/kb/eval is CI fallback)
 ```
 
-`bin/kb/search` is a deprecated wrapper. `--hop` errors (File/FROM_FILE edges
+`bin/kb/search` is a deprecated wrapper. `--hop N` walks
-are not wired yet); do not treat it as a graph walk.
+`FROM_FILE` / `HAS_VERSION` / `AUTHORED` from each hit (1=File, 3=Person).
 
 ## Rules
 
@@ -45,5 +46,6 @@ are not wired yet); do not treat it as a graph walk.
 - If recall looks wrong, run `bin/brain/eval.go`; it gates control questions and
   should stay at or above 95% recall@5.
 - Agents: `GET /openapi.json` and `POST /mcp` on `bin/brain/serve.go` (same
-  handlers; tool names match paths `search`/`get`/`stats`/`audit`).
+  handlers; tool names match paths `search`/`get`/`stats`/`audit`). Generated
+  list: [tools.md](tools.md).
 - Never report an unconfirmed single-source local answer as fact.

skills/brain/tools.md

@@ -0,0 +1,11 @@
+# brain HTTP / MCP tools
+
+Generated from `internal/httpapi.Ops`. Do not edit by hand.
+
+Serve: `bin/brain/serve.go` (`GET /openapi.json`, `POST /mcp`).
+
+- `search` — deduction search (facts → info → web)
+- `get` — read one leaf by id
+- `stats` — index health
+- `audit` — facts confidence histogram
+- `ingest` — add a leaf without rebuild

skills/db-yaml/SKILL.md

@@ -1,39 +0,0 @@
----
-name: db-yaml
-description: >-
-  Read any Postgres as compact YAML through db/psql-yq, with a read-only guard and
-  named profiles. Use when a task needs table contents, column types or a SELECT
-  against cs_brain or another project database.
----
-
-# db-yaml
-
-`bin/db/psql-yq` (vendored in this repo) talks to Postgres and returns YAML,
-which is far cheaper than a psql ASCII table and easy to slice with `yq`.
-
-```bash
-bin/db/psql-yq --profile onlyoffice -s document_asset   # column list
-bin/db/psql-yq --profile onlyoffice -t task_result -l 20  # sample rows as YAML
-bin/db/psql-yq --profile onlyoffice -c 'SELECT ...'     # query -> YAML
-```
-
-Ad-hoc targets without a profile:
-
-```bash
-bin/db/psql-yq --container my-pg --db app -c 'SELECT 1'
-bin/db/psql-yq --dsn 'postgres://user@host:5432/db' -c 'SELECT 1'
-```
-
-## Profiles
-
-Connection details live in `~/.config/brain/db-profiles.yml` (mode 600), never in a
-project repo. A profile names either a `container` or a `host`; passwords are read
-from a separate `password_env_file` and never appear in argv.
-
-## Rules
-
-- **Read-only.** Any `insert|update|delete|drop|truncate|alter|create|grant|
-  revoke|vacuum|copy` is rejected with exit 3. Do not work around it.
-- **PII.** `cs_brain` holds client data. Aggregate and count freely; never copy
-  names or addresses into chat, issues or docs.
-- Use `-l` to keep samples small. Twenty rows answer most questions.

skills/picoclaw/SKILL.md

@@ -0,0 +1,26 @@
+---
+name: picoclaw
+description: >-
+  2dph is the memory/fact gate, not the agent loop. Use when wiring PicoClaw
+  or any MCP client: call brain search/get/audit before a factual reply.
+  throttled is not a negative finding.
+---
+
+# PicoClaw — fact-check before assert
+
+PicoClaw (or any agent) speaks MCP at `POST /mcp` on `bin/brain/serve.go`.
+2dph does not run the agent loop. Compose: `docker compose --profile picoclaw up brain-mcp`
+(see [docs/picoclaw.md](../../docs/picoclaw.md)).
+
+## Tool order (before a factual reply)
+
+1. **`search`** — facts root first, then info. The `web` block is a second
+   source when there is no facts hit. Status `throttled` / `skipped` /
+   `refused` is **not** evidence of absence.
+2. **`get`** — full leaf body only when a hit `id` is needed.
+3. **`audit`** — if recall or confidence looks wrong.
+
+Then answer. Confirmed only from facts (≥2 independent sources). Anything
+else is `(not confirmed)`. Missing graph ≠ “does not exist”.
+
+Generated tool list: [../brain/tools.md](../brain/tools.md).

skills/postgres/SKILL.md

@@ -0,0 +1,39 @@
+---
+name: postgres
+description: >-
+  Read Postgres as compact YAML through bin/postgres/query.go (read-only
+  guard, named profiles). Use when a task needs table contents, column types,
+  or a SELECT against an ops database.
+---
+
+# postgres
+
+`bin/postgres/query.go` wraps vendored `bin/db/psql-yq`. Output is YAML
+(cheaper than psql ASCII, easy to slice with `yq`).
+
+```bash
+bin/postgres/query.go --profile onlyoffice -s document_asset   # column list
+bin/postgres/query.go --profile onlyoffice -t task_result -l 20  # sample rows
+bin/postgres/query.go --profile onlyoffice -c 'SELECT ...'     # query → YAML
+```
+
+Ad-hoc targets without a profile:
+
+```bash
+bin/postgres/query.go --container my-pg --db app -c 'SELECT 1'
+bin/postgres/query.go --dsn 'postgres://user@host:5432/db' -c 'SELECT 1'
+```
+
+## Profiles
+
+Connection details live in `$HOME/.config/brain/db-profiles.yml` (mode 600),
+never in a project repo. A profile names either a `container` or a `host`;
+passwords are read from a separate `password_env_file` and never appear in argv.
+
+## Rules
+
+- **Read-only.** Any `insert|update|delete|drop|truncate|alter|create|grant|
+  revoke|vacuum|copy` is rejected with exit 3. Do not work around it.
+- **PII.** Client CRM databases: aggregate and count freely; never copy names
+  or addresses into chat, issues or docs.
+- Use `-l` to keep samples small. Twenty rows answer most questions.