feat: DuckDB quantiles in-process (gcc CGO), not Ladybug.

OQ3: internal/duckstats + bin/qa/stats.go. Zig stays Ladybug-only. mikefarah/yq for small structured slices. Gitea #30.
feat: OCR scans with tesseract, drop docling from the default path. (#33 )
2026-08-14 11:29:22 +01:00 · 2026-08-14 11:26:04 +01:00 · 2026-08-14 11:03:39 +01:00 · 2026-08-14 10:57:40 +01:00 · 2026-08-14 10:51:35 +01:00
45 changed files with 1487 additions and 1819 deletions
@@ -45,27 +45,53 @@ jobs:
        run: |
          uv run python -m unittest discover -s bin/tools -t .

-      - name: Go tests (root module, no ladybug cgo)
+      - name: Go tests (root module; duckdb-go CGO via gcc, no ladybug)
        run: |
-          go vet ./...
-          go test ./... -count=1
+          CC=gcc CXX=g++ CGO_CFLAGS= CGO_LDFLAGS= go vet ./...
+          CC=gcc CXX=g++ CGO_CFLAGS= CGO_LDFLAGS= go test ./... -count=1

      - name: brain ranking tests (no cgo / no ladybug)
        run: go test ./internal/brain/rank -count=1

      - name: facts/audit self (lexicon consistency, no network)
-        run: |
-          ./bin/facts/audit self 2>/dev/null || echo "audit: not yet implemented; gate skipped"
+        run: ./bin/facts/audit self

-      - name: CGO via Zig (compile brain/search)
+      - name: CGO via Zig (compile brain/search + eval)
        run: |
          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
@@ -12,3 +12,5 @@ __pycache__/
 lib-ladybug/
 go.work.local
 models/
+# Purged from git history. Do not re-add.
+docs/crm-associations-proof.md
@@ -42,13 +42,14 @@ 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 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 (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)
 bin/reasoner/ bakeoff.go (D18 CPU OpenAI tool-call bake-off)
-internal/     shared Go (brain/rank is cgo-free; chats parsers; gitlog; websearch; reasoner)
+internal/     shared Go (brain/rank is cgo-free; chats parsers; gitlog; websearch; reasoner; duckstats)
+bin/qa/       stats.go (DuckDB quantiles / JSONL count; gcc CGO, not Zig)
 bin/watch/    corpus watcher (used by bin/brain/watch.go)
 bin/tools/    vendored python libs behind bin/* (kblib, yamlout, websearch)
 bin/cgo/      zig zcc zc++ (CGO via zig cc, not gcc)
@@ -65,15 +66,15 @@ 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
-  docling (isolated subprocess — its native onnx can segfault the parent).
-  Conversion never touches the brain DB (crash safety).
+  `pdftotext -layout` fast path (~15ms); textless/scanned PDFs use
+  `pdftoppm` + tesseract `eng+deu` (`bin/mail/ocr.go`). Optional
+  `OCR_ENGINE=paddle`. Conversion never touches the brain DB (crash safety).
 - `index_mail` is a deprecation shim for `bin/brain/index.go --rebuild`. Bulk
  rebuild still deletes `var/kb.lbug` and creates FTS/HNSW last. Single-leaf
  write is `bin/brain/add.go` (safe while indexes exist; do not DROP INDEX).
@@ -89,6 +90,7 @@ bin/kb/search "query" [--repo X]                  # deprecated wrapper → bin/b
 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
@@ -100,12 +102,16 @@ bin/git/import.go [REPO] [--json] [--limit N]     # go-git history → commit le
 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/qa/stats.go                                  # D22 DuckDB quantiles / JSONL (gcc CGO)
+bin/mail/ocr.go <image|pdf>                      # tesseract eng+deu (scans)
 bin/md/tables                                     # what the graph holds → YAML
 bin/brain/deduce "question"                       # thinking wrapper
 ```

 Never start a shell command with `cd` — use the tool working-directory
-parameter. Search before reading whole files.
+parameter. Search before reading whole files. For YAML/JSON/XML/CSV/TOML/HCL
+prefer mikefarah/yq (`skills/yq/SKILL.md`). For bulk rows and quantiles use
+duckdb-go (`internal/duckstats`, `skills/duckdb/SKILL.md`), not Ladybug.

 ## GitHub safety rules (ABSOLUTE — never violate)

@@ -16,6 +16,10 @@ 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/*

 COPY requirements.lock.txt /tmp/requirements.lock.txt
 RUN python -m pip install --no-cache-dir -r /tmp/requirements.lock.txt \
@@ -4,8 +4,9 @@ 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** — read path + MCP work; v1 goal is [epic #16](https://git.produktor.io/eSlider/2dph/issues/16)
-(milestone [v1 detective brain](https://git.produktor.io/eSlider/2dph/milestone/12)).
+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
@@ -47,6 +48,7 @@ detective method: **a fact needs ≥2 independent sources or it is
 | 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`/`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`. |
+| D22 | analytics | **duckdb-go** in-process (`internal/duckstats`, `bin/qa/stats.go`) for quantiles/JSONL. Links with **gcc/g++**, not Zig. Ladybug stays the graph; web-search cache stays modernc sqlite. Slice small structured docs with **mikefarah/yq**, not kislyuk/jq. [#30](https://git.produktor.io/eSlider/2dph/issues/30). |

 ## Architecture

@@ -74,6 +76,7 @@ detective method: **a fact needs ≥2 independent sources or it is
    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)
@@ -108,17 +111,20 @@ Common props on every node/edge: `root`, `confidence`, `evidence[]`, `how`,

 - `bin/{subject}/{method}` — line 2 is a usage comment (mirrors `psql-yq`).
 - bash + python primary; golang via Go shebang when a compiled helper is right.
- YAML default output, `--json` for machines. Slice with `yq`.
+- YAML default output, `--json` for machines. Slice with mikefarah/yq.
 - Everything that touches the network / DB is read-only, throttled, cached.
 - Tests (TDD) gate every commit; `gh` + CI/CD on every push.

 ## Open questions (v2)

 - OQ1: mutually-contradicting evidence — how to resolve (authority weighting,
-  temporal freshness, audit adjudication). **v2**; does not block epic #16.
- OQ2: OCR — poppler `pdftotext` fast-path exists; scans still docling.
-  [#6](https://git.produktor.io/eSlider/2dph/issues/6) (v2, does not block #16).
- OQ3: optional duckdb-md layer for `SELECT … FORMAT MARKDOWN` export/write-back.
+  temporal freshness, audit adjudication). **v2**; [#29](https://git.produktor.io/eSlider/2dph/issues/29).
+- OQ2: OCR — **in**. `pdftotext -layout` first; scans `pdftoppm` + tesseract
+  `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: **in** — duckdb-go (`internal/duckstats`, `bin/qa/stats.go`) for
+  quantiles / JSONL count. Not a second graph. [#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.

@@ -127,7 +133,8 @@ 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
@@ -144,8 +151,8 @@ 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` via Zig CGO.
-   CI SoT switch: [#19](https://git.produktor.io/eSlider/2dph/issues/19).
+5. `bin/brain/eval.go` via Zig (recall@5 ≥ 0.95). Python `bin/kb/eval` is an
+   explicit fallback, not the CI SoT.
 6. `bin/cgo/zig go build -tags system_ladybug` (compile search with zig cc; fetches pinned zig+libs).

 Feedback loop: every commit → PR → CI → green/gate → merge. Same discipline as
@@ -159,13 +166,12 @@ Feedback loop: every commit → PR → CI → green/gate → merge. Same discipl
 4. .venv: ladybug + model2vec + mistune
 5. schema + tools with TDD (kb + md + facts + brain)
 6. ~/.config/brain config
-7. corpus extraction (facts/info) — **open**: [#18](https://git.produktor.io/eSlider/2dph/issues/18)
+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)

-Read path + MCP are in. Incremental `brain/add` and `--hop` are in. Remaining:
-facts+chats corpus on rebuild, and CI eval SoT. Board:
+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).
@@ -174,8 +180,8 @@ Narrative: [docs/roadmap.md](docs/roadmap.md).
 |-------|-------|-----|
 | 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) | Rebuild is mostly `info` (repo md + mail). `facts/extract` and chats are not a first-class index input. WhatsApp sync is a stub. |
+| 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) | GitHub CI recall still runs Python `bin/kb/eval`. |
+| 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: [#6](https://git.produktor.io/eSlider/2dph/issues/6) OCR, OQ1, OQ3, OQ4.
+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**.
@@ -121,6 +121,7 @@ Mail is a first-class corpus (retrievable through the same search):
 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
 ```
@@ -168,6 +169,9 @@ 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
@@ -3,7 +3,7 @@
 //
 // 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
 //
@@ -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)
@@ -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, link_from_file,
+    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
@@ -96,12 +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(
@@ -115,44 +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():
-        DB_PATH.unlink()
+    dbpath.parent.mkdir(parents=True, exist_ok=True)
+    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


@@ -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
-already present (message.md exists) is skipped unless --force.
+Requires ONLYOFFICE_URL/USER/PASS in .env (or env) except `--from-raw`.
+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]
@@ -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
+}
@@ -0,0 +1,73 @@
+//usr/bin/env go run -tags=qa_stats "$0" "$@"; exit
+//go:build qa_stats
+//
+// bin/qa/stats.go - DuckDB quantiles over a JSON number array or JSONL count.
+//
+//	./bin/qa/stats.go <<< '[1,2,3,4,5]'
+//	./bin/qa/stats.go --jsonl rows.jsonl
+//
+// NOTE: never run `gofmt -w` on this file — it breaks the shebang.
+// DuckDB CGO needs gcc/g++ (not Zig). After eval "$(bin/cgo/zig env)":
+//   CC=gcc CXX=g++ CGO_CFLAGS= CGO_LDFLAGS= ./bin/qa/stats.go
+package main
+
+import (
+	"encoding/json"
+	"fmt"
+	"io"
+	"os"
+	"strings"
+
+	"github.com/eSlider/2dph/internal/duckstats"
+)
+
+func main() {
+	os.Exit(run(os.Args[1:]))
+}
+
+func run(args []string) int {
+	jsonl := ""
+	for i := 0; i < len(args); i++ {
+		a := args[i]
+		switch {
+		case a == "--jsonl" && i+1 < len(args):
+			i++
+			jsonl = args[i]
+		case strings.HasPrefix(a, "--jsonl="):
+			jsonl = strings.TrimPrefix(a, "--jsonl=")
+		case a == "-h" || a == "--help":
+			fmt.Fprintln(os.Stderr, "bin/qa/stats.go [--jsonl FILE]  # stdin = JSON [float,…]")
+			return 0
+		default:
+			fmt.Fprintln(os.Stderr, "unknown arg:", a)
+			return 2
+		}
+	}
+	if jsonl != "" {
+		n, err := duckstats.CountJSONL(jsonl)
+		if err != nil {
+			fmt.Fprintln(os.Stderr, err)
+			return 1
+		}
+		fmt.Printf("n: %d\n", n)
+		return 0
+	}
+	raw, err := io.ReadAll(os.Stdin)
+	if err != nil {
+		fmt.Fprintln(os.Stderr, err)
+		return 1
+	}
+	var samples []float64
+	if err := json.Unmarshal(raw, &samples); err != nil {
+		fmt.Fprintln(os.Stderr, err)
+		return 1
+	}
+	s, err := duckstats.Quantiles(samples)
+	if err != nil {
+		fmt.Fprintln(os.Stderr, err)
+		return 1
+	}
+	fmt.Printf("n: %d\nmin: %g\np50: %g\np95: %g\nmax: %g\navg: %g\n",
+		s.N, s.Min, s.P50, s.P95, s.Max, s.Avg)
+	return 0
+}
@@ -7,7 +7,7 @@
 //	./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.
+// PicoClaw is compose profile picoclaw; tool names match internal/httpapi MCP ops.
 // NOTE: never run `gofmt -w` on this file — it breaks the shebang.
 package main

@@ -17,6 +17,7 @@ import (
 	"os"
 	"strings"

+	"github.com/eSlider/2dph/internal/duckstats"
 	"github.com/eSlider/2dph/internal/reasoner"
 )

@@ -61,6 +62,14 @@ func run(args []string) int {
 	}
 	c := reasoner.Client{BaseURL: base, Model: model, Device: device}
 	rep := reasoner.Run(c)
+	lat := make([]float64, 0, len(rep.Prompts))
+	for _, p := range rep.Prompts {
+		lat = append(lat, float64(p.LatencyMS))
+	}
+	if st, err := duckstats.Quantiles(lat); err == nil {
+		rep.LatencyP50MS = st.P50
+		rep.LatencyP95MS = st.P95
+	}
 	raw, err := json.MarshalIndent(rep, "", "  ")
 	if err != nil {
 		fmt.Fprintln(os.Stderr, err)
@@ -76,6 +85,8 @@ func run(args []string) int {
 		fmt.Printf("xml_leak: %d\n", rep.XMLLeak)
 		fmt.Printf("rss_mb: %d\n", rep.RSSMB)
 		fmt.Printf("vram_mb: %d\n", rep.VRAMMB)
+		fmt.Printf("latency_p50_ms: %g\n", rep.LatencyP50MS)
+		fmt.Printf("latency_p95_ms: %g\n", rep.LatencyP95MS)
 		for _, p := range rep.Prompts {
 			status := "fail"
 			if p.OK {
@@ -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()
@@ -136,6 +136,33 @@ class BinLayoutTest(unittest.TestCase):
            "index_mail must point at bin/brain/index.go",
        )

+    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()
@@ -190,6 +217,32 @@ class BinLayoutTest(unittest.TestCase):
            if "go-git/go-git" in line:
                self.assertNotIn("indirect", line)

+    def test_duckdb_go_is_direct_require(self) -> None:
+        text = (ROOT / "go.mod").read_text()
+        first = text.split("require (")[1].split(")")[0]
+        self.assertRegex(first, r"github.com/duckdb/duckdb-go/v2\s+v")
+        for line in first.splitlines():
+            if "duckdb/duckdb-go" in line:
+                self.assertNotIn("indirect", line)
+        skill = (ROOT / "skills" / "duckdb" / "SKILL.md").read_text()
+        self.assertIn("github.com/duckdb/duckdb-go", skill)
+        self.assertIn("Ladybug", skill)
+        self.assertIn("sqlite", skill.lower())
+        self.assertIn("gcc", skill.lower())
+        self.assertIn("Zig", skill)
+        plan = (ROOT / "PLAN.md").read_text()
+        self.assertIn("D22", plan)
+        self.assertIn("duckdb-go", plan)
+        self._assert_shebang("bin/qa/stats.go")
+        reasoner = (ROOT / "internal" / "reasoner" / "client.go").read_text()
+        self.assertNotIn("duckdb", reasoner)
+        self.assertNotIn("duckstats", reasoner)
+        bakeoff = (ROOT / "bin" / "reasoner" / "bakeoff.go").read_text()
+        self.assertIn("internal/duckstats", bakeoff)
+        webcache = (ROOT / "internal" / "websearch" / "cache.go").read_text()
+        self.assertNotIn("duckdb", webcache)
+        self.assertIn("modernc.org/sqlite", webcache)
+
    def test_cgo_uses_zig_not_gcc(self) -> None:
        for rel in ("bin/cgo/zig", "bin/cgo/zcc", "bin/cgo/zc++"):
            p = ROOT / rel
@@ -207,3 +260,25 @@ class BinLayoutTest(unittest.TestCase):
        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")
@@ -39,3 +39,59 @@ class IndexAdapterTest(unittest.TestCase):
        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()
@@ -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:
@@ -121,7 +121,7 @@ class PublishedDocsTest(unittest.TestCase):
        self.assertIn("D18", plan)
        self.assertIn("Qwen/Qwen3.5-9B", plan)
        compose = (ROOT / "compose.yaml").read_text()
-        self.assertIn('profiles: ["reasoner"]', compose)
+        self.assertIn('"reasoner"', compose)
        self.assertIn("OLLAMA_NUM_GPU", compose)
        self.assertIn("127.0.0.1:11435", compose)
        dockerfile = (ROOT / "Dockerfile").read_text()
@@ -47,3 +47,17 @@ class SkillsTest(unittest.TestCase):
        self.assertIn("throttled", skill.lower())
        self.assertIn("not a negative finding", agents)
        self.assertIn("Fact-check every", agents)
+
+    def test_yq_is_mikefarah_for_structured_data(self) -> None:
+        skill = (ROOT / "skills" / "yq" / "SKILL.md").read_text()
+        self.assertIn("https://github.com/mikefarah/yq", skill)
+        for fmt in ("YAML", "JSON", "XML", "CSV", "TOML", "HCL"):
+            self.assertIn(fmt, skill)
+        self.assertIn("not kislyuk", skill.lower())
+        plan = (ROOT / "PLAN.md").read_text()
+        self.assertIn("mikefarah/yq", plan)
+        agents = (ROOT / "AGENTS.md").read_text()
+        self.assertIn("mikefarah/yq", agents)
+        web = (ROOT / "skills" / "web-search" / "SKILL.md").read_text()
+        self.assertIn("| yq ", web)
+        self.assertNotIn("| jq ", web)
@@ -0,0 +1,36 @@
+"""qa/system_perf.py is an offline-gated system test (no live brain in CI)."""
+from __future__ import annotations
+
+import ast
+import unittest
+from pathlib import Path
+
+ROOT = Path(__file__).resolve().parents[2]
+
+
+class SystemPerfScriptTest(unittest.TestCase):
+    def test_script_compiles_and_is_read_only(self) -> None:
+        path = ROOT / "qa" / "system_perf.py"
+        src = path.read_text()
+        compile(src, str(path), "exec")
+        self.assertIn("--json", src)
+        self.assertIn("qwen3.5:9b", src)
+        self.assertIn("--picoclaw", src)
+        self.assertIn("BRAIN_URL", src)
+        self.assertIn("tools/list", src)
+        self.assertIn("tools/call", src)
+        self.assertIn("GATE_HEALTH_MS", src)
+        self.assertIn("GATE_GET_P50_MS", src)
+        self.assertNotIn("kb.lbug", src)
+        self.assertNotIn("password", src.lower())
+        self.assertNotIn("token", src.lower())
+
+    def test_script_does_not_write_ladybug(self) -> None:
+        tree = ast.parse((ROOT / "qa" / "system_perf.py").read_text())
+        writes = [
+            n.func.attr
+            for n in ast.walk(tree)
+            if isinstance(n, ast.Call) and isinstance(n.func, ast.Attribute)
+            and n.func.attr in {"write_text", "write_bytes", "dump"}
+        ]
+        self.assertEqual(writes, [], f"system_perf must not write files: {writes}")
@@ -2,14 +2,23 @@
 #
 #   docker compose up -d brain                  # API (Zig CGO serve)
 #   docker compose --profile index run --rm index   # Python rebuild
-#   docker compose --profile picoclaw up brain-mcp
+#   docker compose --profile picoclaw up -d     # brain-mcp + CPU reasoner + PicoClaw gateway
 #   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
 #
 # Secrets never baked in: search.env + db-profiles.yml from ~/.config/brain.

 name: 2dph

+networks:
+  default:
+    name: 2dph_sys
+    driver: bridge
+    ipam:
+      config:
+        - subnet: 10.23.42.0/24
+
 services:
  brain:
    image: ghcr.io/eslider/2dph:api
@@ -97,8 +106,8 @@ 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
+  # MCP endpoint for PicoClaw (and any MCP client).
+  #   docker compose --profile picoclaw up -d
  brain-mcp:
    profiles: ["picoclaw"]
    image: ghcr.io/eslider/2dph:api
@@ -120,7 +129,7 @@ services:
  #   docker compose --profile reasoner up -d reasoner
  #   docker compose --profile reasoner exec reasoner ollama pull qwen3.5:9b
  reasoner:
-    profiles: ["reasoner"]
+    profiles: ["reasoner", "picoclaw"]
    image: docker.io/ollama/ollama:latest
    environment:
      OLLAMA_NUM_GPU: "0"
@@ -131,7 +140,37 @@ services:
      - reasoner-ollama:/root/.ollama
    restart: unless-stopped

+  # Official PicoClaw gateway. Config has no secrets (Ollama + HTTP MCP).
+  # Host network: brain/reasoner bind 127.0.0.1 only, so host.docker.internal
+  # (docker0) cannot reach them. Gateway 127.0.0.1:18790 (not the 18800 launcher).
+  # If :8630/:11435 are already bound, do not start brain-mcp/reasoner:
+  #   docker compose --profile picoclaw up -d --no-deps picoclaw
+  picoclaw:
+    profiles: ["picoclaw"]
+    image: docker.io/sipeed/picoclaw:v0.3.1
+    network_mode: host
+    depends_on:
+      - brain-mcp
+      - reasoner
+    environment:
+      PICOCLAW_GATEWAY_HOST: "127.0.0.1"
+    entrypoint: ["picoclaw", "gateway"]
+    volumes:
+      - picoclaw-home:/root/.picoclaw
+      - ./deploy/picoclaw/config.json:/root/.picoclaw/config.json:ro
+    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:
+  picoclaw-home:
@@ -0,0 +1,33 @@
+{
+  "agents": {
+    "defaults": {
+      "model_name": "qwen3.5-9b",
+      "max_tool_iterations": 8,
+      "max_tokens": 512,
+      "context_window": 8192
+    }
+  },
+  "model_list": [
+    {
+      "model_name": "qwen3.5-9b",
+      "model": "ollama/qwen3.5:9b",
+      "api_base": "http://127.0.0.1:11435/v1",
+      "request_timeout": 600
+    }
+  ],
+  "tools": {
+    "web": {
+      "enabled": false
+    },
+    "mcp": {
+      "enabled": true,
+      "servers": {
+        "2dph": {
+          "enabled": true,
+          "type": "http",
+          "url": "http://127.0.0.1:8630/mcp"
+        }
+      }
+    }
+  }
+}
@@ -20,7 +20,7 @@ Evidence-first knowledge graph. Facts need proof or they are
 | 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 |
+| reference | [PLAN.md](../PLAN.md) — decisions D1–D22 |

 Decisions the public face must name: **D3** SearXNG compose, **D6** Go service /
 Python write sidecar, **D14** `bin/{subject}/{method}.go`, **D15** Gitea origin,
@@ -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.
@@ -1,18 +1,34 @@
 # 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.
+2dph is the memory/fact gate. Compose profile `picoclaw` runs the official
+PicoClaw gateway (`docker.io/sipeed/picoclaw:v0.3.1`) plus `brain-mcp` and the
+CPU reasoner. Default agent model is `qwen3.5:9b` (RAM path, D18). Weights stay
+in the reasoner volume, not in the 2dph image.
+No secrets in git: Ollama needs no key; MCP is local HTTP.

 ```bash
-docker compose --profile picoclaw up brain-mcp
+docker compose --profile picoclaw up -d
+# already serving :8630 / :11435:
+docker compose --profile picoclaw up -d --no-deps picoclaw
 ```

-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).
+Gateway: `127.0.0.1:18790`. Brain MCP: `http://127.0.0.1:8630/mcp`.
+Cursor-style clients can use [deploy/picoclaw/mcp.json.example](../deploy/picoclaw/mcp.json.example).
+PicoClaw itself uses [deploy/picoclaw/config.json](../deploy/picoclaw/config.json)
+(`127.0.0.1` + host network — loopback publishes are not reachable via docker0).

 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.
+System performance (MCP gates + qwen3.5:9b tool_call + PicoClaw gateway):
+
+```bash
+./qa/system_perf.py --json | yq '.gates'
+REASONER_MODEL=qwen3.5:9b ./qa/system_perf.py --reasoner --picoclaw --json | yq '.reasoner'
+```
+
+The default agent model is `qwen3.5:9b`. PicoClaw `context_window` is 8192
+(heuristic `max_tokens*4` at 512 is 2048, too small for MCP tool schemas).
+`request_timeout` is 600s for a CPU turn (tool_call + MCP search + answer).
@@ -1,8 +1,8 @@
 # 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`).
+compose profile `picoclaw` (`sipeed/picoclaw`); the bake-off hits the same
+tool names (`search` → `get` → `audit` from `internal/httpapi.Ops`).

 ```bash
 docker compose --profile reasoner up -d reasoner
@@ -11,6 +11,8 @@ REASONER_BASE_URL=http://127.0.0.1:11435/v1 REASONER_MODEL=qwen3.5:9b \
  ./bin/reasoner/bakeoff.go --json
 ```

+JSON includes `latency_p50_ms` / `latency_p95_ms` from DuckDB (`internal/duckstats`, D22).
+
 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.

@@ -28,9 +28,23 @@ Compose `api` (no CPython) / `index` (Python write). Issues #1–#5, #7–#13.
 `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 — **in**.
+[#30](https://git.produktor.io/eSlider/2dph/issues/30) OQ3 duckdb-go — **in**.
+[#29](https://git.produktor.io/eSlider/2dph/issues/29) OQ1 contradiction
+resolution.

 ## Blockers

+None for epic #16 (closed). Remaining v2: OQ1, OQ4.
+
 ```
 question
    │
@@ -39,22 +53,14 @@ question
    ├─ web (D17)           ← in
    ├─ brain/add ACID      ← in
    ├─ Cypher hop          ← in
-    └─ facts+chats corpus  ← #18
+    └─ facts+chats corpus  ← in
 ```

-1. **[#18](https://git.produktor.io/eSlider/2dph/issues/18) corpus** —
-   rebuild loads repo markdown + mail as `info`. `facts/extract` pairing
-   and `bin/chats` are not indexed. WhatsApp is a stub. PII stays in `var/`.
-2. **[#19](https://git.produktor.io/eSlider/2dph/issues/19) CI eval** —
-   recall SoT should be `bin/brain/eval.go` via Zig, not Python `bin/kb/eval`.
-
 ## Not v1

-[#6](https://git.produktor.io/eSlider/2dph/issues/6) OCR (OQ2), OQ1
-contradiction resolution, OQ3 duckdb-md export, OQ4 YAML-first leafs.
+OQ1 contradiction resolution, OQ4 YAML-first leafs.
+OCR (OQ2) and duckdb-go (OQ3/D22) are in.

 ## Close epic #16 when

- ops pairing + chat import land as leafs on rebuild
- MCP tool order is documented and still gated by tests
- CI recall SoT is `bin/brain/eval.go` via Zig
+Children #14, #15, #17, #18, #19 are closed. MCP tool order stays gated by tests.
@@ -16,6 +16,7 @@ No laptop-absolute paths. Config lives in env files under `$HOME/.config/brain/`
 - 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
@@ -49,7 +50,7 @@ corpus rebuild remains `bin/brain/index.go --rebuild` (Compose profile

 ```bash
 bin/brain/add.go --text "arc-1 runs Matrix" --root facts --source "compose.yml x docker ps"
-bin/brain/index.go --rebuild
+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
@@ -7,6 +7,7 @@ require (
 	github.com/arran4/golang-ical v0.3.5
 	github.com/chewxy/math32 v1.11.2
 	github.com/daulet/tokenizers v1.27.0
+	github.com/duckdb/duckdb-go/v2 v2.10505.0
 	github.com/go-git/go-git/v5 v5.19.2
 	golang.org/x/sys v0.47.0
 	golang.org/x/text v0.40.0
@@ -20,10 +21,17 @@ require (
 	github.com/apache/arrow-go/v18 v18.6.0 // indirect
 	github.com/cloudflare/circl v1.6.3 // indirect
 	github.com/cyphar/filepath-securejoin v0.6.1 // indirect
+	github.com/duckdb/duckdb-go-bindings v0.10505.0 // indirect
+	github.com/duckdb/duckdb-go-bindings/lib/darwin-amd64 v0.10505.0 // indirect
+	github.com/duckdb/duckdb-go-bindings/lib/darwin-arm64 v0.10505.0 // indirect
+	github.com/duckdb/duckdb-go-bindings/lib/linux-amd64 v0.10505.0 // indirect
+	github.com/duckdb/duckdb-go-bindings/lib/linux-arm64 v0.10505.0 // indirect
+	github.com/duckdb/duckdb-go-bindings/lib/windows-amd64 v0.10505.0 // indirect
 	github.com/dustin/go-humanize v1.0.1 // indirect
 	github.com/emirpasic/gods v1.18.1 // indirect
 	github.com/go-git/gcfg v1.5.1-0.20230307220236-3a3c6141e376 // indirect
 	github.com/go-git/go-billy/v5 v5.9.0 // indirect
+	github.com/go-viper/mapstructure/v2 v2.5.0 // indirect
 	github.com/goccy/go-json v0.10.6 // indirect
 	github.com/golang/groupcache v0.0.0-20241129210726-2c02b8208cf8 // indirect
 	github.com/google/flatbuffers v25.12.19+incompatible // indirect
@@ -31,6 +31,20 @@ github.com/davecgh/go-spew v1.1.0/go.mod h1:J7Y8YcW2NihsgmVo/mv3lAwl/skON4iLHjSs
 github.com/davecgh/go-spew v1.1.1/go.mod h1:J7Y8YcW2NihsgmVo/mv3lAwl/skON4iLHjSsI+c5H38=
 github.com/davecgh/go-spew v1.1.2-0.20180830191138-d8f796af33cc h1:U9qPSI2PIWSS1VwoXQT9A3Wy9MM3WgvqSxFWenqJduM=
 github.com/davecgh/go-spew v1.1.2-0.20180830191138-d8f796af33cc/go.mod h1:J7Y8YcW2NihsgmVo/mv3lAwl/skON4iLHjSsI+c5H38=
+github.com/duckdb/duckdb-go-bindings v0.10505.0 h1:/0pPsTLrcCsTGxT0VrHgJWnOcPe1tQL1vrki1v3jbAI=
+github.com/duckdb/duckdb-go-bindings v0.10505.0/go.mod h1:HoD5xePkDj3VZbBnVVfxVVYIljZ9khCprWA7FgwIiC4=
+github.com/duckdb/duckdb-go-bindings/lib/darwin-amd64 v0.10505.0 h1:FrMqquFBQlMsi34h2KZgCku54rqA8xEbXZ0NLVDKwYs=
+github.com/duckdb/duckdb-go-bindings/lib/darwin-amd64 v0.10505.0/go.mod h1:EnAvZh1kNJHp5yF+M1ZHNEvapnmt6anq1xXHVrAGqMo=
+github.com/duckdb/duckdb-go-bindings/lib/darwin-arm64 v0.10505.0 h1:lbRbpQwT1MmUhh/VTwukV9K8bxKByV3UghAP3MvsbBo=
+github.com/duckdb/duckdb-go-bindings/lib/darwin-arm64 v0.10505.0/go.mod h1:IGLSeEcFhNeZF16aVjQCULD7TsFZKG5G7SyKJAXKp5c=
+github.com/duckdb/duckdb-go-bindings/lib/linux-amd64 v0.10505.0 h1:nrsaVYj3XYCRbS2FpdOMD/KHE7egRMr+/NR1IHmjT84=
+github.com/duckdb/duckdb-go-bindings/lib/linux-amd64 v0.10505.0/go.mod h1:KAIynZ0GHCS7X5fRyuFnQMg/SZBPK/bS9OCOVojClxw=
+github.com/duckdb/duckdb-go-bindings/lib/linux-arm64 v0.10505.0 h1:qM6oGDgwXBILJGbTY4fCy6QOczLpucUA6yn6g3ORjh4=
+github.com/duckdb/duckdb-go-bindings/lib/linux-arm64 v0.10505.0/go.mod h1:81SGOYoEUs8qaAfSk1wRfM5oobrIJ5KI7AzYhK6/bvQ=
+github.com/duckdb/duckdb-go-bindings/lib/windows-amd64 v0.10505.0 h1:DjqZl9rYreHkSOqnqLmkrqH5T8UdQNcxZLJVZzGmXXA=
+github.com/duckdb/duckdb-go-bindings/lib/windows-amd64 v0.10505.0/go.mod h1:K25pJL26ARblGDeuAkrdblFvUen92+CwksLtPEHRqqQ=
+github.com/duckdb/duckdb-go/v2 v2.10505.0 h1:SWwvLn2Qx/RQSnQNupwgIF8VbnJ5A6OQU9lYb/mDETI=
+github.com/duckdb/duckdb-go/v2 v2.10505.0/go.mod h1:m0PW4J4FG9hlFlVdXi6Ds9owpyIDaBdE2jyce00fGcE=
 github.com/dustin/go-humanize v1.0.1 h1:GzkhY7T5VNhEkwH0PVJgjz+fX1rhBrR7pRT3mDkpeCY=
 github.com/dustin/go-humanize v1.0.1/go.mod h1:Mu1zIs6XwVuF/gI1OepvI0qD18qycQx+mFykh5fBlto=
 github.com/elazarl/goproxy v1.7.2 h1:Y2o6urb7Eule09PjlhQRGNsqRfPmYI3KKQLFpCAV3+o=
@@ -47,6 +61,8 @@ github.com/go-git/go-git-fixtures/v4 v4.3.2-0.20231010084843-55a94097c399 h1:eMj
 github.com/go-git/go-git-fixtures/v4 v4.3.2-0.20231010084843-55a94097c399/go.mod h1:1OCfN199q1Jm3HZlxleg+Dw/mwps2Wbk9frAWm+4FII=
 github.com/go-git/go-git/v5 v5.19.2 h1:wkfn7vOlUBu8ivAWKBWisTiwJK4jYHzTF8Ndv1LyGqY=
 github.com/go-git/go-git/v5 v5.19.2/go.mod h1:QqCBE1EFN5ddFmrliLQ3/ntRCUjZU3EJuwuB/jWEHjk=
+github.com/go-viper/mapstructure/v2 v2.5.0 h1:vM5IJoUAy3d7zRSVtIwQgBj7BiWtMPfmPEgAXnvj1Ro=
+github.com/go-viper/mapstructure/v2 v2.5.0/go.mod h1:oJDH3BJKyqBA2TXFhDsKDGDTlndYOZ6rGS0BRZIxGhM=
 github.com/goccy/go-json v0.10.6 h1:p8HrPJzOakx/mn/bQtjgNjdTcN+/S6FcG2CTtQOrHVU=
 github.com/goccy/go-json v0.10.6/go.mod h1:oq7eo15ShAhp70Anwd5lgX2pLfOS3QCiwU/PULtXL6M=
 github.com/golang/groupcache v0.0.0-20241129210726-2c02b8208cf8 h1:f+oWsMOmNPc8JmEHVZIycC7hBoQxHH9pNKQORJNozsQ=
@@ -0,0 +1,47 @@
+// Package duckstats runs in-process DuckDB for columnar aggregates.
+// Graph facts stay in Ladybug. Web-search KV cache stays modernc sqlite.
+package duckstats
+
+import (
+	"database/sql"
+	"fmt"
+
+	_ "github.com/duckdb/duckdb-go/v2"
+)
+
+type Stats struct {
+	N   int     `json:"n"`
+	Min float64 `json:"min"`
+	P50 float64 `json:"p50"`
+	P95 float64 `json:"p95"`
+	Max float64 `json:"max"`
+	Avg float64 `json:"avg"`
+}
+
+func Quantiles(samples []float64) (Stats, error) {
+	if len(samples) == 0 {
+		return Stats{}, fmt.Errorf("duckstats: empty samples")
+	}
+	db, err := sql.Open("duckdb", "")
+	if err != nil {
+		return Stats{}, err
+	}
+	defer db.Close()
+	var s Stats
+	err = db.QueryRow(`
+SELECT count(v), min(v), quantile_cont(v, 0.5), quantile_cont(v, 0.95), max(v), avg(v)
+FROM (SELECT unnest(?) AS v)`, samples).Scan(
+		&s.N, &s.Min, &s.P50, &s.P95, &s.Max, &s.Avg)
+	return s, err
+}
+
+func CountJSONL(path string) (int64, error) {
+	db, err := sql.Open("duckdb", "")
+	if err != nil {
+		return 0, err
+	}
+	defer db.Close()
+	var n int64
+	err = db.QueryRow(`SELECT count(*) FROM read_json_auto(?)`, path).Scan(&n)
+	return n, err
+}
@@ -0,0 +1,51 @@
+package duckstats
+
+import (
+	"os"
+	"testing"
+)
+
+func TestQuantilesEmpty(t *testing.T) {
+	_, err := Quantiles(nil)
+	if err == nil {
+		t.Fatal("empty slice must error")
+	}
+}
+
+func TestQuantilesOdd(t *testing.T) {
+	s, err := Quantiles([]float64{1, 2, 3, 4, 5})
+	if err != nil {
+		t.Fatal(err)
+	}
+	if s.N != 5 {
+		t.Fatalf("n=%d", s.N)
+	}
+	if s.Min != 1 || s.Max != 5 {
+		t.Fatalf("min=%v max=%v", s.Min, s.Max)
+	}
+	if s.P50 != 3 {
+		t.Fatalf("p50=%v want 3", s.P50)
+	}
+	if s.Avg != 3 {
+		t.Fatalf("avg=%v want 3", s.Avg)
+	}
+	if s.P95 < 4.5 || s.P95 > 5 {
+		t.Fatalf("p95=%v want in [4.5,5]", s.P95)
+	}
+}
+
+func TestCountJSONL(t *testing.T) {
+	dir := t.TempDir()
+	p := dir + "/rows.jsonl"
+	body := "{\"ms\":1}\n{\"ms\":2}\n{\"ms\":3}\n"
+	if err := os.WriteFile(p, []byte(body), 0o600); err != nil {
+		t.Fatal(err)
+	}
+	n, err := CountJSONL(p)
+	if err != nil {
+		t.Fatal(err)
+	}
+	if n != 3 {
+		t.Fatalf("count=%d want 3", n)
+	}
+}
@@ -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
+}
@@ -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)
+	}
+}
@@ -113,6 +113,8 @@ type Report struct {
 	XMLLeak      int      `json:"xml_leak"`
 	RSSMB        int      `json:"rss_mb"`
 	VRAMMB       int      `json:"vram_mb"`
+	LatencyP50MS float64  `json:"latency_p50_ms,omitempty"`
+	LatencyP95MS float64  `json:"latency_p95_ms,omitempty"`
 	Prompts      []Result `json:"prompts"`
 }

@@ -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",
@@ -0,0 +1,257 @@
+#!/usr/bin/env python3
+"""System performance test: PicoClaw surface (brain MCP) + optional reasoner.
+
+  BRAIN_URL=http://127.0.0.1:8630 ./qa/system_perf.py --json
+  REASONER_BASE_URL=http://127.0.0.1:11435/v1 REASONER_MODEL=qwen3.5:9b \\
+    ./qa/system_perf.py --reasoner --picoclaw --json
+
+Does not write Ladybug. Search includes web (D17); expect ~10s+ per search.
+Exit 1 if health/get/audit gates fail. Reasoner is measured, not gated.
+"""
+from __future__ import annotations
+
+import argparse
+import json
+import os
+import statistics
+import sys
+import time
+import urllib.error
+import urllib.request
+from concurrent.futures import ThreadPoolExecutor
+
+DEFAULT_BRAIN = "http://127.0.0.1:8630"
+DEFAULT_REASONER = "http://127.0.0.1:11435/v1"
+DEFAULT_MODEL = "qwen3.5:9b"
+DEFAULT_PICOCLAW = "http://127.0.0.1:18790"
+
+GATE_HEALTH_MS = 500
+GATE_GET_P50_MS = 50
+GATE_AUDIT_P50_MS = 50
+
+
+def _req(url: str, data: bytes | None = None, timeout: float = 90) -> bytes:
+    headers = {"Content-Type": "application/json"} if data is not None else {}
+    req = urllib.request.Request(url, data=data, headers=headers)
+    with urllib.request.urlopen(req, timeout=timeout) as res:
+        return res.read()
+
+
+def timed(fn):
+    t0 = time.perf_counter()
+    out = fn()
+    return (time.perf_counter() - t0) * 1000.0, out
+
+
+def stats(samples: list[float]) -> dict:
+    s = sorted(samples)
+    n = len(s)
+    return {
+        "n": n,
+        "min_ms": round(s[0], 1),
+        "p50_ms": round(s[n // 2], 1),
+        "p95_ms": round(s[min(n - 1, int(n * 0.95))], 1),
+        "max_ms": round(s[-1], 1),
+        "avg_ms": round(statistics.mean(s), 1),
+    }
+
+
+def mcp(brain: str, method: str, params=None, timeout: float = 90) -> dict:
+    payload: dict = {"jsonrpc": "2.0", "id": 1, "method": method}
+    if params is not None:
+        payload["params"] = params
+    raw = _req(brain.rstrip("/") + "/mcp", json.dumps(payload).encode(), timeout=timeout)
+    return json.loads(raw.decode())
+
+
+def mcp_call(brain: str, name: str, arguments: dict, timeout: float = 90) -> tuple[bool, str]:
+    d = mcp(brain, "tools/call", {"name": name, "arguments": arguments}, timeout=timeout)
+    res = d.get("result") or {}
+    text = ((res.get("content") or [{}])[0].get("text") or "")
+    return (not res.get("isError")), text
+
+
+def reasoner_tool_call(base: str, model: str, user: str) -> str:
+    payload = {
+        "model": model,
+        "messages": [
+            {"role": "system", "content": "You are PicoClaw. Always call search before answering."},
+            {"role": "user", "content": user},
+        ],
+        "tools": [
+            {
+                "type": "function",
+                "function": {
+                    "name": "search",
+                    "description": "deduction search",
+                    "parameters": {
+                        "type": "object",
+                        "properties": {"q": {"type": "string"}},
+                        "required": ["q"],
+                    },
+                },
+            }
+        ],
+        "tool_choice": "required",
+    }
+    raw = _req(
+        base.rstrip("/") + "/chat/completions",
+        json.dumps(payload).encode(),
+        timeout=600,
+    )
+    chat = json.loads(raw.decode())
+    tcs = chat["choices"][0]["message"].get("tool_calls") or []
+    if not tcs:
+        return ""
+    return tcs[0]["function"]["name"]
+
+
+def run(args: argparse.Namespace) -> dict:
+    brain = args.brain.rstrip("/")
+    report: dict = {
+        "brain": brain,
+        "device": "cpu",
+        "ok": True,
+        "gates": {},
+        "mcp": {},
+    }
+    ms, _ = timed(lambda: _req(brain + "/health", timeout=5))
+    report["mcp"]["health"] = {"n": 1, "avg_ms": round(ms, 1)}
+    report["gates"]["health"] = ms <= GATE_HEALTH_MS
+    if ms > GATE_HEALTH_MS:
+        report["ok"] = False
+
+    list_ms = []
+    for _ in range(args.n):
+        ms, d = timed(lambda: mcp(brain, "tools/list", timeout=10))
+        names = [t["name"] for t in ((d.get("result") or {}).get("tools") or [])]
+        if "search" not in names:
+            report["ok"] = False
+        list_ms.append(ms)
+    report["mcp"]["tools_list"] = stats(list_ms)
+
+    audit_ms = []
+    for _ in range(args.n):
+        ms, (ok, _) = timed(lambda: mcp_call(brain, "audit", {}))
+        if not ok:
+            report["ok"] = False
+        audit_ms.append(ms)
+    report["mcp"]["audit"] = stats(audit_ms)
+    report["gates"]["audit_p50"] = report["mcp"]["audit"]["p50_ms"] <= GATE_AUDIT_P50_MS
+    if not report["gates"]["audit_p50"]:
+        report["ok"] = False
+
+    ok, text = mcp_call(brain, "search", {"q": "LadybugDB", "n": 2}, timeout=90)
+    inner = json.loads(text) if ok else {}
+    hits = inner.get("results") or []
+    leaf_id = hits[0]["id"] if hits else ""
+    report["mcp"]["search_seed"] = {
+        "ok": ok,
+        "count": inner.get("count"),
+        "web": (inner.get("web") or {}).get("status"),
+    }
+
+    get_ms = []
+    if leaf_id:
+        for _ in range(args.n):
+            ms, (ok, _) = timed(lambda: mcp_call(brain, "get", {"id": leaf_id, "body": True}))
+            if not ok:
+                report["ok"] = False
+            get_ms.append(ms)
+        report["mcp"]["get"] = stats(get_ms)
+        report["gates"]["get_p50"] = report["mcp"]["get"]["p50_ms"] <= GATE_GET_P50_MS
+        if not report["gates"]["get_p50"]:
+            report["ok"] = False
+
+        def one_get() -> float:
+            t0 = time.perf_counter()
+            mcp_call(brain, "get", {"id": leaf_id, "body": True})
+            return (time.perf_counter() - t0) * 1000.0
+
+        t0 = time.perf_counter()
+        with ThreadPoolExecutor(max_workers=8) as ex:
+            conc = list(ex.map(lambda _: one_get(), range(8)))
+        wall = (time.perf_counter() - t0) * 1000.0
+        report["mcp"]["get_concurrent_8"] = {**stats(conc), "wall_ms": round(wall, 1)}
+
+    search_ms = []
+    for q in ("LadybugDB", "model2vec"):
+        ms, (ok, text) = timed(lambda q=q: mcp_call(brain, "search", {"q": q, "n": 3}, timeout=90))
+        inner = json.loads(text) if ok else {}
+        search_ms.append(ms)
+        report.setdefault("mcp", {}).setdefault("search_samples", []).append(
+            {
+                "q": q,
+                "ms": round(ms, 1),
+                "ok": ok,
+                "count": inner.get("count"),
+                "web": (inner.get("web") or {}).get("status"),
+            }
+        )
+    if search_ms:
+        report["mcp"]["search"] = stats(search_ms)
+
+    if args.reasoner:
+        base = args.reasoner_url
+        model = args.model
+        report["reasoner"] = {"base_url": base, "model": model, "calls": []}
+        for user in (
+            "Use tools. Search the 2dph brain for LadybugDB. Call search.",
+            "Use tools. Search the 2dph brain for model2vec. Call search.",
+        ):
+            ms, name = timed(lambda user=user: reasoner_tool_call(base, model, user))
+            report["reasoner"]["calls"].append({"ms": round(ms, 1), "tool": name})
+        tools = [c["tool"] for c in report["reasoner"]["calls"]]
+        report["gates"]["reasoner_tool_call"] = bool(tools) and all(t == "search" for t in tools)
+        if not report["gates"]["reasoner_tool_call"]:
+            report["ok"] = False
+
+    if args.picoclaw:
+        gw = args.picoclaw_url.rstrip("/")
+        ms, raw = timed(lambda: _req(gw + "/health", timeout=5))
+        body = json.loads(raw.decode())
+        report["picoclaw"] = {
+            "url": gw,
+            "health_ms": round(ms, 1),
+            "status": body.get("status"),
+        }
+        report["gates"]["picoclaw_health"] = body.get("status") == "ok" and ms <= GATE_HEALTH_MS
+        if not report["gates"]["picoclaw_health"]:
+            report["ok"] = False
+    return report
+
+
+def main(argv: list[str]) -> int:
+    p = argparse.ArgumentParser(description="2dph system performance (MCP + optional reasoner)")
+    p.add_argument("--brain", default=os.environ.get("BRAIN_URL", DEFAULT_BRAIN))
+    p.add_argument("--n", type=int, default=20)
+    p.add_argument("--json", action="store_true")
+    p.add_argument("--reasoner", action="store_true")
+    p.add_argument("--picoclaw", action="store_true")
+    p.add_argument("--picoclaw-url", default=os.environ.get("PICOCLAW_URL", DEFAULT_PICOCLAW))
+    p.add_argument("--reasoner-url", default=os.environ.get("REASONER_BASE_URL", DEFAULT_REASONER))
+    p.add_argument("--model", default=os.environ.get("REASONER_MODEL", DEFAULT_MODEL))
+    args = p.parse_args(argv)
+    try:
+        report = run(args)
+    except (urllib.error.URLError, TimeoutError, OSError) as e:
+        print(f"system_perf: {e}", file=sys.stderr)
+        return 1
+    if args.json:
+        print(json.dumps(report, indent=2))
+    else:
+        print(f"ok={report['ok']} brain={report['brain']}")
+        for name, block in report.get("mcp", {}).items():
+            if isinstance(block, dict) and "p50_ms" in block:
+                print(f"  {name}: p50={block['p50_ms']} p95={block['p95_ms']} n={block['n']}")
+            elif name == "health":
+                print(f"  health: {block.get('avg_ms')} ms")
+        for k, v in report.get("gates", {}).items():
+            print(f"  gate {k}: {v}")
+        for c in (report.get("reasoner") or {}).get("calls") or []:
+            print(f"  reasoner {c['tool']}: {c['ms']} ms")
+    return 0 if report["ok"] else 1
+
+
+if __name__ == "__main__":
+    raise SystemExit(main(sys.argv[1:]))
@@ -0,0 +1,38 @@
+---
+name: duckdb
+description: >-
+  Use https://github.com/duckdb/duckdb-go in-process for columnar analytics
+  (quantiles, GROUP BY, JSON/CSV/Parquet/JSONL scans) when that is faster than
+  nested Go loops. Not Ladybug. Not the web-search sqlite cache. Use when
+  aggregating samples, counting JSONL, or SQL over tabular files.
+---
+
+# duckdb-go
+
+Use https://github.com/duckdb/duckdb-go where it makes sense to get better performance in code.
+
+In-process DuckDB (`internal/duckstats`, `database/sql` driver `duckdb`).
+Vectorized SQL over tables, JSONL, CSV, Parquet. CGO with bundled libs
+(linux/darwin amd64/arm64). Links with **gcc/g++** (libstdc++), not Zig.
+D21 Zig (`bin/cgo/zcc`) is Ladybug/tokenizers only. After
+`eval "$(bin/cgo/zig env)"`:
+
+```bash
+CC=gcc CXX=g++ CGO_CFLAGS= CGO_LDFLAGS= ./bin/qa/stats.go <<< '[1,2,3,4,5]'
+CC=gcc CXX=g++ CGO_CFLAGS= CGO_LDFLAGS= go test ./internal/duckstats
+```
+
+| Store | Job |
+|-------|-----|
+| Ladybug | graph + FTS + HNSW (facts/info) |
+| modernc sqlite | web-search KV cache + throttle |
+| duckdb-go | OLAP: quantiles, counts, scans of many rows/files |
+| mikefarah/yq | small YAML/JSON/XML/CSV/TOML/HCL slice, not bulk |
+
+```bash
+./bin/qa/stats.go <<< '[1,2,3,4,5]'
+./bin/qa/stats.go --jsonl path/to/rows.jsonl
+```
+
+Do not open Ladybug through DuckDB. Do not put secrets or client PII into
+DuckDB files under the repo.
@@ -1,15 +1,15 @@
 ---
 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.
+  2dph is the memory/fact gate. Compose runs the official PicoClaw gateway.
+  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`
+PicoClaw speaks MCP at `POST /mcp` on `bin/brain/serve.go`. Compose profile
+`picoclaw` runs the official `sipeed/picoclaw` gateway plus `brain-mcp`
 (see [docs/picoclaw.md](../../docs/picoclaw.md)).

 ## Tool order (before a factual reply)
@@ -9,7 +9,7 @@ description: >-
 # postgres

 `bin/postgres/query.go` wraps vendored `bin/db/psql-yq`. Output is YAML
-(cheaper than psql ASCII, easy to slice with `yq`).
+(cheaper than psql ASCII, easy to slice with mikefarah/yq).

 ```bash
 bin/postgres/query.go --profile onlyoffice -s document_asset   # column list
@@ -11,7 +11,7 @@ description: >-
 ```bash
 bin/web/search.go "LadybugDB vector index"
 bin/web/search.go "model2vec multilingual" --category it
-bin/web/search.go "hypervisor" --site example.com --json | jq -r '.results[].url'
+bin/web/search.go "hypervisor" --site example.com --json | yq -r '.results[].url'
 bin/web/search.go "postgres partial index" --lang en --fresh year
 ```

@@ -43,7 +43,7 @@ found" unless the client refuses to call it absence.

 ```bash
 for i in $(seq 10); do
-  bin/web/search.go "test $i" -n 1 --refresh --json | jq -r .status
+  bin/web/search.go "test $i" -n 1 --refresh --json | yq -r '.status'
 done
 ```

@@ -0,0 +1,33 @@
+---
+name: yq
+description: >-
+  Use https://github.com/mikefarah/yq to work with YAML, JSON, XML, CSV,
+  TOML, HCL where it's efficient and less code. Use when slicing compose,
+  config, --json tool output, CSV/TOML/HCL/XML, or converting between those
+  formats. Not kislyuk Python yq. Not jq when yq already does the job.
+---
+
+# yq (mikefarah)
+
+Use https://github.com/mikefarah/yq to work with YAML, JSON, XML, CSV, TOML, HCL where it's efficient and less code.
+
+This is the Go `yq` (`yq --version` contains `mikefarah`). It is not
+kislyuk/yq (Python, jq-syntax, YAML-only wrapper). `bin/db/psql-yq` already
+calls this binary.
+
+Prefer `yq` over `python3 -c`, `jq`, or ad-hoc parsers when one expression
+reads or converts the file. Keep Python/Go for HTTP, binary protocols, and
+in-process tests.
+
+```bash
+yq '.services.picoclaw.image' compose.yaml
+yq -P . deploy/picoclaw/config.json          # JSON → YAML
+yq -o=json '.gates'                           # JSON stdin (qa/system_perf.py --json)
+yq -p=csv -o=json .
+yq -p=xml -o=json .
+yq -p=toml '.package.name' file.toml
+bin/brain/search.go "LadybugDB" --json | yq '.[].ref'
+bin/web/search.go "hypervisor" --json | yq -r '.results[].url'
+```
+
+Do not print secrets, PII, or `$HOME/.config/brain/` through `yq`.
Author	SHA1	Message	Date
eSlider	168a0a3b53	feat: DuckDB quantiles in-process (gcc CGO), not Ladybug. Tests / Test (push) Skipped Tests / OCR (tesseract fixture) (push) Skipped Tests / Release (semver) (push) Skipped Tests / Test (pull_request) Failing after 5s Details Tests / OCR (tesseract fixture) (pull_request) Failing after 4s Details Tests / Release (semver) (pull_request) Skipped Details OQ3: internal/duckstats + bin/qa/stats.go. Zig stays Ladybug-only. mikefarah/yq for small structured slices. Gitea #30.	2026-08-14 11:29:22 +01:00
eSliderandGitHub	f99dfea104	feat: OCR scans with tesseract, drop docling from the default path. (#33 ) Tests / Test (push) Failing after 5s Details Tests / OCR (tesseract fixture) (push) Failing after 4s Details Tests / Release (semver) (push) Skipped Details pdftotext still wins on born-digital PDFs. Empty text layers go through pdftoppm + tesseract eng+deu. Optional OCR_ENGINE=paddle. Gitea #6.	2026-08-14 11:26:04 +01:00
eSliderandGitHub	bae1494258	ci: recall@5 SoT is Zig bin/brain/eval.go (#19 ) (#32 ) Tests / Test (push) Failing after 6s Details Tests / Release (semver) (push) Skipped Details * ci: recall@5 SoT is Zig bin/brain/eval.go. GitHub Actions rebuilds the repo corpus and runs the Go eval gate instead of Python bin/kb/eval. Audit self is a hard fail (Gitea #19). * ci: point eval at the checkout and keep DevOps in the corpus. /tmp/brain-eval cannot walk up to var/kb.lbug; KB_ROOT is required. Recall fragments must exist in README/docs so a repo-only rebuild gates.	2026-08-14 11:03:39 +01:00
eSliderandGitHub	c5be3f19be	feat: index facts extract and chats markdown on rebuild. (#31 ) Tests / Test (push) Failing after 5s Details Tests / Release (semver) (push) Skipped Details --with-facts / --facts-json write root=facts leafs; --with-chats picks up var/chats/md as info. WhatsApp sync stays out of v1 (Gitea #18).	2026-08-14 10:57:40 +01:00
eSliderandGitHub	a88dbb490c	feat: search --hop walks FROM_FILE to Person. (#30 ) Tests / Test (push) Failing after 4s Details Tests / Release (semver) (push) Skipped Details Parser no longer errors; hop 1 returns File, hop 3 reaches Person. Rebuild writes Leaf-[:FROM_FILE]->File so the walk is not empty on a fresh index (Gitea #17).	2026-08-14 10:51:35 +01:00