11 changed files with 125 additions and 1137 deletions
--- a/CLAUDE.md
+++ b/CLAUDE.md
@ -1,297 +1,122 @@
 # second-brain

-Personal knowledge pipeline. Two machines coordinate purely through the
-shared Postgres queue: the **dev side** (herbys-dev) does article pull
-+ all extraction + embedding, and the **tower** (EndeavourOS, RTX 3080)
-does video pull + transcription on the GPU. A FastAPI/Jinja/HTMX web
-app fronts the queue, settings, and add-to-queue form; the file-based
-Obsidian wiki is downstream of `accepted` extractions.
+A personal knowledge system: video/article extraction pipeline + LLM-maintained wiki.

-## Two-machine pipeline
+## Architecture
+
+Three-layer system:
+
+1. **Wiki** — long-term structured memory. Markdown files in an Obsidian vault, maintained by the wiki compiler.
+2. **Context assembler** — builds the prompt bundle for each extraction. Phase 1 is thin: domain template + focus field + transcript.
+3. **Extractor** — stateless single-shot LLM call per source. Pure function: bundle in, structured extraction out.
+
+## Pipeline stages

 ```
-                add (CLI or POST /sources/add — single video, article,
-                     or YouTube playlist URL → fan-out into N rows)
-                                  │
-                                  ▼
-   ┌────────── second_brain.sources (the queue) ──────────┐
-   │                                                       │
-   │   PENDING  ── (article)──►  dev `second-brain process`
-   │                              trafilatura pull         │
-   │   PENDING  ── (video)  ──►  tower transcribe-worker   │
-   │                              yt-dlp pull → PULLED     │
-   │   PULLED   ── (video)  ──►  tower transcribe-worker   │
-   │                              faster-whisper           │
-   │              both paths land at  ──►  TRANSCRIBED     │
-   │                                                       │
-   │   TRANSCRIBED ─────────►  dev `second-brain schedule` │
-   │                              Claude CLI extract +     │
-   │                              Ollama embed             │
-   │              ──►  ANALYZED  ──►  accept in web UI     │
-   │              ──►  ACCEPTED  ──►  `compile` writes vault│
-   │              ──►  PUBLISHED                           │
-   │                                                       │
-   │   FAILED is the terminal error state                  │
-   └───────────────────────────────────────────────────────┘
+pending → pulled → transcribed → analyzed → accepted → published
 ```

-**Source claiming** — `sources.claimed_by` / `claimed_at` columns +
-`SELECT ... FOR UPDATE SKIP LOCKED` keep the dev and tower workers from
-grabbing the same row. Both stages filter by `source_type`:
+## Project layout

- tower transcribe-worker: `status IN (PENDING, PULLED) AND source_type = 'video'`
- dev `process` (article pull): `status = PENDING AND source_type = 'article'`
- dev `schedule` (extract+embed): `status = TRANSCRIBED` (any type)
+```
+src/second_brain/
+├── adapters/      # Pull adapters (youtube.py, article.py)
+├── context/       # Context assembler
+├── embeddings/    # Chunk + nomic-embed-text + upsert into public.embeddings
+├── extractor/     # LLM extraction engine + Pydantic schema
+├── llm/           # claude_cli.py — thin subprocess wrapper around `claude -p`
+├── compiler/      # Wiki compiler (reads vault, writes pages, git commits)
+├── scheduler/     # Rate-limit smoother (call-count for CLI, token-budget for API)
+└── web/           # FastAPI + Jinja2 + HTMX UI

-A 30-minute stale-claim reaper runs at worker startup so a crashed
-machine doesn't dam the queue.
+alembic/           # Schema migrations for the second_brain Postgres schema
+```
+
+## Setup
+
+```bash
+cp config/settings.toml.example config/settings.toml
+# Edit settings.toml — at minimum, point the paths and set SECOND_BRAIN_DATABASE_URL.
+
+# Connection string format. Containerised (default):
+#   postgresql+psycopg://lovebug:<password>@homelab-postgres:5432/petalbrain
+# Host-side dev runs:
+#   postgresql+psycopg://lovebug:<password>@127.0.0.1:5433/petalbrain
+export SECOND_BRAIN_DATABASE_URL=...
+
+# One-time, before first migration: bootstrap the schema as postgres
+# superuser. lovebug doesn't have CREATE on petalbrain, so this can't be
+# done from the app side — same pattern as wiki/trellis/ob1.
+#   CREATE SCHEMA second_brain AUTHORIZATION lovebug;
+#   GRANT ALL PRIVILEGES ON SCHEMA second_brain TO lovebug;
+
+uv sync
+uv run alembic upgrade head
+uv run second-brain serve
+```

 ## CLI commands

- `second-brain add <url>` — queue a source. YouTube `playlist?list=…`
-  URLs auto-expand into one row per video (cap 50, the
-  `watch?v=…&list=…` form stays a single video — see
-  `is_youtube_playlist_url` for the rule).
- `second-brain process` — dev-side: pull pending **articles**, extract
-  + embed any TRANSCRIBED source. Skips videos (tower's job).
- `second-brain schedule` — dev-side overnight extract+embed loop,
-  honouring `pipeline_settings.extraction_enabled` / active-hours /
-  max-items.
- `second-brain transcribe-worker` — tower-side daemon. `--once` for
-  ad-hoc runs. Honours `pipeline_settings.transcription_*` knobs.
- `second-brain serve` — local dev launch of the web UI (port 8000).
-  Production runs as a container (see `deploy/web/`).
- `second-brain compile` — wiki compiler: ACCEPTED → markdown in
-  `vault_path`, git-committed. File-based Obsidian vault only; we do
-  **not** dual-write into petalbrain.wiki.
-
-## Storage
-
-**Postgres**: `petalbrain` DB on `homelab-postgres` (pgvector/pg17, docker
-container). Connection role: `lovebug`. Dev reaches it at
-`127.0.0.1:5433`; tower reaches it over WireGuard at
-`db.wg.herbylab.dev:5432`; the web container reaches it at
-`homelab-postgres:5432` on the `homelab` docker bridge. **Keep both port
-binds** — the host-side 5433 publish and the docker-net 5432 service.
-
-**Schema `second_brain`** (Alembic-managed in `alembic/`):
-
- `sources` — the queue. Int SERIAL PK. Naive UTC `timestamp`
-  (`without time zone`) — never `timestamptz`. `transcript_text` is the
-  joined full text the extractor reads; `transcript_segments` is JSONB
-  with the faster-whisper per-segment / per-word breakdown.
-  `claimed_by` / `claimed_at` for the queue dance.
- `extractions` — one row per source. JSON columns for the structured
-  output (summary, key_points, entities, claims, …).
- `pipeline_settings` — **single-row** config (PK pinned to id=1 with a
-  CHECK constraint). What the dashboard edits and workers read each
-  iteration: transcription_enabled / active_hours / max_gpu_jobs /
-  max_video_length / max_items_per_run plus the equivalent extraction_*
-  fields. DB values override settings.toml.
- `wiki_pages` — bookkeeping for the file-based wiki compiler.
-
-**Embeddings** live in `public.embeddings` — the **shared** pgvector
-table owned by `postgres` and consumed by vault-mcp / ob1 / second-brain.
-We do not manage this in Alembic. Rows are keyed by
-`(source_schema='second_brain', source_table='extractions', source_id, embedding_model='nomic-embed-text')`,
-with `delete-before-insert` semantics on the 4-tuple so a re-embed
-never leaves orphans. We embed extraction **summaries only** for now —
-not key_points, not transcript chunks.
-
-## Web app
-
-FastAPI + Jinja2 + HTMX, in `src/second_brain/web/`. Production lives in
-a slim Python 3.12 container (see `deploy/web/`) on the `homelab` docker
-network, reaching Postgres + Ollama by service DNS. Published at
-`10.0.21.207:8080` and fronted by the Traefik VM (10.0.11.20,
-file-provider — **no docker-label routing**) as
-`https://brain.herbylab.dev`.
-
-Routes:
-
- `/` — sources list (filterable by domain + status)
- `/queue` — in-flight sources (PENDING / PULLED / TRANSCRIBED) plus the
-  add-to-queue form
- `/dashboard` — status counts, recent activity, settings snapshot, plus
-  the add-to-queue form
- `/sources/{id}` — extraction detail + accept/reject HTMX actions
- `/settings` — full HTMX editor for `pipeline_settings`
- `POST /sources/add` — shared endpoint. Detects YouTube playlist URLs
-  and fans out via yt-dlp `extract_flat`. Reads `HX-Current-URL` to
-  dispatch the response partial (`_queue_body.html` vs
-  `_dashboard_body.html`) so each page refreshes its own region inline.
-
-**No auth.** UI mutates state (accept/reject sources, edit settings) but
-has no SSO / CSRF / rate-limiting. The Traefik route is currently
-LAN-only by Travis's call — see `deploy/web/README.md` §SECURITY.
-
-## Transcription
-
-`src/second_brain/transcribe.py` wraps **faster-whisper** (CTranslate2
-backend). Defaults: model `large-v3`, device `cuda`, compute_type
-`float16`. CPU fallback drops to `int8`. `word_timestamps=True` is
-always passed; the segment dicts persisted to
-`sources.transcript_segments` carry per-word `{start, end, word,
-probability}`.
-
-The tower worker (`src/second_brain/scheduler/transcribe_worker.py`)
-runs as a systemd service — see `deploy/tower/README.md` for the full
-install runbook. ExecStart goes through
-`deploy/tower/run-transcribe-worker.sh` which prepends the venv's
-CUDA-12 lib dirs to `LD_LIBRARY_PATH` (see gotcha (b) below).
+- `second-brain add <url>` — queue a source URL (video or article)
+- `second-brain process` — run the pipeline on pending items
+- `second-brain serve` — launch the web UI (default port 8000)
+- `second-brain compile` — run the wiki compiler
+- `second-brain schedule` — start the overnight scheduler

 ## Extractor backends

-`[extractor].backend` in `config/settings.toml`:
+`[extractor].backend` in `config/settings.toml` picks the LLM transport.

- `cli` (default) — `src/second_brain/llm/claude_cli.py` shells out to
-  `claude -p --output-format json --model … --json-schema …`. Runs
-  under the Max OAuth subscription. Hermetic spawn: throwaway
-  `tempfile.TemporaryDirectory`, `ANTHROPIC_API_KEY` / `_AUTH_TOKEN`
-  stripped from env so it uses OAuth, and `--no-session-persistence
-  --disable-slash-commands --tools "" --setting-sources ""` so host
-  CLAUDE.md / hooks / agents don't leak in.
- `api` — `anthropic` SDK, lazy-imported. Requires `ANTHROPIC_API_KEY`.
-  Install with `uv sync --extra api`.
+- **`cli`** (default) — `src/second_brain/llm/claude_cli.py` shells out to `claude -p --output-format json --model <model> --json-schema <schema>`. Runs under the Max OAuth subscription (no per-token billing). Hermetic isolation: spawned inside `tempfile.TemporaryDirectory(prefix="sb-claude-cli-")`, with `ANTHROPIC_API_KEY` / `ANTHROPIC_AUTH_TOKEN` stripped from env so the CLI uses OAuth instead of an API key, and with `--no-session-persistence --disable-slash-commands --tools "" --setting-sources ""` so host CLAUDE.md / hooks / agents don't leak in.
+- **`api`** — uses the `anthropic` SDK. Lazy-imported only when this backend is selected. The SDK is an optional dependency (`uv sync --extra api`).

-Schema-validated CLI output lands at `payload["structured_output"]`, not
-`payload["result"]`. The wrapper checks both.
+**Schema-validated output gotcha:** when `--json-schema` is used, the parsed object lands at `payload["structured_output"]`, not `payload["result"]`. The wrapper checks both. The engine prefers the structured payload via `LLMExtraction.model_validate(structured)` and falls back to parsing the text field.

 ## Environment variables

- `SECOND_BRAIN_DATABASE_URL` — Postgres DSN. Falls back to
-  `HERBYLAB_DATABASE_URL`. Use `postgresql+psycopg://` for SQLAlchemy;
-  the embeddings + claim modules strip the `+psycopg` prefix for raw
-  psycopg.
- `OLLAMA_URL` — overrides `[embeddings].ollama_url`. Default
-  `http://ollama:11434` (homelab docker net); host-side runs use
-  `http://127.0.0.1:11434`.
- `EMBEDDING_MODEL` — default `nomic-embed-text` (768-d).
- `WHISPER_MODEL` / `WHISPER_DEVICE` / `WHISPER_COMPUTE_TYPE` — override
-  the `[whisper]` block on the tower.
- `ANTHROPIC_API_KEY` — only for `extractor.backend = "api"`.
- `SECOND_BRAIN_CONFIG` — settings file path override.
- `SECOND_BRAIN_WORKER_ID` — stamped into `sources.claimed_by`.
- `SECOND_BRAIN_DB_POOL_MIN` / `_MAX` — SQLAlchemy pool sizing.
- `SECOND_BRAIN_LIVE_NETWORK_TESTS=1` — opt-in for the yt-dlp playlist
-  test that hits real YouTube.
+- `ANTHROPIC_API_KEY` — only required when `extractor.backend = "api"`. The default `cli` backend uses your Max OAuth and does not need it.
+- `SECOND_BRAIN_CONFIG` — optional override for the settings file path.
+- `SECOND_BRAIN_DATABASE_URL` — Postgres connection string. Falls back to `HERBYLAB_DATABASE_URL` if unset. Required at runtime; both `second_brain.database` and `second_brain.embeddings` read it. Use the SQLAlchemy `postgresql+psycopg://` form; the embeddings module strips the `+psycopg` prefix internally for raw psycopg.
+- `OLLAMA_URL` — overrides `[embeddings].ollama_url`. Default `http://ollama:11434` (homelab docker network); for host-side runs use `http://127.0.0.1:11434`.
+- `EMBEDDING_MODEL` — overrides `[embeddings].model`. Default `nomic-embed-text` (768-d).
+- `SECOND_BRAIN_DB_POOL_MIN` / `SECOND_BRAIN_DB_POOL_MAX` — SQLAlchemy pool sizing knobs (defaults 1 / 9 overflow).

 ## Domains

-`development` / `content` / `business` / `homelab` — drives the
-per-domain prompt template loaded by the context assembler.
+- `development` — software, programming, systems
+- `content` — content creation strategy, video production
+- `business` — entrepreneurship, marketing, operations
+- `homelab` — self-hosted infrastructure, networking, DevOps

 ## Key files

- `src/second_brain/models.py` — all ORM in the `second_brain` schema.
- `src/second_brain/sources_service.py` — `add_source`, `add_playlist`,
-  `is_youtube_playlist_url`, `expand_youtube_playlist`. CLI + web both
-  call into here so there's no behavior drift between entry points.
- `src/second_brain/claim.py` — the SKIP LOCKED dance + stale reaper.
- `src/second_brain/settings_store.py` — single-row `pipeline_settings`
-  get/update + the active-window math the workers share.
- `src/second_brain/transcribe.py` — faster-whisper wrapper + segment
-  serialiser.
- `src/second_brain/scheduler/runner.py` — dev extract scheduler.
- `src/second_brain/scheduler/transcribe_worker.py` — tower poll loop.
- `src/second_brain/embeddings/__init__.py` — chunk → Ollama → upsert
-  into `public.embeddings`; best-effort with graceful degradation.
- `alembic/` + `alembic.ini` — schema migrations for the `second_brain`
-  schema only. `public.embeddings` is **not** ours to migrate.
- `deploy/web/` — Dockerfile + compose.yml + README for the web
-  container. Travis applies the Knot DNS A-record + Traefik
-  file-provider router from the README §2.
- `deploy/tower/` — systemd unit + `run-transcribe-worker.sh` wrapper +
-  env template + EndeavourOS install runbook.
- `prompts/<domain>.md` — extraction prompt templates.
- `config/settings.toml` — `[database]`, `[extractor]`, `[embeddings]`,
-  `[scheduler]`, `[whisper]`.
+- `config/settings.toml` — vault path, `[database]`, `[extractor]`, `[embeddings]`, `[scheduler]`
+- `prompts/<domain>.md` — per-domain extraction prompt templates
+- `src/second_brain/extractor/schema.py` — `LLMExtraction` (LLM-only fields, drives the `--json-schema` contract) + `ExtractionResult` (LLMExtraction + SourceMeta filled from the DB)
+- `src/second_brain/llm/claude_cli.py` — CLI subprocess wrapper
+- `src/second_brain/models.py` — SQLAlchemy models (all tables live in the `second_brain` schema) + `utcnow()` helper
+- `src/second_brain/database.py` — Postgres engine + URL resolution
+- `src/second_brain/embeddings/__init__.py` — chunk → Ollama → upsert into `public.embeddings`; best-effort with graceful degradation
+- `alembic/` + `alembic.ini` — schema migrations for the `second_brain` schema (Alembic does **not** manage `public.embeddings`)
+- `postgres-migration-planning.md` — original planning doc; superseded by this file once the migration shipped

-## Load-bearing gotchas
+## Conventions / gotchas

-These have all bitten us; don't relearn them.
+- **Naive UTC timestamps everywhere.** Use `from second_brain.models import utcnow` rather than `datetime.utcnow()` (deprecated in 3.12) or `datetime.now()` (local time). Postgres columns are `timestamp without time zone` — do not switch to `timestamptz`.
+- **PK style.** Int autoincrement (`SERIAL`) — locked for compatibility with the existing schema shape. Do not switch to `BIGINT GENERATED ALWAYS AS IDENTITY` or UUID.
+- **Re-query inside the session.** SQLAlchemy ORM attribute access after the session closes raises `DetachedInstanceError`. The `process` command and the web accept/reject endpoints both collect IDs first, then re-query inside the per-iteration / per-request session block, and pre-compute any dict views needed for templates *inside* the `with db.session()` block.
+- **Starlette `TemplateResponse` signature.** Starlette ≥ 0.36 requires `templates.TemplateResponse(request, name, context)`. The old `(name, context)` form raises `TypeError: unhashable type: 'dict'`. Do not put `"request": request` into the context dict — it's auto-injected now.
+- **Vault git repo auto-init.** The wiki compiler calls `_ensure_git_repo()` on every run, which `git init`s the vault and creates an empty seed commit if no `.git` is present. `_git_commit()` also checks `git diff --cached --quiet` before committing to avoid empty commits.
+- **Scheduler throttling shape.** The CLI backend is throttled by `max_calls_per_hour` (Max subscription limits are call-shaped, not token-shaped). The API backend uses `max_tokens_per_hour`. Both share `min_gap_seconds`. The scheduler targets sources in `TRANSCRIBED` state — earlier code mistakenly filtered on `ANALYZED`, making it a no-op.
+- **Alembic + lovebug privilege gap.** `lovebug` owns the `second_brain` schema but does **not** have `CREATE` on the `petalbrain` database. `alembic/env.py` therefore guards `CREATE SCHEMA` behind a `pg_namespace` lookup, and the v1 migration intentionally drops the `IF NOT EXISTS` form (a bare `CREATE SCHEMA IF NOT EXISTS` raises permission denied even when the schema is already present). The schema is bootstrapped once, out-of-band, by `postgres` superuser.
+- **public.embeddings is shared, not ours.** Alembic only manages the `second_brain` schema. The vector table itself is owned by `postgres` and shared across vault-mcp / ob1 / second-brain — we just INSERT into it. Rows are keyed by `(source_schema='second_brain', source_table='extractions', source_id, embedding_model='nomic-embed-text')`.
+- **Embedding is best-effort.** The pipeline writes the relational `extractions` row first and `sess.flush()`es; only then calls `embed_extraction()`. Any failure (no DB URL, no Ollama, embedding 500) logs at WARNING and returns None — the relational row and file vault stay canonical.
+- **File-based wiki, not petalbrain.wiki.** The wiki compiler still writes markdown into the configured `vault_path` and commits via git. We do **not** dual-write into `petalbrain.wiki` — that's vault-mcp's territory and intentionally out of scope.

- **(a) Schema bootstrap requires postgres superuser.** `lovebug` does
-  NOT have `CREATE` on the `petalbrain` database, even though it owns
-  the `second_brain` schema. The first migration needs a one-time
-  `CREATE SCHEMA second_brain AUTHORIZATION lovebug; GRANT ALL PRIVILEGES
-  ON SCHEMA second_brain TO lovebug;` as `postgres`. `alembic/env.py`
-  guards `CREATE SCHEMA` behind a `pg_namespace` lookup so subsequent
-  runs as `lovebug` don't trip the permission check.
+## In-flight work

- **(b) Arch ships CUDA 13, ctranslate2 needs CUDA 12.** EndeavourOS's
-  `pacman -S cuda cudnn` installs `libcublas.so.13` / `libcudnn.so.10`,
-  which `ctranslate2` 4.7.x cannot load — it wants the .12 / .9 ABI.
-  We pin `nvidia-cublas-cu12` and `nvidia-cudnn-cu12>=9,<10` in the
-  `tower` optional-dependency group, and the systemd ExecStart calls
-  `deploy/tower/run-transcribe-worker.sh` which resolves the venv's
-  wheel-bundled `.so` directories at runtime and prepends them to
-  `LD_LIBRARY_PATH`. The wrapper uses the namespace packages'
-  `__path__` (not `__file__`, which is `None` for PEP 420 packages).
-  systemd does **not** inherit shell exports.
-
- **(c) pg_hba needs BOTH subnets.** Two host rules are load-bearing:
-  ```
-  host all all 172.19.0.0/16  scram-sha-256   # homelab docker bridge
-  host all all 10.99.0.0/24   scram-sha-256   # WireGuard (tower)
-  ```
-  Dropping the 172.19/16 rule locks out the web container + vault-mcp +
-  ob1-* + trellis-mcp (all on the homelab net). Dropping the 10.99/24
-  rule locks out the tower. Both have been removed once each during
-  unrelated postgres edits — the symptom is `FATAL: no pg_hba.conf
-  entry for host "…"` in service logs.
-
- **(d) Two postgres port binds, both load-bearing.** `homelab-postgres`
-  publishes `127.0.0.1:5433 → 5432` for dev-host access AND is reachable
-  on `homelab-postgres:5432` over the docker net. Tower hits it over WG
-  at `db.wg.herbylab.dev:5432`. Don't consolidate; each path serves a
-  different consumer set.
-
- **(e) Traefik is file-provider on a separate VM (10.0.11.20).** It
-  cannot read this host's docker labels. The web `compose.yml` therefore
-  publishes back to a fixed host port (`10.0.21.207:8080`) that Traefik
-  reaches over LAN, and intentionally carries NO `traefik.*` labels.
-  Router config lives in Traefik's dynamic-config YAML — see
-  `deploy/web/README.md` §2b for the snippet.
-
- **Code-level conventions, since they bit us early:**
-  - **Naive UTC timestamps.** Use `from second_brain.models import utcnow`
-    — never `datetime.utcnow()` (deprecated) or `datetime.now()` (local).
-    Postgres columns are `timestamp without time zone`.
-  - **Int autoincrement PKs.** Don't migrate to bigint-identity or UUID;
-    the existing schema shape is locked.
-  - **Re-query inside the session.** SQLAlchemy ORM attribute access
-    after the session closes raises `DetachedInstanceError`. Web routes
-    and CLI loops collect IDs first, then re-fetch inside the session
-    block; dict snapshots for templates are built inside `with
-    db.session()`.
-  - **Starlette TemplateResponse signature.** `templates.TemplateResponse(request, name, context)` — the old `(name, context)` form raises
-    `TypeError: unhashable type: 'dict'`. Do not put `"request":
-    request` into the context dict.
-  - **Embedding is best-effort.** The extract path flushes the
-    relational `extractions` row first, then calls `embed_extraction()`.
-    DB / Ollama failures log and return None; pipeline state stays
-    canonical in the relational + file vault.
-  - **Worker LD_LIBRARY_PATH.** See (b). When pre-warming or debugging
-    faster-whisper manually, run via `deploy/tower/run-transcribe-worker.sh`
-    or replicate the path probe in your shell.
-
-## In-flight / parked
-
- **Vault location** — recommended `/opt/projects/second-brain-vault/`
-  as a peer dir to the project. Not yet confirmed.
- **Media + subtitles on NAS** — currently land on the tower's local
-  SSD. Move to NAS once we have a stable mount path.
- **SSO** — Authentik forward-auth middleware in Traefik is the
-  obvious next step before opening the UI past LAN.
- **Dev-side extraction systemd timer** — currently invoked manually;
-  no systemd unit yet by Travis's call.
+- **Vault location.** Recommended `/opt/projects/second-brain-vault/` as the production location (peer to the project dir). Not yet confirmed.

 ## Validation

-Per the user's global CLAUDE.md, run `zero-check` after any code
-change. The pipeline's at `~/zero-check-pipeline/validate/run-all.sh
-"$(pwd)"`. Run `a-review` after non-trivial diffs (>50 lines):
-`python3 ~/.claude/skills/a-review/run_review.py`. Test suite is `uv
-run pytest`; live DB + Ollama paths auto-skip when their endpoints
-aren't configured.
+Per Travis's global CLAUDE.md, run `zero-check` after any code change, and `a-review` after non-trivial diffs (> 50 lines). `a-review` script: `python3 ~/.claude/skills/a-review/run_review.py`, output goes to `reviews/a-review-<date>.md`.
--- a/deploy/tower/README.md
+++ b/deploy/tower/README.md
@ -50,9 +50,9 @@ your CLI PATH.

 ```bash
 sudo mkdir -p /opt/projects/homelab
-sudo chown trucktrav:trucktrav /opt/projects/homelab
+sudo chown herbyadmin:herbyadmin /opt/projects/homelab
 cd /opt/projects/homelab
-git clone gitea-pbs-trucktrav:petal-power/second-brain.git
+git clone gitea-lovebug:petal-power/second-brain.git
 cd second-brain

 # Install the runtime deps + the `tower` extra (faster-whisper / CTranslate2).
@ -90,7 +90,7 @@ The worker assumes the tunnel is already configured and that
 | Peer | WG IP |
 |------|-------|
 | hub (herbydev) | `10.99.0.1` |
-| tower | `10.99.0.3` |
+| tower | `10.99.0.2` |

 `db.wg.herbylab.dev` resolves to `10.99.0.1`. Confirm connectivity from
 the tower:
--- a/deploy/tower/second-brain-transcribe.service
+++ b/deploy/tower/second-brain-transcribe.service
@ -16,8 +16,8 @@ StartLimitBurst=5

 [Service]
 Type=simple
-User=trucktrav
-Group=trucktrav
+User=herbyadmin
+Group=herbyadmin

 # Adjust to wherever you `git clone`d the repo on the tower.
 WorkingDirectory=/opt/projects/homelab/second-brain
--- a/reviews/a-review-2026-05-25.md
+++ b/reviews/a-review-2026-05-25.md
@ -1,37 +0,0 @@
-# a-review — 2026-05-25
-
-**Model:** gemini
-**Diff base:** `HEAD~1` (5 files changed, 615 insertions(+), 30 deletions(-))
-**Session reference:** none
-**Context:** CLAUDE.md + session notes + expanded diff
-
---
-
-Excellent work. The changes are well-structured, follow existing project conventions, and include a comprehensive test suite. This is a high-quality contribution. I have only one minor finding and one informational note.
-
-### [WARNING] Logic Error
-
-*   **File:** `src/second_brain/web/templates/source_detail.html`
-*   **Location:** Line 151
-*   **Issue:** The "Re-extract only" button is conditionally rendered with `{% if source.has_transcript and source.source_type == 'video' %}`. This is too restrictive. The backend service function `reset_source_for_retry` correctly allows an "extract" retry for any source that has content in `transcript_text`, which includes processed articles. Limiting this feature to only videos in the UI prevents a valid and useful action for non-video sources.
-*   **Suggestion:** Change the conditional to `{% if source.has_transcript %}`. This will correctly show the button for any source (article or video) that has been pulled and has content ready for extraction, which matches the backend's capability.
-
-```html
-    <!-- Change this -->
-    {% if source.has_transcript and source.source_type == 'video' %}
-    
-    <!-- To this -->
-    {% if source.has_transcript %}
-```
-
-### [INFO] Security
-
-*   **Files:** `src/second_brain/web/app.py`, `src/second_brain/web/templates/source_detail.html`
-*   **Observation:** The changes introduce new endpoints (`/edit`, `/retry`) that modify database state. According to `CLAUDE.md`, the application has "No auth" and relies on being deployed in a trusted, LAN-only environment. These new endpoints increase the number of ways an unauthorized user on the network could alter data.
-*   **Context:** This is not a new vulnerability, but an expansion of the existing trust model. As long as the application remains on a trusted network, this is acceptable per the project's documented security posture. It's worth keeping in mind that no CSRF protection is being used for these HTMX `POST` actions.
-
-### Summary
-
-The overall code quality is very high. The changes are thoughtfully implemented, with a clean separation between the CLI, the web layer, and a shared service layer. The refactoring in `web/app.py` to create `_render_source_detail` is a good move for maintainability. The addition of a thorough, live-DB test suite for the new service functions is commendable and demonstrates a commitment to quality.
-
-Once the minor template logic issue is resolved, this change will be in perfect shape.
--- a/src/second_brain/config.py
+++ b/src/second_brain/config.py
@ -85,28 +85,6 @@ def _merge(default: dict, override: dict) -> dict:
    return out


-def _overlay_sections(base: dict, overlay: dict) -> dict:
-    """Section-level overlay: one level deep.
-
-    For each key in `overlay`, if both `base[key]` and `overlay[key]` are
-    dicts, merge their keys (overlay wins). Otherwise the overlay value
-    replaces the base value outright. Sections present in only one of
-    the two pass through untouched.
-
-    Use this — not `_merge` — when combining `settings.toml` with
-    `settings.local.toml`, so a local `[database] url = "..."` doesn't
-    silently wipe other `[database]` keys defined in the base file.
-    """
-    out = dict(base)
-    for key, ov_val in (overlay or {}).items():
-        base_val = out.get(key)
-        if isinstance(base_val, dict) and isinstance(ov_val, dict):
-            out[key] = {**base_val, **ov_val}
-        else:
-            out[key] = ov_val
-    return out
-
-
 class Config:
    """Holds resolved configuration values."""

@ -197,24 +175,7 @@ _config_instance: Config | None = None


 def load_config(path: Path | None = None) -> Config:
-    """Load (and cache) configuration from a TOML file.
-
-    After loading the base file, look for a sibling `settings.local.toml`
-    and overlay it (section-level merge — see `_overlay_sections`). The
-    local file is gitignored and intended for host-specific secrets
-    (e.g. `[database].url` for a dev shell) without forking the tracked
-    base.
-
-    Path resolution:
-      - explicit `path=` arg wins
-      - else `SECOND_BRAIN_CONFIG` env var
-      - else walks up from CWD for `config/settings.toml` or `settings.toml`
-
-    The local-overlay sibling is looked up next to whichever base file
-    was chosen, so `SECOND_BRAIN_CONFIG=/etc/sb/settings.toml` will also
-    pick up `/etc/sb/settings.local.toml` if present. If no base file is
-    found at all, no overlay is attempted.
-    """
+    """Load (and cache) configuration from a TOML file."""
    global _config_instance
    if _config_instance is not None:
        return _config_instance
@ -236,15 +197,6 @@ def load_config(path: Path | None = None) -> Config:
        with open(path, "rb") as fh:
            raw = tomllib.load(fh)

-        # Section-level overlay from a sibling settings.local.toml.
-        # Filename is the one already covered in .gitignore — host
-        # secrets live here and never get committed.
-        local_path = path.parent / "settings.local.toml"
-        if local_path.exists():
-            with open(local_path, "rb") as fh:
-                local_raw = tomllib.load(fh)
-            raw = _overlay_sections(raw, local_raw)
-
    _config_instance = Config(raw)
    return _config_instance

--- a/src/second_brain/main.py
+++ b/src/second_brain/main.py
@ -4,7 +4,6 @@ second-brain CLI entry point.
 Commands:
  add       — queue a source URL (video or article)
  process   — run the full pipeline on pending/pulled/transcribed sources
-  retry     — reset a source so the pipeline re-processes it
  serve     — launch the FastAPI web UI
  compile   — run the wiki compiler on accepted sources
  schedule  — start the overnight scheduler
@ -308,47 +307,6 @@ def serve(host: str, port: int, reload: bool) -> None:
    )


-# ---------------------------------------------------------------------------
-# retry
-# ---------------------------------------------------------------------------
-
-
-@cli.command()
-@click.argument("source_id", type=int)
-@click.option(
-    "--mode",
-    type=click.Choice(["full", "extract"]),
-    default="full",
-    show_default=True,
-    help="`full` resets to PENDING (re-pull + re-process); "
-    "`extract` resets to TRANSCRIBED (requires existing transcript).",
-)
-def retry(source_id: int, mode: str) -> None:
-    """Reset a source so the pipeline re-processes it.
-
-    Clears the claim pair and error_message and rewinds status to the
-    appropriate stage. Parity with the dashboard re-run buttons.
-    """
-    from second_brain.sources_service import reset_source_for_retry
-
-    db = get_database()
-    try:
-        with db.session() as sess:
-            source = reset_source_for_retry(
-                sess, source_id=source_id, mode=mode
-            )
-            new_status = source.status.value
-            url = source.url
-    except LookupError as exc:
-        click.echo(f"[retry] {exc}", err=True)
-        sys.exit(2)
-    except ValueError as exc:
-        click.echo(f"[retry] {exc}", err=True)
-        sys.exit(2)
-
-    click.echo(f"[retry] id={source_id} status → {new_status}  ({url})")
-
-
 # ---------------------------------------------------------------------------
 # compile
 # ---------------------------------------------------------------------------
--- a/src/second_brain/sources_service.py
+++ b/src/second_brain/sources_service.py
@ -275,126 +275,14 @@ def add_playlist(
    return result


-# ---------------------------------------------------------------------------
-# Edit + retry helpers (shared by web routes and CLI)
-# ---------------------------------------------------------------------------
-
-
-# Fields the edit-mode form is allowed to mutate. URL is intentionally
-# excluded — it's the UNIQUE dedupe key and editing it risks collisions
-# plus breaks the embeddings 4-tuple identity (source_id is stable but
-# semantic identity drifts).
-EDITABLE_METADATA_FIELDS = ("domain", "title", "focus")
-
-
-def update_source_metadata(
-    sess: Session,
-    *,
-    source_id: int,
-    domain: Optional[str] = None,
-    title: Optional[str] = None,
-    focus: Optional[str] = None,
-) -> Source:
-    """Update editable metadata on a source row.
-
-    Empty strings collapse to NULL for title/focus (matches `add_source`
-    semantics). Domain is required to stay in the configured set —
-    bumping the constraint is a separate decision.
-
-    Raises ValueError on validation failures and LookupError if the row
-    is missing. Caller owns the transaction boundary.
-    """
-    source = sess.query(Source).filter(Source.id == source_id).first()
-    if source is None:
-        raise LookupError(f"source id={source_id} not found")
-
-    if domain is not None:
-        if domain not in DOMAINS:
-            raise ValueError(
-                f"unknown domain {domain!r}; valid: {', '.join(DOMAINS)}"
-            )
-        source.domain = domain
-
-    if title is not None:
-        source.title = title.strip() or None
-
-    if focus is not None:
-        source.focus = focus.strip() or None
-
-    source.updated_at = utcnow()
-    sess.flush()
-    return source
-
-
-# Retry modes — keep the surface small. Anything else is a separate ask.
-RETRY_FULL = "full"
-RETRY_EXTRACT = "extract"
-RETRY_MODES = (RETRY_FULL, RETRY_EXTRACT)
-
-
-def reset_source_for_retry(
-    sess: Session,
-    *,
-    source_id: int,
-    mode: str = RETRY_FULL,
-) -> Source:
-    """Reset a source so the pipeline re-processes it.
-
-    Modes:
-      - "full": status → PENDING. Article pull or video download +
-        transcribe will happen again on the appropriate worker.
-      - "extract": status → TRANSCRIBED. Requires the row to already
-        carry a transcript (transcript_text). The dev scheduler will
-        re-run extraction without re-pulling the source. Cheap.
-
-    Always clears `claimed_by` / `claimed_at` / `error_message` so the
-    queue dance can pick it up cleanly and stale errors don't bleed
-    into the UI.
-
-    Raises ValueError on bad mode / missing transcript for extract mode,
-    and LookupError if the row is missing.
-    """
-    if mode not in RETRY_MODES:
-        raise ValueError(
-            f"unknown retry mode {mode!r}; valid: {', '.join(RETRY_MODES)}"
-        )
-
-    source = sess.query(Source).filter(Source.id == source_id).first()
-    if source is None:
-        raise LookupError(f"source id={source_id} not found")
-
-    if mode == RETRY_EXTRACT:
-        if not (source.transcript_text and source.transcript_text.strip()):
-            raise ValueError(
-                "extract-only retry requires an existing transcript; "
-                "use full retry instead"
-            )
-        source.status = SourceStatus.TRANSCRIBED
-    else:
-        source.status = SourceStatus.PENDING
-
-    source.claimed_by = None
-    source.claimed_at = None
-    source.error_message = None
-    source.updated_at = utcnow()
-    sess.flush()
-    return source
-
-
 __all__ = [
    "AddResult",
    "DEFAULT_PLAYLIST_MAX_ITEMS",
-    "EDITABLE_METADATA_FIELDS",
    "PlaylistAddResult",
-    "RETRY_EXTRACT",
-    "RETRY_FULL",
-    "RETRY_MODES",
    "add_playlist",
    "add_source",
    "expand_youtube_playlist",
    "is_article_url",
    "is_youtube_playlist_url",
-    "reset_source_for_retry",
-    "update_source_metadata",
    "validate_url",
 ]
--- a/src/second_brain/web/app.py
+++ b/src/second_brain/web/app.py
@ -31,14 +31,9 @@ from second_brain.settings_store import (
 )
 from second_brain.sources_service import (
    DEFAULT_PLAYLIST_MAX_ITEMS,
-    RETRY_EXTRACT,
-    RETRY_FULL,
-    RETRY_MODES,
    add_playlist,
    add_source,
    is_youtube_playlist_url,
-    reset_source_for_retry,
-    update_source_metadata,
 )

 # ---------------------------------------------------------------------------
@ -125,33 +120,16 @@ async def index(
    )


-def _render_source_detail(
-    request: Request,
-    source_id: int,
-    *,
-    edit: bool = False,
-    flash: Optional[str] = None,
-    error: Optional[str] = None,
-    form_overrides: Optional[dict] = None,
-) -> HTMLResponse:
-    """Single render path for /sources/{id} — used by GET + every HTMX
-    mutation endpoint so each action gets a consistent re-render.
-
-    `form_overrides` lets a failed edit-save bounce show the user's
-    in-progress values instead of the row's stored values.
-    """
+@app.get("/sources/{source_id}", response_class=HTMLResponse)
+async def source_detail(request: Request, source_id: int):
+    """Source detail page with extraction results and accept/reject actions."""
    db = get_database()
    with db.session() as sess:
        source = sess.query(Source).filter(Source.id == source_id).first()
        if not source:
            raise HTTPException(status_code=404, detail="Source not found")
        source_data = _source_to_dict(source)
-        extraction_data = (
-            _extraction_to_dict(source.extraction) if source.extraction else None
-        )
-
-    if form_overrides:
-        source_data = {**source_data, **form_overrides}
+        extraction_data = _extraction_to_dict(source.extraction) if source.extraction else None

    return templates.TemplateResponse(
        request,
@ -159,25 +137,10 @@ def _render_source_detail(
        {
            "source": source_data,
            "extraction": extraction_data,
-            "edit_mode": edit,
-            "flash": flash,
-            "error": error,
-            "domains": DOMAINS,
        },
    )


-@app.get("/sources/{source_id}", response_class=HTMLResponse)
-async def source_detail(request: Request, source_id: int, edit: bool = False):
-    """Source detail page with extraction results and accept/reject actions.
-
-    `?edit=1` opens the metadata editor inline. Editing exposes domain,
-    title, and focus — URL stays read-only since it's the UNIQUE dedupe
-    key for the queue and the embeddings identity tuple.
-    """
-    return _render_source_detail(request, source_id, edit=edit)
-
-
 _IN_FLIGHT_STATUSES = [
    SourceStatus.PENDING,
    SourceStatus.PULLED,
@ -229,11 +192,17 @@ async def accept_source(request: Request, source_id: int):
        if not source:
            raise HTTPException(status_code=404, detail="Source not found")
        source.status = SourceStatus.ACCEPTED
+        source_data = _source_to_dict(source)
+        extraction_data = _extraction_to_dict(source.extraction) if source.extraction else None

-    return _render_source_detail(
+    return templates.TemplateResponse(
        request,
-        source_id,
-        flash="Accepted — will be included in next wiki compile.",
+        "source_detail.html",
+        {
+            "source": source_data,
+            "extraction": extraction_data,
+            "flash": "Accepted — will be included in next wiki compile.",
+        },
    )


@ -246,80 +215,19 @@ async def reject_source(request: Request, source_id: int):
        if not source:
            raise HTTPException(status_code=404, detail="Source not found")
        source.status = SourceStatus.FAILED
+        source_data = _source_to_dict(source)
+        extraction_data = _extraction_to_dict(source.extraction) if source.extraction else None

-    return _render_source_detail(request, source_id, flash="Rejected.")
-
-
-@app.post("/sources/{source_id}/edit", response_class=HTMLResponse)
-async def edit_source_metadata(
-    request: Request,
-    source_id: int,
-    domain: str = Form(...),
-    title: Optional[str] = Form(None),
-    focus: Optional[str] = Form(None),
-):
-    """HTMX edit-mode save. Updates domain/title/focus only — URL is
-    intentionally not editable (see service-layer comment)."""
-    db = get_database()
-    try:
-        with db.session() as sess:
-            update_source_metadata(
-                sess,
-                source_id=source_id,
-                domain=domain,
-                title=title,
-                focus=focus,
-            )
-    except LookupError as exc:
-        raise HTTPException(status_code=404, detail=str(exc)) from exc
-    except ValueError as exc:
-        # Bounce back to edit mode with the user's in-progress values so
-        # they don't have to retype them.
-        return _render_source_detail(
+    return templates.TemplateResponse(
        request,
-            source_id,
-            edit=True,
-            error=str(exc),
-            form_overrides={
-                "domain": domain,
-                "title": title or "",
-                "focus": focus or "",
+        "source_detail.html",
+        {
+            "source": source_data,
+            "extraction": extraction_data,
+            "flash": "Rejected.",
        },
    )

-    return _render_source_detail(request, source_id, flash="Saved.")
-
-
-@app.post("/sources/{source_id}/retry", response_class=HTMLResponse)
-async def retry_source(
-    request: Request,
-    source_id: int,
-    mode: str = Form(RETRY_FULL),
-):
-    """HTMX retry/re-run. `mode=full` → PENDING (re-pull + re-process).
-    `mode=extract` → TRANSCRIBED (re-extract only; requires an existing
-    transcript). Always clears the claim pair and error_message."""
-    if mode not in RETRY_MODES:
-        raise HTTPException(status_code=400, detail=f"unknown mode {mode!r}")
-
-    db = get_database()
-    try:
-        with db.session() as sess:
-            source = reset_source_for_retry(
-                sess, source_id=source_id, mode=mode
-            )
-            new_status = source.status.value
-    except LookupError as exc:
-        raise HTTPException(status_code=404, detail=str(exc)) from exc
-    except ValueError as exc:
-        return _render_source_detail(request, source_id, error=str(exc))
-
-    if mode == RETRY_EXTRACT:
-        flash = f"Re-queued for extraction (status → {new_status})."
-    else:
-        flash = f"Re-queued from scratch (status → {new_status})."
-    return _render_source_detail(request, source_id, flash=flash)
-

 # ---------------------------------------------------------------------------
 # JSON API endpoints
@ -378,7 +286,6 @@ def _source_to_dict(s: Source) -> dict:
        "focus": s.focus,
        "ingested_at": s.ingested_at.strftime("%Y-%m-%d %H:%M") if s.ingested_at else "",
        "has_extraction": s.extraction is not None,
-        "has_transcript": bool((s.transcript_text or "").strip()),
        "error_message": s.error_message,
    }

--- a/src/second_brain/web/templates/source_detail.html
+++ b/src/second_brain/web/templates/source_detail.html
@ -11,19 +11,9 @@
        </svg>
        Back to sources
    </a>
-    <div class="flex items-start justify-between gap-4">
-        <div class="min-w-0">
-            <h2 class="text-2xl font-semibold text-gray-900 break-words">{{ source.title }}</h2>
+    <h2 class="text-2xl font-semibold text-gray-900">{{ source.title }}</h2>
    <a href="{{ source.url }}" target="_blank"
       class="text-sm text-indigo-500 hover:text-indigo-700 break-all">{{ source.url }}</a>
-        </div>
-        {% if not edit_mode %}
-        <a href="/sources/{{ source.id }}?edit=1"
-           class="shrink-0 px-3 py-1.5 text-sm border border-gray-300 text-gray-700 rounded hover:bg-gray-50 transition-colors">
-            Edit
-        </a>
-        {% endif %}
-    </div>
 </div>

 {% if flash %}
@ -31,54 +21,7 @@
    {{ flash }}
 </div>
 {% endif %}
-{% if error %}
-<div class="mb-4 px-4 py-3 bg-red-50 border border-red-200 text-red-800 rounded text-sm">
-    {{ error }}
-</div>
-{% endif %}

-{% if edit_mode %}
-<!-- Edit mode: domain / title / focus. URL is intentionally read-only -->
-<form hx-post="/sources/{{ source.id }}/edit"
-      hx-target="body"
-      hx-swap="innerHTML"
-      class="mb-8 bg-white rounded-lg border border-gray-200 p-5 space-y-4">
-    <div>
-        <label class="block text-xs font-semibold text-gray-500 uppercase tracking-wide mb-1">URL <span class="font-normal normal-case text-gray-400">(read-only — dedupe key)</span></label>
-        <input type="text" value="{{ source.url }}" disabled
-               class="w-full px-3 py-2 border border-gray-200 bg-gray-50 text-gray-500 text-sm rounded">
-    </div>
-    <div>
-        <label for="edit-title" class="block text-xs font-semibold text-gray-500 uppercase tracking-wide mb-1">Title</label>
-        <input type="text" id="edit-title" name="title" value="{{ source.title or '' }}"
-               class="w-full px-3 py-2 border border-gray-300 text-sm rounded focus:outline-none focus:ring-2 focus:ring-indigo-300">
-    </div>
-    <div>
-        <label for="edit-domain" class="block text-xs font-semibold text-gray-500 uppercase tracking-wide mb-1">Domain</label>
-        <select id="edit-domain" name="domain"
-                class="w-full px-3 py-2 border border-gray-300 text-sm rounded focus:outline-none focus:ring-2 focus:ring-indigo-300">
-            {% for d in domains %}
-            <option value="{{ d }}" {% if d == source.domain %}selected{% endif %}>{{ d }}</option>
-            {% endfor %}
-        </select>
-    </div>
-    <div>
-        <label for="edit-focus" class="block text-xs font-semibold text-gray-500 uppercase tracking-wide mb-1">Focus</label>
-        <textarea id="edit-focus" name="focus" rows="3"
-                  class="w-full px-3 py-2 border border-gray-300 text-sm rounded focus:outline-none focus:ring-2 focus:ring-indigo-300">{{ source.focus or '' }}</textarea>
-    </div>
-    <div class="flex gap-2 pt-1">
-        <button type="submit"
-                class="px-4 py-2 bg-indigo-600 text-white text-sm font-semibold rounded hover:bg-indigo-700 transition-colors">
-            Save
-        </button>
-        <a href="/sources/{{ source.id }}"
-           class="px-4 py-2 border border-gray-300 text-gray-700 text-sm font-semibold rounded hover:bg-gray-50 transition-colors">
-            Cancel
-        </a>
-    </div>
-</form>
-{% else %}
 <!-- Meta row -->
 <div class="flex flex-wrap gap-2 mb-6">
    {% set status_colors = {
@ -117,9 +60,9 @@
 </div>
 {% endif %}

-<!-- Actions row: accept/reject (when analyzed) + retry (always) -->
-<div class="flex flex-wrap gap-3 mb-8">
-    {% if source.status == 'analyzed' %}
+<!-- Accept / Reject actions (only when analyzed) -->
+{% if source.status == 'analyzed' %}
+<div class="flex gap-3 mb-8">
    <button hx-post="/sources/{{ source.id }}/accept"
            hx-target="body"
            hx-swap="innerHTML"
@ -132,31 +75,6 @@
            class="px-5 py-2 bg-red-500 text-white text-sm font-semibold rounded hover:bg-red-600 transition-colors">
        Reject
    </button>
-    {% endif %}
-
-    <!-- Re-run: full reset → PENDING. Re-pulls and re-processes from scratch. -->
-    <button hx-post="/sources/{{ source.id }}/retry"
-            hx-vals='{"mode": "full"}'
-            hx-target="body"
-            hx-swap="innerHTML"
-            hx-confirm="Re-run this source from scratch? Status will reset to PENDING and the pipeline will re-pull{% if source.source_type == 'video' %} and re-transcribe (expensive){% endif %}."
-            class="px-5 py-2 bg-amber-500 text-white text-sm font-semibold rounded hover:bg-amber-600 transition-colors">
-        Re-run from scratch
-    </button>
-
-    {% if source.has_transcript %}
-    <!-- Light retry: skip the re-pull, just re-run LLM extraction. Works
-         for any source with stored body text (videos have transcripts,
-         articles have the trafilatura-extracted body). -->
-    <button hx-post="/sources/{{ source.id }}/retry"
-            hx-vals='{"mode": "extract"}'
-            hx-target="body"
-            hx-swap="innerHTML"
-            hx-confirm="Re-extract only — keep the existing transcript/body and re-run the LLM extraction?"
-            class="px-5 py-2 bg-amber-100 text-amber-800 text-sm font-semibold rounded hover:bg-amber-200 border border-amber-300 transition-colors">
-        Re-extract only
-    </button>
-    {% endif %}
 </div>
 {% endif %}

--- a/tests/test_config_overlay.py
+++ b/tests/test_config_overlay.py
@ -1,165 +0,0 @@
-"""Tests for the settings.local.toml overlay loader.
-
-The overlay is the dev-side path for keeping host-specific secrets
-(notably `[database].url`) out of the tracked `settings.toml` without
-forcing every interactive shell to export env vars. The footgun this
-test suite guards against: a one-key local override (e.g. a single
-`[database] url = "..."`) silently wiping every other key in that
-section. The merge is section-level, not shallow.
-"""
-
-from __future__ import annotations
-
-from pathlib import Path
-
-import pytest
-
-
-# ---------------------------------------------------------------------------
-# Pure-Python: _overlay_sections
-# ---------------------------------------------------------------------------
-
-
-def test_overlay_preserves_sibling_keys_in_overlapping_section():
-    """The exact regression target — local `url` must not wipe `pool_size`."""
-    from second_brain.config import _overlay_sections
-
-    base = {"database": {"url": "", "pool_size": 5}}
-    local = {"database": {"url": "postgresql://x"}}
-
-    out = _overlay_sections(base, local)
-    assert out["database"]["url"] == "postgresql://x"
-    assert out["database"]["pool_size"] == 5
-
-
-def test_overlay_section_only_in_base_survives():
-    from second_brain.config import _overlay_sections
-
-    base = {"database": {"url": "a"}, "scheduler": {"interval": 30}}
-    local = {"database": {"url": "b"}}
-
-    out = _overlay_sections(base, local)
-    assert out["scheduler"] == {"interval": 30}
-    assert out["database"]["url"] == "b"
-
-
-def test_overlay_section_only_in_local_appears():
-    from second_brain.config import _overlay_sections
-
-    base = {"database": {"url": "a"}}
-    local = {"embeddings": {"ollama_url": "http://x"}}
-
-    out = _overlay_sections(base, local)
-    assert out["database"] == {"url": "a"}
-    assert out["embeddings"] == {"ollama_url": "http://x"}
-
-
-def test_overlay_scalar_in_overlay_replaces_dict_in_base():
-    """Non-dict overlay values replace outright — no merge attempt."""
-    from second_brain.config import _overlay_sections
-
-    base = {"feature": {"a": 1, "b": 2}}
-    local = {"feature": "off"}  # scalar wins outright
-
-    out = _overlay_sections(base, local)
-    assert out["feature"] == "off"
-
-
-def test_overlay_dict_in_overlay_replaces_scalar_in_base():
-    from second_brain.config import _overlay_sections
-
-    base = {"feature": "off"}
-    local = {"feature": {"a": 1}}
-
-    out = _overlay_sections(base, local)
-    assert out["feature"] == {"a": 1}
-
-
-def test_overlay_empty_overlay_is_identity():
-    from second_brain.config import _overlay_sections
-
-    base = {"database": {"url": "a", "pool_size": 5}}
-    assert _overlay_sections(base, {}) == base
-    assert _overlay_sections(base, None) == base
-
-
-def test_overlay_only_merges_one_level():
-    """Doc: the overlay is section-level. Nested dicts inside a section
-    replace, not deep-merge. This pins the documented depth."""
-    from second_brain.config import _overlay_sections
-
-    base = {"db": {"url": "a", "pool": {"min": 1, "max": 10}}}
-    local = {"db": {"pool": {"max": 20}}}
-
-    out = _overlay_sections(base, local)
-    # pool replaced wholesale — `min` does NOT survive. This is deliberate.
-    assert out["db"]["pool"] == {"max": 20}
-    assert out["db"]["url"] == "a"
-
-
-# ---------------------------------------------------------------------------
-# load_config: end-to-end with real TOML files on disk
-# ---------------------------------------------------------------------------
-
-
-def _write(path: Path, body: str) -> None:
-    path.write_text(body, encoding="utf-8")
-
-
-@pytest.fixture(autouse=True)
-def _reset_config():
-    """Drop the cached singleton before/after each test so overlay state
-    doesn't leak between cases."""
-    from second_brain.config import reset_config_singleton
-
-    reset_config_singleton()
-    yield
-    reset_config_singleton()
-
-
-def test_load_config_picks_up_sibling_local_overlay(tmp_path):
-    from second_brain.config import load_config
-
-    base = tmp_path / "settings.toml"
-    local = tmp_path / "settings.local.toml"
-    _write(
-        base,
-        '[database]\nurl = ""\n\n[embeddings]\nollama_url = "http://ollama:11434"\nmodel = "nomic-embed-text"\n',
-    )
-    _write(
-        local,
-        '[database]\nurl = "postgresql+psycopg://lovebug:x@127.0.0.1:5433/petalbrain"\n\n[embeddings]\nollama_url = "http://127.0.0.1:11434"\n',
-    )
-
-    cfg = load_config(base)
-    # url from local wins; default model survives (not in local).
-    assert cfg.database["url"].startswith("postgresql+psycopg://")
-    assert cfg.embeddings["ollama_url"] == "http://127.0.0.1:11434"
-    assert cfg.embeddings["model"] == "nomic-embed-text"
-
-
-def test_load_config_no_local_overlay_is_inert(tmp_path):
-    from second_brain.config import load_config
-
-    base = tmp_path / "settings.toml"
-    _write(base, '[database]\nurl = "from-base"\n')
-
-    cfg = load_config(base)
-    assert cfg.database["url"] == "from-base"
-
-
-def test_load_config_local_overlay_at_explicit_path(tmp_path):
-    """Doc: when an explicit (or SECOND_BRAIN_CONFIG) path is used, the
-    overlay sibling is resolved next to THAT path — not next to the
-    repo's config dir."""
-    from second_brain.config import load_config
-
-    subdir = tmp_path / "etc"
-    subdir.mkdir()
-    base = subdir / "settings.toml"
-    local = subdir / "settings.local.toml"
-    _write(base, '[database]\nurl = "base"\n')
-    _write(local, '[database]\nurl = "overlay-wins"\n')
-
-    cfg = load_config(base)
-    assert cfg.database["url"] == "overlay-wins"
--- a/tests/test_source_edit_retry.py
+++ b/tests/test_source_edit_retry.py
@ -1,258 +0,0 @@
-"""Tests for the source edit-metadata + retry-reset service helpers.
-
-Live-DB pattern matches `tests/test_claim.py` / `tests/test_playlist.py`:
-both groups auto-skip when no petalbrain connection is configured. The
-service layer is where edit/retry logic lives (the web routes + CLI are
-thin wrappers around it), so this is where the meaningful coverage is.
-"""
-
-from __future__ import annotations
-
-import os
-import uuid
-
-import psycopg
-import pytest
-
-
-def _db_reachable() -> bool:
-    url = os.environ.get("SECOND_BRAIN_DATABASE_URL") or os.environ.get(
-        "HERBYLAB_DATABASE_URL"
-    )
-    if not url:
-        return False
-    raw = url.replace("postgresql+psycopg://", "postgresql://", 1)
-    try:
-        with psycopg.connect(raw, connect_timeout=2):
-            return True
-    except Exception:
-        return False
-
-
-pytestmark = pytest.mark.skipif(
-    not _db_reachable(),
-    reason="needs SECOND_BRAIN_DATABASE_URL pointed at a real petalbrain",
-)
-
-
-def _seed_source(db, **overrides):
-    """Insert a Source row and return (id, url). Caller cleans up by id."""
-    from second_brain.models import Source, SourceStatus, SourceType, utcnow
-
-    tag = uuid.uuid4().hex[:8]
-    defaults = dict(
-        url=f"https://example.test/edit-retry/{tag}",
-        title=f"seed-{tag}",
-        domain="development",
-        source_type=SourceType.ARTICLE,
-        status=SourceStatus.PENDING,
-        ingested_at=utcnow(),
-    )
-    defaults.update(overrides)
-    with db.session() as sess:
-        src = Source(**defaults)
-        sess.add(src)
-        sess.flush()
-        return src.id, src.url
-
-
-def _cleanup(db, source_id):
-    from second_brain.models import Source
-
-    with db.session() as sess:
-        sess.query(Source).filter(Source.id == source_id).delete(
-            synchronize_session=False
-        )
-
-
-# ---------------------------------------------------------------------------
-# update_source_metadata
-# ---------------------------------------------------------------------------
-
-
-def test_update_metadata_writes_domain_title_focus():
-    from second_brain.database import Database, reset_database_singleton
-    from second_brain.models import Source
-    from second_brain.sources_service import update_source_metadata
-
-    reset_database_singleton()
-    db = Database()
-    sid, _ = _seed_source(db)
-    try:
-        with db.session() as sess:
-            update_source_metadata(
-                sess,
-                source_id=sid,
-                domain="homelab",
-                title="new title",
-                focus="new focus directive",
-            )
-        with db.session() as sess:
-            row = sess.query(Source).filter(Source.id == sid).first()
-            assert row.domain == "homelab"
-            assert row.title == "new title"
-            assert row.focus == "new focus directive"
-            assert row.updated_at is not None
-    finally:
-        _cleanup(db, sid)
-        reset_database_singleton()
-
-
-def test_update_metadata_blank_strings_become_null():
-    from second_brain.database import Database, reset_database_singleton
-    from second_brain.models import Source
-    from second_brain.sources_service import update_source_metadata
-
-    reset_database_singleton()
-    db = Database()
-    sid, _ = _seed_source(db)
-    try:
-        with db.session() as sess:
-            update_source_metadata(
-                sess, source_id=sid, domain="development", title="   ", focus=""
-            )
-        with db.session() as sess:
-            row = sess.query(Source).filter(Source.id == sid).first()
-            assert row.title is None
-            assert row.focus is None
-    finally:
-        _cleanup(db, sid)
-        reset_database_singleton()
-
-
-def test_update_metadata_rejects_unknown_domain():
-    from second_brain.database import Database, reset_database_singleton
-    from second_brain.sources_service import update_source_metadata
-
-    reset_database_singleton()
-    db = Database()
-    sid, _ = _seed_source(db)
-    try:
-        with db.session() as sess:
-            with pytest.raises(ValueError, match="unknown domain"):
-                update_source_metadata(sess, source_id=sid, domain="not-a-domain")
-    finally:
-        _cleanup(db, sid)
-        reset_database_singleton()
-
-
-def test_update_metadata_missing_source_raises_lookup():
-    from second_brain.database import Database, reset_database_singleton
-    from second_brain.sources_service import update_source_metadata
-
-    reset_database_singleton()
-    db = Database()
-    try:
-        with db.session() as sess:
-            with pytest.raises(LookupError):
-                update_source_metadata(
-                    sess, source_id=-99999, domain="development"
-                )
-    finally:
-        reset_database_singleton()
-
-
-# ---------------------------------------------------------------------------
-# reset_source_for_retry
-# ---------------------------------------------------------------------------
-
-
-def test_retry_full_resets_to_pending_and_clears_claim():
-    from second_brain.database import Database, reset_database_singleton
-    from second_brain.models import Source, SourceStatus, utcnow
-    from second_brain.sources_service import reset_source_for_retry
-
-    reset_database_singleton()
-    db = Database()
-    sid, _ = _seed_source(
-        db,
-        status=SourceStatus.FAILED,
-        error_message="boom",
-        claimed_by="worker-x",
-        claimed_at=utcnow(),
-    )
-    try:
-        with db.session() as sess:
-            reset_source_for_retry(sess, source_id=sid, mode="full")
-        with db.session() as sess:
-            row = sess.query(Source).filter(Source.id == sid).first()
-            assert row.status == SourceStatus.PENDING
-            assert row.claimed_by is None
-            assert row.claimed_at is None
-            assert row.error_message is None
-    finally:
-        _cleanup(db, sid)
-        reset_database_singleton()
-
-
-def test_retry_extract_sets_transcribed_when_transcript_present():
-    from second_brain.database import Database, reset_database_singleton
-    from second_brain.models import Source, SourceStatus
-    from second_brain.sources_service import reset_source_for_retry
-
-    reset_database_singleton()
-    db = Database()
-    sid, _ = _seed_source(
-        db,
-        status=SourceStatus.ANALYZED,
-        transcript_text="word " * 50,
-        error_message="prior failure",
-    )
-    try:
-        with db.session() as sess:
-            reset_source_for_retry(sess, source_id=sid, mode="extract")
-        with db.session() as sess:
-            row = sess.query(Source).filter(Source.id == sid).first()
-            assert row.status == SourceStatus.TRANSCRIBED
-            assert row.error_message is None
-            assert row.transcript_text  # transcript preserved
-    finally:
-        _cleanup(db, sid)
-        reset_database_singleton()
-
-
-def test_retry_extract_without_transcript_raises():
-    from second_brain.database import Database, reset_database_singleton
-    from second_brain.models import SourceStatus
-    from second_brain.sources_service import reset_source_for_retry
-
-    reset_database_singleton()
-    db = Database()
-    sid, _ = _seed_source(db, status=SourceStatus.PENDING, transcript_text=None)
-    try:
-        with db.session() as sess:
-            with pytest.raises(ValueError, match="requires an existing transcript"):
-                reset_source_for_retry(sess, source_id=sid, mode="extract")
-    finally:
-        _cleanup(db, sid)
-        reset_database_singleton()
-
-
-def test_retry_unknown_mode_raises():
-    from second_brain.database import Database, reset_database_singleton
-    from second_brain.sources_service import reset_source_for_retry
-
-    reset_database_singleton()
-    db = Database()
-    sid, _ = _seed_source(db)
-    try:
-        with db.session() as sess:
-            with pytest.raises(ValueError, match="unknown retry mode"):
-                reset_source_for_retry(sess, source_id=sid, mode="bogus")
-    finally:
-        _cleanup(db, sid)
-        reset_database_singleton()
-
-
-def test_retry_missing_source_raises_lookup():
-    from second_brain.database import Database, reset_database_singleton
-    from second_brain.sources_service import reset_source_for_retry
-
-    reset_database_singleton()
-    db = Database()
-    try:
-        with db.session() as sess:
-            with pytest.raises(LookupError):
-                reset_source_for_retry(sess, source_id=-99999, mode="full")
-    finally:
-        reset_database_singleton()