dd75a67964
Brings the bureau from 4 → 8 detectives. All eight run as Bun + claude-CLI
subprocesses against the same Supabase + investigation_jobs LISTEN/NOTIFY
queue, sharing search.ts hybridSearch and writer-side validators that
gate writes against schema + FK.
New detectives:
Poirot (witness_analysis)
- prompts/poirot.md — credibility / access / bias / corroboration /
verdict; uses entity_mentions JOIN chunks to pull 12 chunks per
person; resolves corroboration_refs chunk_ids defensively (accepts
bare cNNNN even when the model emits pNNN/cNNNN).
- INSERT into public.witnesses with W-NNNN naming.
- Tone: purple (#9b5de5).
Taleb (outlier_scan)
- prompts/taleb.md — "surprise is relative to a model"; at most 3
outliers; each requires explicit dominant_model + why_surprising +
what_it_implies; fan-out into public.gaps with scope.kind="outlier".
- Same unscoped-fallback as Dupin (Pass 1 with doc_id, Pass 2 widens
to corpus if hits < 3).
- Tone: yellow (#ffd23f).
Tetlock (calibrate_hypothesis)
- prompts/tetlock.md — honest Bayesian update; emits new_posterior +
Δ + recommended_action ∈ {keep, downgrade, upgrade, supersede}.
- write_calibration UPDATEs public.hypotheses + APPENDS a
"## Calibration history" section to the H-NNNN.md case file
(calibration is append-only — each datapoint matters). Posterior
band auto-corrected to match Tetlock thresholds.
- NO_NEW_EVIDENCE sentinel handled; pure 'keep' with |Δ|<0.005 only
touches updated_at + reviewed_by.
- Tone: teal (#26d4cc).
Case-Writer (case_report)
- prompts/case-writer.md — Dr. Watson assembles all artefacts
(E-NNNN, H-NNNN, R-NNNN, W-NNNN, G-NNNN) into a five-act narrative.
ILIKE filter on topic; doc_id optional scope.
- Larger budget cap (≥ $0.50) + longer timeout for prose generation.
- Writes case/reports/<slug>.md with frontmatter (topic + counts);
no DB table for v0.
- New page /c/[slug] renders the report via MarkdownBody + stat chips.
- Tone: gold (#e0c080).
Hardening across the bureau:
- Sentinel parsing now accepts backticked AND prose-trailing forms
(Holmes NO_HYPOTHESES, Dupin NO_CONTRADICTIONS, Schneier
INSUFFICIENT_HYPOTHESIS, Poirot INSUFFICIENT_TESTIMONY, Taleb
NO_OUTLIERS, Tetlock NO_NEW_EVIDENCE, Case-Writer
INSUFFICIENT_ARTEFACTS). Avoids the failure mode where the model
refuses honestly but the runtime treated it as a parse error
(observed live with Poirot+Hoover identifying the DIRECTOR
false-positive disambiguation issue in entity_mentions).
Chat tool extensions (web/lib/chat/tools.ts):
- request_investigation now accepts 7 kinds. Each routes to its
detective with appropriate validation (hypothesis_id regex,
person_id kebab-case, topic non-empty, doc_id for evidence_chain).
- ETA per kind: Holmes/Dupin 60s, Poirot 45s, Schneier/Tetlock 30s,
Taleb 50s, Case-Writer 180s (longer prose), Locard 30×n_chunks.
UI integration:
- chat-bubble inline card paints each detective in its tone color.
- /jobs/[id] page header swaps name/subtitle/tone per detective;
question label adapts ("Topic" / "Hypothesis under attack" /
"Witness under analysis" / "Topic to outlier-scan" / "Hypothesis
under recalibration" / "Case to assemble").
- job-status-poller renders: case-report link card (gold), outlier
cards (yellow), witness cards (purple) — alongside existing
hypothesis, evidence, contradiction cards.
- /api/jobs/[id] hydrates witnesses (JOIN entities for canonical_name)
+ gaps (with scope JSONB).
- /c/[slug] page reads /data/ufo/case/reports/<slug>.md and renders
with MarkdownBody, frontmatter parsed for stat chips.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
|
||
|---|---|---|
| .. | ||
| app | ||
| components | ||
| lib | ||
| .dockerignore | ||
| .env.local.example | ||
| Dockerfile | ||
| instrumentation.ts | ||
| middleware.ts | ||
| next-env.d.ts | ||
| next.config.ts | ||
| package-lock.json | ||
| package.json | ||
| postcss.config.mjs | ||
| README.md | ||
| sentry.client.config.ts | ||
| sentry.server.config.ts | ||
| tailwind.config.ts | ||
| tsconfig.json | ||
web — Disclosure Bureau Next.js app
Next.js 15 + React 19 + Tailwind + Supabase + assistant-ui.
Quick start (local dev)
# 1. Install deps
npm install
# 2. (Optional) Start local Supabase
# Requires Docker. Skip if pointing at remote Supabase.
npx supabase init # first time only — creates supabase/ folder
npx supabase start # spins up Postgres/GoTrue/Storage on :54321
# 3. Configure env
cp .env.local.example .env.local
# Edit .env.local — paste local Supabase keys (printed by `supabase start`)
# 4. Apply migrations
psql postgresql://postgres:postgres@localhost:54322/postgres \
-f ../infra/supabase/migrations/0001_chat_schema.sql
# 5. Start dev
npm run dev
# http://localhost:3030
Without Supabase
The app degrades gracefully if Supabase env vars are unset:
- Wiki browsing works (read-only from filesystem)
- Auth bar shows "auth: disabled (dev)"
- Chat bubble shows "Auth not configured"
Useful for quick UI work without spinning up Docker.
Production (Coolify on VPS)
See ../infra/coolify/. Stack:
- Coolify orchestrates everything
- Supabase self-hosted:
db.disclosure.top,studio.disclosure.top - Next.js:
disclosure.top - Meilisearch (shared):
search.disclosure.top - Imgproxy (shared):
img.disclosure.top - Caddy: TLS + reverse proxy (built into Coolify)
Architecture
app/
├── page.tsx # home — 116 docs grouped by collection
├── auth/
│ ├── signin/page.tsx # magic-link form
│ ├── callback/route.ts # exchanges code for session
│ └── signout/route.ts
├── d/[docId]/
│ ├── page.tsx # doc detail
│ └── [page]/page.tsx # page reader (OCR + entity highlights + crops + sidebar PNG)
├── api/
│ ├── me/route.ts # GET current profile
│ ├── sessions/route.ts # GET list, POST new
│ ├── sessions/[id]/route.ts # GET detail, PATCH, DELETE
│ ├── sessions/[id]/messages/route.ts # POST send → assistant reply
│ ├── documents/, pages/, entities/, tables/ # read-only data
│ └── static/[...path]/route.ts # sandboxed file serve
components/
├── chat-bubble.tsx # floating Sherlock — auth-aware, session list
├── entity-modal.tsx # opens on entity click
├── reader-content.tsx # OCR + highlights + crops
└── auth-bar.tsx # sign in / out + budget tracker
lib/
├── wiki.ts # markdown reader (gray-matter)
├── entity-index.ts # match loader + text segmentation
└── supabase/{server,client}.ts # SSR helpers
middleware.ts # session refresh on every request
Tech notes
- No RAG: chat agent reads markdown directly. Wiki-link traversal substitutes for vector search.
- RLS-first: Supabase Row Level Security enforces "user sees only own sessions" at the DB layer.
- Magic-link auth: no passwords. GoTrue handles email delivery.
- Anti-abuse: per-user budget cap (default $5) + daily message quota (default 100) enforced via
check_budgetRPC before each Claude call.
Cost
Each chat turn costs ~$0.005-0.05 depending on context size (mostly Haiku $1/M input, $5/M output).