AI-powered document review with verifiable citations. Upload PDFs, define extraction columns, and let AI populate every cell with answers, reasoning, and exact source references — all auto-saved, all verifiable, all agent-ready.
| Layer | Technology | Why |
|---|---|---|
| API Server | Python / FastAPI | Best CRUD framework, Pydantic validation, AI ecosystem |
| Extraction Worker | Python | OpenAI/Anthropic SDKs, RAG pipeline, IO-bound |
| Document Worker | Python + LiteParse | Local PDF parsing with OCR + bounding boxes for citations |
| SSE Service | Go | Goroutines, 2KB/connection, purpose-built for long-lived connections |
| Frontend | React / TypeScript / shadcn | Type safety, component library, Zustand state |
| Database | PostgreSQL + pgvector | Relational data + vector similarity search in one |
| Queue / Events | Redis | Task queues + pub/sub for real-time events |
| Storage | S3 / MinIO | PDF storage, presigned URL uploads |
| Monorepo | Taskfile | Language-agnostic task runner, no framework lock-in |
| Document | Description |
|---|---|
| Product | What we're building, for whom, scope, user journey |
| Architecture | System design, data flows, decisions, scale estimates |
| API Reference | Every endpoint — request, response, status codes |
| Database Schema | Tables, columns, types, indexes, relationships |
| API Server | FastAPI internals, route organization, config |
| Document Worker | LiteParse pipeline, chunking, embeddings |
| Extraction Worker | RAG pipeline, LLM prompts, cell extraction |
| SSE Service | Go service, connection management, event routing |
| Frontend | Components, state, layout system, design system |
| Infrastructure | Docker, Redis queues, S3, event catalog |
| Local Setup | Clone to running in 5 minutes |
git clone https://github.com/june-legal/review-table.git
cd review-table
cp .env.example .env
task dev
# Open http://localhost:3000