Signalboard - RAG Intelligence Copilot

Local-first MVP for ingesting project knowledge (meeting notes, CSV trackers, manual uploads), processing it asynchronously, and answering PM/TPM questions with citations.

Architecture

• Frontend: React + Vite
• Backend: FastAPI
• Database: Postgres + pgvector
• Queue: Redis + RQ
• Worker: Python RQ worker
• Embeddings: Ollama nomic-embed-text (768 dimensions)
• LLM: Ollama qwen2.5:7b (fallback: qwen2.5:3b on low-RAM machines)
• Packaging: Docker Compose

Prerequisites

1. Docker Desktop (with Compose)
2. Ollama running on the host machine

Pull required models on the host:

ollama pull nomic-embed-text
ollama pull qwen2.5:7b

For machines with 8 GB RAM or less, use qwen2.5:3b instead and set OLLAMA_LLM_MODEL=qwen2.5:3b in .env.

Quick Start

cd rag_intelligence_copilot/deccan-ai-copilot
cp .env.example .env
docker compose up --build

Services:

Service	URL
Frontend	http://localhost:5173
Backend	http://localhost:8000
Postgres	localhost:5432
Redis	localhost:6379

API docs: http://localhost:8000/docs

Smoke Test

Recommended: use the dummy data pack at ../dummy_data/ — 5 PM-themed text files with full upload and chat test instructions. No manual metadata required.

Quick test:

1. Open http://localhost:5173
2. Upload tab → select all 5 files from ../dummy_data/ (bulk upload)
3. Click Upload & Process and wait for all jobs completed
4. Chat: ask Why is Project Phoenix delayed? with project filter Phoenix
5. Confirm answer includes citations (see dummy_data/README.md for more questions)
6. Dashboard shows document/chunk/job counts

Supported upload formats: .txt, .md, .csv, .pdf, .docx (max 10 MB per file).

API Endpoints

Method	Path	Description
GET	/health	Health check
POST	/documents/upload	Upload document (JSON)
POST	/documents/upload/bulk	Upload files (multipart)
GET	/documents	List documents
GET	/jobs/{job_id}	Job status
POST	/chat/query	Ask question
GET	/dashboard/summary	Stats

Environment Variables

See [.env.example](.env.example). Key settings:

• OLLAMA_BASE_URL — default http://host.docker.internal:11434 for Docker → host Ollama
• EMBEDDING_DIMENSION — must match model (768 for nomic-embed-text)
• CHUNK_SIZE_CHARS / CHUNK_OVERLAP_CHARS — default 3500 / 500
• UPLOAD_MAX_BYTES — max file size (default 10 MB)
• ALLOWED_UPLOAD_EXTENSIONS — .txt,.md,.csv,.pdf,.docx

Project Structure

main/
  backend/app/          # FastAPI, services, workers
  frontend/src/         # React UI
  docker-compose.yml