Personal AI knowledge base with hybrid RAG retrieval. Ingest articles, PDFs, or URLs — ask questions and get cited answers grounded in your own research.
- Hybrid Search — every query runs vector similarity + full-text keyword search in parallel, then fuses the results for comprehensive recall
- Cited Q&A — ask a question in natural language, get a grounded answer with source citations and relevance scores
- Multi-Source Ingest — paste text, drop a URL, or upload a PDF/DOCX — content is extracted, chunked, embedded, and indexed automatically
- Wiki Browser — browse, search, and filter your knowledge base with instant FTS5-powered keyword search
- Eval-Gated Quality — a retrieval quality harness gates every change to the system; if search quality drops, the change doesn't ship
| Layer | Technology |
|---|---|
| Backend | Python 3.12 · FastAPI · SQLite (FTS5 + sqlite-vec) |
| LLM | OpenRouter (Gemma 4 26B, free) · fastembed (local ONNX embeddings) |
| Frontend | Next.js 16 · React 19 · Tailwind v4 · shadcn/ui |
| Database | Single-file SQLite with FTS5 full-text search + sqlite-vec vector search |
| Deployment | Render (Docker + persistent disk) · Vercel (Next.js) |
| Testing | pytest (150+ tests) · Vitest · Biome |
Portal (Next.js) Server (FastAPI)
┌──────────────────────┐ ┌──────────────────────────┐
│ Wiki Browser │──SQLite──→ │ │
│ Keyword Search │ (readonly) │ Hybrid Retrieval │
│ Landing Page │ │ vec_search + fts_search │
│ │ │ → score-based merge │
│ Ask AI ────────────│──POST /chat─│ │
│ Add Source ─────────│──POST /api──│ Provider (OpenRouter) │
│ File Upload ────────│──POST /api──│ Embeddings (fastembed) │
└──────────────────────┘ └────────────┬─────────────┘
│
manifest.db
SQLite + FTS5 + sqlite-vec
- Python 3.12+, uv
- Node.js 20+, bun
- OpenRouter API key (free)
# Clone and setup
git clone https://github.com/samueldanso/evomind.git
cd evomind
# Seed the database with demo content + embeddings
uv run scripts/seed.py
# Start the backend
OPENROUTER_API_KEY=your-key uvicorn server:app --port 8765
# In another terminal — start the portal
cd portal && bun install && bun devFrom the UI: Go to Add Source → paste text, drop a URL, or upload a PDF.
From the CLI:
uv run scripts/ingest.py --title "..." --slug "..." --tags "..." --summary "..." --html /path/to/file.html
uv run scripts/embed.py --incrementaluv run pytest # Python (150+ tests)
cd portal && bun test # Portal tests
cd portal && bun run build # Build check| Service | Platform | Config |
|---|---|---|
| Backend (FastAPI) | Render | render.yaml — Docker + persistent disk |
| Frontend (Next.js) | Vercel | portal/vercel.json — root dir portal |
Render (backend):
| Variable | Value |
|---|---|
OPENROUTER_API_KEY |
Your key (set as secret) |
EVO_STORE |
/data |
ALLOWED_ORIGIN |
Your Vercel URL |
Vercel (frontend):
| Variable | Value |
|---|---|
EVO_SERVER_URL |
Your Render URL |
MIT