A desktop RAG chatbot that indexes your entire home directory and lets you search it with natural language using local LLMs.
- Crawls and indexes files in your home directory using FAISS vector storage
- Embeds document chunks via Ollama (nomic-embed-text)
- Answers natural language queries about your files using local LLMs (deepseek-r1, gemma3, etc.)
- Floating desktop bubble UI — always on top, click to expand
| Layer |
Technology |
| LLM + Embeddings |
Ollama (local, no API keys) |
| Backend |
Flask (Python) |
| Vector Store |
FAISS |
| Desktop UI |
Electron |
| RAG Pipeline |
Custom chunking + embedding + retrieval |
User Query
↓
Flask API (:8787)
↓
RAG Engine (retrieves relevant chunks from FAISS)
↓
Ollama LLM generates response
↓
Electron UI displays answer + source files
# Install Python deps
pip install flask flask-cors httpx faiss-cpu
# Install Ollama + pull models
ollama pull deepseek-r1:1.5b
ollama pull nomic-embed-text
# Install Electron deps
npm install
# Start the server
python3 server.py
# Launch the desktop app
npx electron .
| Method |
Path |
Description |
| GET |
/api/status |
Index status and model info |
| POST |
/api/index |
Trigger file indexing |
| POST |
/api/query |
Ask a question about your files |
| GET |
/api/recent |
Recently modified files |
- Local-first: No cloud APIs, no data leaves your machine
- Floating bubble UI: Click to expand into chat, click away to shrink
- Source citation: Every answer includes the file paths it came from
- Incremental indexing: Only re-embeds new or changed files