Documentation

Build with Bee

Comprehensive guides for running, training, evaluating, and deploying Bee. From local inference on a MacBook to cloud GPU training.

Documentation Sections

Server Architecture

FastAPI server with REST and WebSocket chat endpoints. Domain adapter switching, document upload, and feedback collection.

Model & LoRA

SmolLM2-360M-Instruct base model with domain-specific LoRA adapters. 5.2M trainable parameters per domain.

RAG & Document Grounding

FAISS + sentence-transformers for semantic retrieval. Upload documents, get cited answers.

Training Pipeline

Real instruction data from OpenOrca, CodeAlpaca, OpenHermes. LoRA training on cloud GPU with eval gating.

Evaluation Harness

5 benchmarks: coding, reasoning, instruction following, grounded facts, domain knowledge. Compare before/after every training run.

Cloud Training

Offload training to RunPod, Vast.ai, Google Colab, or Modal.com. Sync checkpoints via HuggingFace Hub.

Quantum Integration

Optional IBM Quantum for certified randomness and experimental HPO. Opt-in only, free-tier aware.

API Reference

OpenAI-compatible /v1/chat/completions, document upload, feedback endpoints, WebSocket streaming.

Quick Start

1. Start the Server

cd /path/to/Bee
source .venv/bin/activate
BEE_HOST=127.0.0.1 BEE_PORT=8000 python3 -m bee.server

2. Test the API

curl http://127.0.0.1:8000/health

curl -X POST http://127.0.0.1:8000/v1/chat/completions \\
  -H "Content-Type: application/json" \\
  -d '{"model":"bee","messages":[{"role":"user","content":"Hello"}]}'

3. Upload a Document

curl -X POST http://127.0.0.1:8000/v1/documents/upload \\
  -H "Content-Type: application/json" \\
  -d '{
    "source": "guide.txt",
    "content": "Bee supports cybersecurity, quantum, fintech..."
  }'

Ready to start building?

Full documentation is available in the docs/ directory of the repository.

Launch Bee