Skip to content

Releases: DriftOS/driftos-embed

v0.1.0 - Initial Release

09 Dec 04:45

Choose a tag to compare

driftos-embed v0.1.0

Lightweight semantic conversation routing engine using local embeddings.

Features

  • Embedding-based drift detection - Sub-200ms latency, zero LLM costs for routing
  • Three routing actions: STAY, BRANCH, ROUTE
  • Python sidecar embedding server - Uses paraphrase-MiniLM-L6-v2
  • Optional LLM-powered fact extraction - Groq, OpenAI, or Anthropic
  • Prometheus metrics - Full observability
  • Grafana dashboards - Pre-configured visualizations
  • Load testing script - Realistic conversation simulation

Quick Start

git clone https://github.com/DriftOS/driftos-embed
cd driftos-embed
make setup    # Install dependencies
make up       # Start Docker services + database
make dev      # Start embedding server + API

API

curl -X POST http://localhost:3000/api/v1/drift/route \
  -H "Content-Type: application/json" \
  -d '{"conversationId": "conv-1", "content": "Hello world", "role": "user"}'

Useful Commands

make help       # Show all available commands
make dev        # Start both servers
make down       # Stop everything
make logs       # View Docker logs
make test-drift # Test drift detection