Synth AI OSS
Tools to run evals, trace LLM calls, and fine‑tune open models using a simple Python/CLI toolkit.Install
Try it quickly
- Check your balance (prod backend):
- Explore examples:
What you can do
- Configure simple and modular evals and metrics for agent runs
- Run inference on and finetune Open models via Synth endpoints to iterate on hypotheses
- Inspect every LLM/tool call with structured tracing
- And more
Jump in
- Evals Demo: compare models and export datasets
- Rejection Training: generate → filter → train → run
- On-Policy RL: deploy task app → mint env key → check → run RL
API
- HTTP endpoints (inference, files, SFT, models, balance):
- See the concise spec: HTTP API Reference