Python SDK Reference
This reference documents the Synth AI Python SDK, auto-generated from source docstrings.Status Legend
Pages are tagged with their stability status:Experimental
Early-stage features available by request. APIs may change significantly.
Alpha
Generally available but actively evolving. APIs may have breaking changes.
Beta
Stable APIs with minor changes possible. Suitable for production with caution.
Stable
Production-ready APIs with backwards compatibility guarantees.
Module Overview
Training APIs
| Module | Status | Description |
|---|---|---|
| GraphGen | Alpha | Automated design of agentic systems |
| Prompt Learning | Alpha | GEPA (Alpha) and MIPRO (Experimental) prompt optimization |
| RL/GSPO | Experimental | Reinforcement learning training |
| SFT | Experimental | Supervised fine-tuning |
Inference APIs
| Module | Status | Description |
|---|---|---|
| Graph Completions | Alpha | Graph inference (policies, verifiers, RLM) |
Evaluation APIs
| Module | Status | Description |
|---|---|---|
| Eval | Alpha | Task app rollout evaluation |
Task & Environment
| Module | Description |
|---|---|
| Task App | Task application framework |
| In-Process Task | In-process task execution |