osmosis-ai 0.2.17

A Python SDK for Osmosis LLM training workflows: reward/rubric validation and remote rollout.

osmosis-ai

⚠️ Warning: osmosis-ai is still in active development. APIs may change between versions.

Python SDK for Osmosis AI training workflows. Supports two training modes with shared tooling for testing and evaluation.

Osmosis AI is a platform for training LLMs with reinforcement learning. You define custom reward functions, LLM-as-judge rubrics, and agent tools -- then Osmosis handles the training loop on managed GPU clusters. This SDK provides everything you need to build and test those components locally, from @osmosis_reward decorators and MCP tool definitions to a full CLI for running agents against datasets before submitting training runs.

Quick Start

Pick a training mode and follow the example repo:

• Local Rollout (recommended for most users): osmosis-git-sync-example
• Remote Rollout (custom agent architectures): osmosis-remote-rollout-example

Two Training Modes

Osmosis supports Local Rollout and Remote Rollout as parallel approaches to training with reinforcement learning:

	Local Rollout	Remote Rollout
How it works	Osmosis manages the agent loop. You provide reward functions, rubrics, and MCP tools via a GitHub-synced repo.	You implement and host a RolloutAgentLoop server. Full control over agent behavior.
Best for	Standard tool-use agents, fast iteration, zero infrastructure	Custom agent architectures, complex orchestration, persistent environments
Example repo	osmosis-git-sync-example	osmosis-remote-rollout-example

Installation

Requires Python 3.10 or newer. For development setup, see CONTRIBUTING.md.

Prerequisites

• Python 3.10+
• An LLM API key (e.g., OpenAI, Anthropic, Groq) -- required for osmosis test and osmosis eval. See supported providers.
• Osmosis account (optional) -- needed for platform features like osmosis login, workspace management, and submitting training runs. Sign up at platform.osmosis.ai.

pip

pip install osmosis-ai            # Core SDK
pip install osmosis-ai[server]    # FastAPI server for Remote Rollout
pip install osmosis-ai[mcp]       # MCP tool support for Local Rollout
pip install osmosis-ai[full]      # All features

uv

uv add osmosis-ai                 # Core SDK
uv add osmosis-ai[server]         # FastAPI server for Remote Rollout
uv add osmosis-ai[mcp]            # MCP tool support for Local Rollout
uv add osmosis-ai[full]           # All features

Local Rollout

Osmosis manages the agent loop. You provide reward functions, rubrics, and MCP tools via a GitHub-synced repo.

Get started: osmosis-git-sync-example | Docs

Remote Rollout

You implement and host a RolloutAgentLoop server. Full control over agent behavior.

Get started: osmosis-remote-rollout-example | Docs

Testing & Evaluation

Both modes share the same CLI tools: Test Mode | Eval Mode | CLI Reference

Contributing

See CONTRIBUTING.md for development setup, testing, linting, and PR guidelines.

License

MIT License - see LICENSE file for details.

Links

• Homepage
• Issues
• Local Rollout Example
• Remote Rollout Example