omium 0.4.0

Omium SDK for Python - Fault-tolerant agent operating system

Omium Python SDK

Python SDK and CLI for the Omium platform: tracing, checkpoints, and workflow management. Use the HTTP/remote API with api.omium.ai for the hosted platform, or the CLI for local runs and project push.

Documentation: docs.omium.ai — install, configure, CLI reference, configuration.

Installation

From PyPI (recommended)

python -m pip install --upgrade pip
python -m pip install omium

The package is published via PyPI Trusted Publisher. Verify with omium --version.

From source (development)

cd omium-platform/sdk/python
pip install -e ".[dev]"

Quick start (platform)

1. Get an API key from app.omium.ai and configure the CLI:

   omium init --api-key omium_xxx --api-url https://api.omium.ai
   

2. Initialize the SDK in your code (or run with omium run so the CLI injects config):

   import omium
   omium.init(api_key="omium_xxx", project="my-agent")
   

3. Use decorators for tracing and checkpoints:

   from omium import trace, checkpoint
   
   @trace("my_step")
   def my_step(data):
       return process(data)
   
   @checkpoint("important_step")
   async def important_step(data):
       return await do_work(data)
   

4. Run with the CLI to send traces to the platform:

   omium run your_script.py --project my-agent
   

5. Push a project to see it on the Automations dashboard:

   omium project init --name my-agent
   omium project push
   

Checkpoint API (advanced / local)

For direct checkpoint-manager access (e.g. local gRPC), use OmiumClient and the @checkpoint decorator or Checkpoint context manager:

from omium import OmiumClient, checkpoint, Checkpoint

client = OmiumClient(checkpoint_manager_url="localhost:7001")
await client.connect()
client.set_execution_context(execution_id="exec_123", agent_id="agent_1")

@checkpoint("validate_data", preconditions=["data is not None"])
async def validate_data(data: dict) -> dict:
    return {"validated": True, "data": data}

async with Checkpoint("important_state", client=client) as cp:
    result = await do_critical_thing()
    cp.update_state(step="complete")

See the package docstrings and docs.omium.ai for full API details.

Configuration

• API URL: Use https://api.omium.ai for the hosted platform. Override with OMIUM_API_URL or omium configure --api-url <url>. Do not add /api/v1 to the base URL.
• API key: Set via omium init (stored in ~/.omium/config.json) or OMIUM_API_KEY.
• Project config: Use omium.toml in your project root; see docs.omium.ai.

Requirements

• Python >= 3.9 (3.11+ recommended)
• grpcio >= 1.69.0, < 2.0
• protobuf >= 5.26.1, < 6.0

Testing

pip install -e ".[dev]"
pytest tests/
pytest tests/ --cov=omium --cov-report=html

Publishing (maintainers)

Before deploying via GitHub Actions:

1. Bump version in both places (must match or the workflow will fail):
   • pyproject.toml → version = "X.Y.Z"
   • omium/__init__.py → __version__ = "X.Y.Z"
2. Commit and push, then run Actions → Publish Omium Python SDK → Run workflow.
3. The workflow runs tests, builds, and publishes to PyPI (skips if version already exists).

License

Proprietary - Omium Platform