Prerequisites
- A SambaCloud account and API key
- Python 3.10 or later installed
Setup
1
Install Oumi
2
Export your API key
Usage
Create aSambanovaInferenceEngine and run inference:
For full usage — including how to pass conversations and retrieve responses — see the Oumi SambaNova inference engine API reference.
Troubleshooting
ModuleNotFoundError: oumi
ModuleNotFoundError: oumi
- Run
pip install oumiinside your active virtual environment - Confirm the environment is active:
which pythonshould point inside.venv
Authentication error or missing API key
Authentication error or missing API key
- Confirm the key is exported:
echo $SAMBANOVA_API_KEY - Re-export with
export SAMBANOVA_API_KEY="your-key"and retry RemoteParams(api_key_env_varname="SAMBANOVA_API_KEY")reads the variable at runtime — it must be set in the same shell session
Model not found or invalid model name
Model not found or invalid model name
- Verify the model name on the SambaCloud models page
- Model names are case-sensitive — use the exact string shown in the portal
Python version error on install
Python version error on install
- Oumi requires Python 3.10 or later — confirm with
python --version - If using an older version, create a new virtual environment with Python 3.10+
Resources
Oumi API reference
Full SambanovaInferenceEngine API — conversations, responses, and config options.
Oumi documentation
Official Oumi guides covering training, evaluation, and inference.
Supported models
Current list of models available on SambaCloud.
SambaCloud portal
Generate and manage your SambaCloud API keys.