LUML SDK

Python SDK for ML experiment tracking. LUML lets you log metrics, parameters, models, evaluations, and LLM traces as your experiments run — either locally (SQLite) or synced to the LUML cloud platform.

Main use cases:

• Track training runs: log hyperparameters, metrics over time, final models
• Run and score evaluations on LLM or classical ML outputs
• Capture LLM traces automatically via OpenTelemetry instrumentation
• Upload experiments and models to the LUML platform for sharing and collaboration

Installation

pip install luml-sdk

With optional extras:

pip install "luml-sdk[tracing]"          # LLM tracing via OpenTelemetry
pip install "luml-sdk[datasets]"         # pandas / pyarrow support
pip install "luml-sdk[datasets-hf]"      # Hugging Face datasets
pip install "luml-sdk[datasets-polars]"  # polars
pip install "luml-sdk[llm]"              # OpenAI-compatible LLM client
pip install "luml-sdk[flow]"             # expose a running Flow through a LUML relay

Getting your API key

To upload experiments to the LUML platform you need an API key.

1. Go to luml.ai and log in
2. Open Settings → API Keys
3. Click Generate new key and copy it

Set it as an environment variable:

export LUML_API_KEY=your_api_key_here

Documentation

https://docs.luml.ai/sdk/experiments/backends/sqlite