Installation¶

Requirements¶

• Python 3.10+
• pip or uv

1. Install¶

# pip
pip install octuner

# or uv (recommended)
uv add octuner

2. Configure provider credentials¶

Set the API keys for the providers you plan to use:

export OPENAI_API_KEY="sk-your-openai-key"
export GOOGLE_API_KEY="your-google-api-key"

3. Create an explicit YAML config¶

Octuner is library-first and uses explicit configs only (no global defaults). Copy a starter template and keep it in your repo:

mkdir -p configs
cp config_templates/multi_provider.yaml configs/llm.yaml   # or openai_basic.yaml / gemini_basic.yaml

Edit configs/llm.yaml to include the providers/models you want enabled.

4. Use in your application¶

See Adapt existing code in Getting Started for concise before/after examples and a service wrapper pattern.

• Getting Started → Adapt existing code
• Then return to Optimization to tune parameters

Verify¶

import octuner
print("Octuner version:", octuner.__version__)

from octuner import MultiProviderTunableLLM
_ = MultiProviderTunableLLM(config_file="configs/llm.yaml")
print("Config loaded OK")

Next steps¶

• Read the short Getting Started guide
• Explore the API Reference for detailed documentation
• Learn how to contribute to Octuner

Note: Always pass your YAML config_file explicitly when constructing MultiProviderTunableLLM.