LiteLLM Integration with Callbacks - Helicone OSS LLM Observability

Warning

This integration method is maintained but no longer actively developed. For the best experience and latest features, use our new AI Gateway with unified API access to 100+ models.

LiteLLM is a model I/O library to standardize API calls to Azure, Anthropic, OpenAI, etc. Here's how you can log your LLM API calls to Helicone from LiteLLM using callbacks.

Note

Note: Custom Properties are available in metadata starting with LiteLLM version 1.41.23.

Warning

System Instructions Limitation: When using LiteLLM callbacks, system instructions for Gemini and Claude may not appear as "role": "system" messages in Helicone logs. This is because LiteLLM processes the request before sending it to Helicone.

For full system instruction support, consider using proxy-based integration instead.

1 line integration#

Add HELICONE_API_KEY to your environment variables.

export HELICONE_API_KEY=sk-<your-api-key># You can also set it in your code (See below)

Tell LiteLLM you want to log your data to Helicone

litellm.success_callback=["helicone"]

Complete code#

from litellm import completionimport os## set env variablesos.environ["HELICONE_API_KEY"] = "your-helicone-key"os.environ["OPENAI_API_KEY"], os.environ["COHERE_API_KEY"] = "", ""# set callbackslitellm.success_callback=["helicone"]#openai callresponse = completion(  model="gpt-4o-mini",  messages=[{"role": "user", "content": "Hi 👋 - i'm openai"}],  metadata={    "Helicone-Property-Hello": "World"  })#cohere callresponse = completion(  model="command-r",  messages=[{"role": "user", "content": "Hi 👋 - i'm cohere"}],  metadata={    "Helicone-Property-Hello": "World"  })print(response)

Feel free to check it out and tell us what you think 👋

For proxy-based integration with LiteLLM, see our LiteLLM Proxy Integration guide.