​
LiteLLM Integration

Lucidic integrates with LiteLLM via a custom callback bridge that maps LiteLLM events into Lucidic sessions/steps/events. This enables automatic tracking for any provider supported by LiteLLM.

​
Setup

import lucidicai as lai

# Initialize with the LiteLLM provider enabled
lai.init(session_name="LiteLLM Demo", providers=["litellm"])  

# Configure LiteLLM (example: OpenAI-compatible model)
import litellm
litellm.set_verbose = False

response = litellm.completion(
    model="gpt-4o-mini",
    messages=[{"role": "user", "content": "Say 'hello from litellm'"}]
)

# Event is auto-created. Close the session when done.
lai.end_session()

​
How it works

  • Lucidic registers a LiteLLM CustomLogger (LucidicLiteLLMCallback) under the hood
  • Before-call: captures messages to form the event description
  • After-call: extracts response text, usage/cost (when available), duration, and model/provider
  • Images in multimodal messages are detected and attached as screenshots
  • On failure: an error event is recorded with error metadata

​
Notes

  • Ensure your provider-specific API keys are configured per LiteLLM’s docs
  • The bridge waits briefly on shutdown for pending callbacks to finish
  • Cost is estimated using Lucidic’s pricing tables when usage is provided

​
See Also