Description
SDK
Python SDK
Description
Users have reported confusion regarding the clarity of our documentation for LLM Monitoring, specifically about what works out of the box versus what requires manual setup or decorators.
-
For the OpenAI Python library, it is unclear if monitoring is automatically enabled upon installation or if manual instrumentation using @ai_track is required.
-
For LangChain, users expected automatic tracking of LLM calls without explicitly initializing Sentry with the LangChain integration or using decorators. However, LLM monitoring only activates once explicitly initialized and traces only appear when decorators are used.
Suggested Solution
-
Explicitly state for the AI integrations (e.g., OpenAI, LangChain) support automatic instrumentation, and what auto instrumentation means in that context. Just errors? Tracing? LLM specific insights?
-
Provide clear examples for each integration scenario highlighting necessary setup steps.