arize-instrumentation

Automating LLM observability involves adding tracing to your application to capture LLM calls, tool usage, and agent loop execution. By implementing manual CHAIN and TOOL spans, you get comprehensive trace data that helps understand the performance and behavior of your LLM application.

Yes, you can instrument tracing for LangChain applications using OpenAI providers. The instrumentation process detects your specific frameworks and LLM providers, then automatically implements the necessary tracing to show LLM calls, tool usage, and overall chain execution in your observability platform.

Yes, Arize AX instrumentation supports TypeScript, JavaScript, and Java stacks, alongside Python. It detects your application's dependencies and implements the necessary tracing logic for various LLM providers and agent frameworks across these languages.

The best way to trace tool and function calling is by manually adding CHAIN and TOOL spans to your application. This instrumentation captures detailed execution data for tool usage and agent loops, providing comprehensive observability of your tool-calling workflows.

Yes, you need the OpenTelemetry SDK and Arize-specific packages to export traces to the Arize platform. These dependencies are required to properly format and transmit your LLM observability data from your application to the platform.