agent-debugger-cli

You can analyze agent execution traces by running a quality check or posing QA questions directly against the trace file to identify issues and their causes. The CLI automates downloading, cleaning, and evaluating these traces.

You can download and clean Langfuse traces for debugging using the agent debugger CLI, which retrieves execution data and prepares it for analysis. Optional network access is required to fetch the traces.

Yes, you need Python 3.9+ and pip to run quality checks on agent traces using this CLI. Local shell access is also required, and you must configure python-dotenv for environment variables.

The best way to configure LLM settings for agent debugging is through the CLI's configuration commands, which manage language model setups for trace analysis. This ensures the debugger uses the correct model parameters.

Yes, you can ask specific questions about a trace file to debug your agent by using the ask command with the trace and query parameters. This returns targeted answers regarding issues within the execution data.

Trace analysis fails without network access if the CLI needs to download Langfuse traces or make model calls. While local trace files can be analyzed offline, retrieving remote data requires an active connection.