crisai-diagnostics-and-tooling

To analyze crisAI run costs, execute the 'crisai spend' command, which reviews logs, run histories, and usage data to output a detailed cost analysis for your latest execution.

Run interpretation analyzes crisAI execution outcomes by evaluating routing decisions, contract enforcement, and checkpoints, providing detailed insights into how the system behaves during a specific run.

Yes, you need Python 3.10 or higher and a uv environment to run crisAI diagnostics, as the behavior measurement and directory cleanup tools require this specific setup.

You can clean up stray directories at the repository root using the directory cleanup tools included in the crisai-diagnostics toolkit, ensuring a clean environment for future runs.

Yes, you can verify mode parity between pipeline and peer execution modes using the mode parity check feature, which confirms that different execution modes maintain consistent behavior.

Unexpected routing decisions in crisAI run histories can be diagnosed using the run interpretation tools, which analyze logs and checkpoints to explain the system's behavior during execution.