retrospective

To extract learnings from a session transcript, analyze the text to identify corrections, reversals, and successes, then write immediate learned rules to the local rules directory for rapid AI behavior improvement.

Session analysis is the process of aggregating transcript learnings across current or past sessions to detect recurring issues, generate metrics, and propose rules that systematically improve AI behavior.

Yes, you can automatically run a retrospective on session start for the previous session, or manually invoke it to analyze the current session, review accumulated learnings, or target specific past sessions.

Generate rules from accumulated learnings by detecting recurring patterns in session transcripts and writing those immediate learned rules directly into the local rules directory to adjust behavior.

Extracted session learnings and proposed rules are stored locally in the learnings directory, while immediate learned rules are written directly to the rules directory for rapid application.