learn

Extract actionable lessons from incident post-mortem transcripts by applying causal-pattern detection to identify signals like 'because' or 'due to', then scoring insights for complexity and impact. This filters noise and surfaces durable patterns for your team's knowledge base.

Causal-pattern extraction for session analysis identifies explicit realizations and causal signals within transcripts to distill high-value insights. It separates durable lessons from scattered observations by matching local patterns and checking novelty against existing knowledge.

You can filter duplicate knowledge from code review transcripts using built-in novelty checks. By querying your knowledge base daemon, the pipeline compares newly extracted patterns against stored lessons to avoid duplicates and promote only new learnings.

The best way to score lesson usefulness from debugging sessions is evaluating insights across novelty, complexity, pattern repeatability, and impact. This threshold-based scoring ensures only high-value insights are stored to your knowledge base.

You do not need a CKS daemon to run the transcript learning pipeline. The process uses local pattern matching for self-contained extraction, while the daemon is optional for novelty checks and cross-referencing when storing lessons to a knowledge base.

Manually store a pre-written lesson from a post-mortem by running the command with a lesson flag, including the text, category, and severity. This bypasses automatic transcript extraction and directly saves the formatted insight to your knowledge base.