What problem does it solve? Valuable engineering knowledge—design decisions, business rules, coding conventions, and bug patterns—stays buried in code and commit history, forcing teams to rediscover it repeatedly. ## Core Features & Use Cases - Five LLM-Powered Extractors: pattern (architecture), decision (lightweight ADRs), rule (business rules), convention (coding standards), and bug-pattern (defect patterns from fix commits). - OpenSpec Integration: Run extraction from an archived OpenSpec change with --from-change to close the knowledge feedback loop. - Quality Guardrails: Content-hash caching avoids redundant LLM calls, low-confidence entries (< 0.7) are flagged for review, and supersede links keep knowledge current. - Use Case: After archiving an OpenSpec change like add-refund-fee, run extraction to automatically capture the new business rules and design decisions into the knowledge base. ## Quick Start Ask the AI to run codekb extract on the current project to generate structured knowledge entries from the codebase and git history.