crystallize-workflow

A crystallized knowledge layer caches proven QA answers to serve repeated questions directly, avoiding redundant RAG runs. It maintains frontmatter metadata, versioning, and atomic writes to ensure stable answers are delivered quickly.

You cache proven RAG answers by triggering crystallization after a successful response, which writes the outcome to a crystallized skill layer. This enforces frontmatter metadata and freshness TTLs for future similar questions.

Use crystallized content when QA answers are stable over time and similar questions are frequently repeated. It bypasses redundant RAG generation by serving cached answers, refreshing them only when their freshness TTL expires.

The crystallization workflow handles stale answers using freshness TTLs and selective refresh triggers. It enforces versioning and atomic writes to index.json during updates, ensuring data integrity when refreshing outdated crystallized skills.

Yes, the crystallization workflow supports automatic integration with existing QA workflows. After your QA workflow generates a final answer, it triggers crystallization to cache the outcome and bypass redundant RAG for future queries.

Creating crystallized skills enforces frontmatter metadata, versioning, and freshness TTLs. It uses atomic writes to save data to a crystallized directory and update an index.json file, preventing data corruption during creation or updates.