reflect:corpus

You create a filtered knowledge base by running a script that applies tags, categories, and project IDs to your learnings. This generates a long-lived slice of code history designed specifically for sustained and repeated Q&A sessions.

A long-lived corpus is a persistent, filtered slice of your learnings that enables durable Q&A sessions against a specific area of code history. It allows you to repeatedly query specific segments without losing conversational context.

Yes, you can filter your knowledge base using specific tags, categories, and project IDs. The Skill accepts a filter string during corpus creation to ensure the resulting Q&A context targets a specific code history segment.

You need a sustained Q&A session against code history when you must ask the same set of questions repeatedly, such as continuously querying an auth subsystem or retrieving everything tagged with a specific label.

The best way to query everything tagged with a specific label is to build a dedicated corpus using a filter parameter. This creates a durable knowledge base slice that can be queried repeatedly across conversational interactions.