knowledge-agent

Semantic search over a filtered observation corpus works by priming a conversational session with your selected knowledge base. The session then retrieves and answers questions using the primed knowledge, ensuring responses are grounded in your historical work patterns rather than general data.

The best way to answer questions using a filtered set of past observations is to define a corpus with specific filter parameters, prime a session from that corpus, and then ask questions conversationally. This approach grounds answers in your historical work patterns for a chosen topic or time window.

Yes, you can filter your observation history by project, types, concepts, files, semantic query, and date ranges. This filtering compiles focused expertise corpora, allowing you to create a compact knowledge base tailored to a specific topic, project, or time window for conversational Q&A.

To keep your primed knowledge base current when new observations are added, you use the supported corpus listing, rebuild, and reprime workflows. These workflows allow you to refresh the filtered observation corpus and re-prime your conversational session with the updated knowledge.

Yes, this question answering approach requires defining a corpus with filter parameters before priming a session. You must compile the knowledge base from prior observation history first, which then gets loaded into a persistent AI session to ground your conversational queries.