spectra-ask

Ground Q&A responses by querying openspec documents with vector search to locate authoritative specs and proposals. The assistant answers strictly from document content, ensuring responses are derived exclusively from the knowledge base rather than general knowledge.

When documentation doesn't contain the answer, the grounding mechanism responds that the content is not present rather than speculating. This ensures answers never rely on training data or general knowledge outside the openspec knowledge base.

Archived spec proposals provide historical context about how ideas evolved. Vector search prioritizes current specs over archives for accuracy, using archive documents only when historical context about the spec is needed.

Vector search uses embedding-based retrieval to locate the most relevant main spec and related proposals within openspec documents. It prioritizes current specs over archives to ensure accurate, document-backed answers.

Spectra-ask operates without external dependencies, querying the openspec knowledge base directly. It uses vector search to identify relevant specs and proposals, producing grounded answers exclusively from document content.

Spec Q&A answers include inaccurate general knowledge when they are not grounded in openspec documents. Using vector search to retrieve authoritative specs ensures answers are derived strictly from document content, preventing speculation.