decision-graph-analyzer

Query the decision graph using semantic search to find past deliberations similar to your current question. The skill retrieves relevant historical decisions from a SQLite backend with adaptive k-nearest neighbor retrieval, two-tier caching, and noise filtering to surface the most applicable context.

Yes. Semantic search analyzes the decision graph to locate deliberations semantically similar to new queries across all stored projects. This prevents redundant debates and injects relevant historical consensus into current decisions.

The decision graph uses a SQLite backend with three core tables: decision_nodes storing deliberations, participant_stances capturing positions, and decision_similarities indexing semantic relationships. This structure enables efficient retrieval and maintenance of decision history.

Use context injection debugging endpoints to inspect decision graph statistics, validate database health, and analyze retrieval performance. The skill identifies whether context retrieval is failing, caching is inefficient, or injected context simply lacks relevance to current deliberations.

Monitor and optimize the two-tier caching system for fast historical retrieval, inspect database growth statistics, estimate archival benefits, and run health checks. These maintenance endpoints surface performance bottlenecks and inform optimization decisions.

The skill requires a populated decision graph to function meaningfully. You need prior deliberations stored in the SQLite backend to retrieve context; new projects with no history will yield limited semantic search results until decisions accumulate.