Intelligence & Knowledge

What it is

The per-project understanding layer. It profiles a codebase, ingests and queries a knowledge corpus, runs analytical "prisms," and proposes governance, scaffolding, and onboarding.

Operator surface

• Profiling & discovery — intelligence_profile, intelligence_profile_build, intelligence_discover, intelligence_recommendations, intelligence_briefing_generate.
• Knowledge corpus — intelligence_knowledge_ingest, intelligence_knowledge_query, intelligence_knowledge_sources, intelligence_knowledge_stats, plus discovery / suggestion / approval (intelligence_knowledge_discover, intelligence_knowledge_suggestions, intelligence_knowledge_approve).
• Code index / search — intelligence_code_index, intelligence_code_search.
• Prisms (analytical passes) — intelligence_prism_run, intelligence_prisms_list, intelligence_prisms_run_all, intelligence_prism_findings, intelligence_prism_report.
• Governance rules — intelligence_governance_list / _add / _update.
• Scaffolding — intelligence_scaffolding_generate / _approve / _execute / _status; plus alerts and maintenance config.
• Aspects (tracked dimensions of project health) — aspect_summary_get / aspect_summary_refresh, aspect_open_questions, aspect_investigate, aspect_resolve_question.

REST is IntelligenceController (+ CooccurrenceController, SuggestionsController).

How it works

Knowledge sources and project profiles persist per tenant (KnowledgeSource, ProjectProfile, CodebaseSnapshot, IntelligenceBriefing). Code is chunked and embedded for semantic search (EmbeddingChunk, an all-MiniLM-L6-v2 model ships in-process), file relationships are mined into FileCooccurrence, and prisms emit PrismReport → PrismFinding. Aspects are tracked dimensions of project health, each with a refreshable summary snapshot (AspectSnapshot) and open questions agents can investigate and resolve.