Making the Context Across 46 Repositories Semantically Searchable for AI (Part 2)

Short summary

Ryan, CTO at airCloset, shares how he made 46 repositories semantically searchable for AI by joining three graphs: code-graph (structure), db-graph (database), and annotation-graph (intent). The key insight is annotating only boundary nodes—APIs, pages, events—with @graph-* tags rather than all functions, enabling AI agents to query codebases in natural language without overhead. This pattern has scaled to their db-graph system spanning 1,133 tables and keeps engineer workflows unchanged.

• •Three-graph architecture (code-graph + db-graph + annotation-graph) enables semantic search across 46 repositories
• •Boundary-focused annotation strategy (@graph-* tags) reduces annotation burden while maintaining meaning
• •Pattern proven at scale with 1,133 tables and integrates existing tools like MCP servers for AI agent access

Generated with AI, which can make mistakes.