Repository Ingestion Engine

Why This Matters

Engineering teams spend weeks maintaining service catalogs in Backstage or Confluence. The docs go stale the moment they're written. StackSight's ingestion engine reads the code directly — NuGet packages, cloud service references, internal API calls — and builds an always-current dependency map.

What You'll Get

• Connect any Git repo (Azure DevOps, GitHub, Bitbucket)
• Automatic extraction of dependencies from project files, imports, and configuration
• Living dependency graph updated on every push
• Confidence scoring — auto-detected facts vs developer-confirmed knowledge

How It Works

A webhook fires on push. The ingestion service clones the relevant changes, runs deterministic parsers for known file types (.csproj, package.json, Dockerfile), then uses an LLM to extract semantic understanding from code patterns. Results are written to the Cosmos DB knowledge graph with confidence levels.