
DriftGuard solves a DevOps pain point: stale documentation. When developers rename functions, change API endpoints, or modify parameters, the README and docs often stay outdated. This causes onboarding delays, integration bugs, and compliance failures. DriftGuard automates the detection process. Users simply paste a public GitHub repository URL into the dashboard. The backend auto-detects the latest commit or pull request, fetches the code diff and all documentation files via the GitHub API, then sends them through a LangGraph multi-step pipeline powered by Groq's LLM. The AI cross-references every code change against the documentation and identifies exactly which sections are now stale. Each drift is classified by severity (Critical, High, Medium, Low) with specific suggested fixes and line numbers. A documentation health score from 0 to 100 summarizes the overall state. The tech stack includes FastAPI for the async backend, LangGraph for pipeline orchestration, LangChain with ChatGroq for structured LLM output using Pydantic models, and a React + Vite frontend with a premium dark-mode glassmorphism UI. The entire project was built using IBM Bob IDE. Three scan modes are supported: Quick Scan (auto-detect latest commit), PR Scan (analyze any pull request), and Manual (compare any two branches or commits).
17 May 2026