RepoGuardian

powered by RHD - Repository Health Director

Autonomous Engineering Intelligence. Evidence-grounded decisions. Human-controlled execution.

Connect a GitHub repository. RHD investigates its issues, pull requests, source code, releases, engineering risks and repository health using agentic RAG, deterministic multi-agent orchestration, specialized model governance and evidence-grounded reasoning.

RHD analyzes automatically. RHD recommends automatically. Humans authorize external actions.

Related MCP server: GithubMCP

Live Application

Public web app: https://repoguardian-rhd.vercel.app

FastAPI: https://repoguardian-rhd-api.vercel.app

API docs: https://repoguardian-rhd-api.vercel.app/docs

Architecture: https://repoguardian-rhd.vercel.app/architecture

RHD v4 Mission Control: https://repoguardian-rhd.vercel.app/mission-control

RHD v5 Chat Workspace: https://repoguardian-rhd.vercel.app

Demo repository: https://github.com/romil569/RepoGuardian-Demo

Public v5.1 flow: paste a GitHub repository into the home chat, run analysis, wait for the persisted job stages to complete, then review the inline SVG architecture artifact and ask follow-up questions in the same repository context.

Architecture

flowchart LR
  G[GitHub] --> W[Webhooks / Sync]
  W --> Q[Event Queue]
  Q --> RI[Repository Intelligence]
  RI --> S[(SQL + Vector + Graph)]
  S --> AR[Agentic RAG]
  AR --> RHD[RHD Supervisor]
  RHD --> A[Specialist Agents]
  A --> ML[ML / DL Intelligence]
  ML --> EC[Evidence Critic]
  EC --> PG[Policy Gate]
  PG --> HR[Human Review]
  HR --> GA[GitHub Action]

Deployment modes:

• LIGHTWEIGHT_LOCAL: SQLite, local vectors, deterministic/RHD tools.

• INDUSTRY_LOCAL: optional Docker PostgreSQL/Redis when Docker exists.

• MANAGED_CLOUD: Vercel frontend, Vercel Python FastAPI backend, Neon PostgreSQL/pgvector, Postgres queue.

• ENTERPRISE_AWS: Terraform foundation; not provisioned.

Feature Matrix

Truthful capability levels:

• Working: production-compatible and covered by regression tests.

• Beta: implemented as an additive v4 path and covered by tests, but depends on synced repository data quality.

• Partial: foundation exists, with explicitly documented constraints.

• Optional: requires local/configured provider; never claimed active in public cloud without configuration.

• Roadmap: documented only, not represented as shipped behavior.

Quick Start

Lightweight Local

cd C:\Users\HP\Desktop\RepoGuardian
copy .env.example .env
copy backend\.env.example backend\.env
.\scripts\start-dev.ps1

Open http://127.0.0.1:3000.

Industry Local

cd C:\Users\HP\Desktop\RepoGuardian
.\scripts\start-industry-local.ps1
.\scripts\doctor.ps1

Docker PostgreSQL/Redis are used only when Docker is installed. Otherwise the stable lightweight path remains available.

Managed Cloud

Backend Vercel entrypoint: api/index.py.

Required backend environment:

• DEPLOYMENT_MODE=MANAGED_CLOUD

• DATABASE_URL=postgresql://...

• POSTGRES_RUNTIME_MODE=managed

• QUEUE_BACKEND=postgres

• PUBLIC_ANALYSIS_MODE=true

• GITHUB_WRITE_MODE=disabled

• FRONTEND_URL=https://...

• CORS_ORIGINS=https://...

• ENABLE_STARTUP_SCHEMA_CREATE=false

Frontend environment:

NEXT_PUBLIC_API_URL=https://your-fastapi-service.example.com

See docs/deployment-managed-cloud.md and docs/vercel-backend-audit.md.

MCP

cd mcp-server
npm install
$env:REPOGUARDIAN_API_URL="http://127.0.0.1:8000"
npm start

MCP exposes RHD tools, resources, and prompts. Write-gated actions remain human/policy gated. See docs/mcp.md.

Testing

cd backend
.\.venv\Scripts\python -m pytest

cd frontend
npm run lint
npm run typecheck
npm run build
npm run e2e

cd mcp-server
npm run typecheck
npm test

Safety

• Repository writes are allow-listed and require human approval.

• Public repositories are analyzed in read-only mode unless policy explicitly allows writes.

• Private repositories default to local/deterministic processing.

• Issue text, comments, README files, code, and PR descriptions are treated as untrusted evidence.

• Evidence must correspond to synchronized repository records.

• Frontend NEXT_PUBLIC_ variables never contain backend secrets.

• Audit logs store safe summaries, not secrets or private reasoning.

Current Limits

• Managed PostgreSQL/pgvector is ready for credentials but not connected in this local run.

• Redis is optional and not connected locally.

• Docker remains optional and unavailable on the current machine.

• qwen3:1.7b Ollama was validated locally; qwen3:8b was pulled but too slow for demo.

• ML/DL training is not claimed without a defensible labeled dataset.

• Production is not marked validated until deployed infrastructure and live production testing exist.

Documentation

• API

• MCP

• Managed Cloud Deployment

• Provider Benchmark

• Industry Readiness

• Public Source Audit

• RHD v4 Architecture

• RHD v4 Demo Script

• RHD v5 Chat Workspace