-
Notifications
You must be signed in to change notification settings - Fork 0
Home
Guruprasath Annadurai edited this page Jul 3, 2026
·
2 revisions
The Trust Layer for AI — Enterprise Governance Platform
ResponsibleAI is an open-source AI governance platform that provides trust scoring, bias detection, guardrails, hallucination detection, compliance auditing, cost intelligence, drift monitoring, and an MCP server for Claude Code integration.
| Section | Description |
|---|---|
| Architecture | System design, module map, data flow |
| Getting Started | Install, first run, 5-minute quickstart |
| API Reference | All REST endpoints with request/response examples |
| MCP Server | Claude Code integration — tools, resources, setup |
| Configuration | All environment variables and their effects |
| Database & Migrations | SQLite, PostgreSQL, Alembic usage |
| Authentication & RBAC | Bearer tokens, roles, org management |
| Webhooks | Event delivery, HMAC signing, retry persistence |
| BiasBuster | Bias probes, scoring methodology, CI gate |
| PrivacyLabel | Federated learning, differential privacy |
| Compliance Frameworks | NIST AI RMF, EU AI Act, ISO 42001 |
| Deployment | Docker, Kubernetes, Helm, PostgreSQL+Redis |
| Contributing | Dev setup, test suite, code style |
| Roadmap | What's shipped, what's next |
┌─────────────────────────────────────────────────────────────────┐
│ ResponsibleAI v1.1.0 │
│ │
│ Claude Code ──► MCP Server (10 tools, 5 resources) │
│ │ │
│ REST Client ─────────┤ │
│ ▼ │
│ ┌──────────────────────────────────────────────────────────┐ │
│ │ Governance Dashboard (FastAPI) │ │
│ │ Auth · Per-org rate limit · Audit log · OTEL · WS │ │
│ └──────────────────────────────────────────────────────────┘ │
│ │ │ │ │ │ │
│ Trust Compliance Guardrails Cost Red Team │
│ Score NIST/EU/ISO PII+Tox Tracking Simulator │
│ │ │ │ │ │ │
│ ┌──────────────────────────────────────────────────────────┐ │
│ │ SQLAlchemy (SQLite / PostgreSQL) │ │
│ │ Alembic migrations · WAL mode · async │ │
│ └──────────────────────────────────────────────────────────┘ │
└─────────────────────────────────────────────────────────────────┘
src/
├── responsibleai/ # Core governance engine + dashboard
│ ├── dashboard/ # FastAPI app, routes, WebSocket, static SPA
│ ├── db/ # Async SQLAlchemy engine, repositories
│ ├── trust/ # TrustScoreEngine (6-dim, A–F grade)
│ ├── compliance/ # NIST AI RMF, EU AI Act, ISO 42001
│ ├── guardrails/ # PII detection, toxicity filtering
│ ├── hallucination/ # TF-IDF self-consistency + hedging analysis
│ ├── redteam/ # 10 adversarial attack vectors
│ ├── cost/ # Token tracking, model routing, budget alerts
│ ├── mcp/ # MCP server (stdio, 10 tools, 5 resources)
│ ├── webhooks/ # HMAC-signed event delivery + DB retry queue
│ ├── rbac/ # OWNER/ADMIN/ANALYST/VIEWER roles
│ └── auth/ # Bearer token + OIDC/JWKS validation
├── biasbuster/ # Bias evaluation CLI + 6 demographic probes
└── privacylabel/ # Federated learning + differential privacy
migrations/ # Alembic versioned schema
tests/ # 942 tests