verifywise-ai/verifywise

VerifyWise – Open‑source AI Governance Platform

What it is – VerifyWise is a source‑available web application that helps organisations manage the risks and compliance requirements of using AI models. It bundles a set of tools for tracking AI assets, evaluating large language models (LLMs), documenting policies, and generating audit‑ready reports. The platform can be run on‑premises or in a private cloud, giving companies full control over their data and governance processes.

Key capabilities

• Compliance frameworks – Built‑in support for the EU AI Act, ISO 42001, NIST AI RMF, ISO 27001 and 21 other standards (SOC 2, GDPR, HIPAA, etc.).
• Dashboard & risk views – Executive and operational dashboards show AI use‑case risks, vendor risk, model inventory, and incident management.
• LLM evaluation suite – An "EvalServer" service runs LLM benchmarks and an "LLM Arena" for side‑by‑side comparisons.
• AI‑powered advisor – Chat‑based assistant that suggests governance actions and answers compliance questions.
• AI detection – Scans code repositories to flag AI‑generated content and “shadow AI” that may be used without oversight.
• Policy & workflow engine – Create policy templates, map controls to standards, and run approval workflows with email or Slack notifications.
• Integrations – Connectors for Slack, Jira, Azure AI Foundry, MLflow, and import of risk data from MIT/IBM repositories.
• Automation & webhooks – Trigger actions when entities change, schedule reports, or push data to external systems.

Technology stack

• Front‑end: React.js (served on port 5173 in dev).
• Back‑end: Node.js API (port 3000) with PostgreSQL for persistence and Redis for caching.
• Evaluation service: Python 3.12 + FastAPI‑style server (EvalServer) handling LLM benchmarks.
• Deployment: Docker & Docker‑Compose, with optional Kubernetes manifests; can be run on‑prem, private cloud, or services like Render.com.
• Authentication: Local accounts, Google OAuth2, Entra ID, RBAC roles.

How to get started

1. Clone the repo and install front‑end (npm install in Clients) and back‑end (npm install in Servers).
2. Set up a PostgreSQL database and Redis container.
3. Copy .env.dev → .env, set a strong JWT_SECRET, and define a super‑admin email/password.
4. (Optional) Start the Python EvalServer for LLM evaluation features.
5. Run the back‑end (npm run watch) and front‑end (npm run dev) for development, or use the provided install.sh/Docker‑Compose script for production.
6. Configure SSL via Nginx and Certbot if exposing the service publicly.

Who it’s for – Enterprises, regulated industries, or any team that needs a centralized place to inventory AI models, assess their risks, and demonstrate compliance to auditors. Because the code is source‑available under the Business Source License 1.1 (with a commercial dual‑license option), organisations can audit the implementation or extend it to meet internal policies.

Where to learn more – Official site verifywise.ai, user guide, Discord community, and the repository’s README contain detailed docs, screenshots, and deployment scripts.