Cognipeer/console

git clone https://github.com/Cognipeer/cognipeer-console.git
npm install
cp .env.example .env.local && npm run dev

Usage

• •Requires Node.js 20+; the gateway starts with SQLite by default — no external database required.
• •Set JWT_SECRET (required); dashboard at http://localhost:3000.
• •Call the OpenAI-compatible API at /api/client/v1/chat/completions and /api/client/v1/embeddings with a Bearer API token.
• •Docker: docker compose up -d or docker run -p 3000:3000 -v ./data:/app/data -e JWT_SECRET=... cognipeer-console.

Key features

• ✓Multi-tenant architecture with complete data isolation via per-tenant databases.
• ✓LLM gateway with OpenAI-compatible chat completions and embeddings across OpenAI, Anthropic, AWS Bedrock, Google Vertex AI, vLLM, Ollama, and more.
• ✓Vector store management (multi-provider, built-in SQLite vectors), Knowledge Engine ingestion/chunking/embedding/retrieval, and semantic memory.
• ✓Agent tracing, guardrails (PII detection, moderation, prompt shields, LLM evaluators), prompt management with versioned env-based deployment, inference monitoring, alerts, and quota/rate limiting.
• ✓Next.js 15 + TypeScript with Mantine v8 and Tailwind; SQLite or MongoDB backend; official TypeScript SDK in console-sdk.

Best for

Teams that want a self-hosted, OpenAI-compatible multi-tenant AI gateway with guardrails, tracing, prompt management, and vector/knowledge tooling.

Caveats

• ⚠AGPL-3.0 (community edition); embedding in a closed-source product or offering a proprietary hosted derivative requires a commercial license.
• ⚠JWT_SECRET is required and the rate-limit backend defaults to MongoDB (use mongodb, memory, or redis).
• ⚠BROWSER_BLOCK_PRIVATE_NETWORK defaults to true, blocking private-network egress from managed browser sessions.