MultiMind API Reference

Reference documentation for MultiMind’s REST APIs. For the Python API, start with the quickstart and cookbook.

All HTTP servers need the gateway extra (pip install "multimind-sdk[gateway]"); the RAG routes additionally need the rag extra.

REST APIs

• MultiMind API (multimind backend, default http://127.0.0.1:8080, Swagger at /docs) — one FastAPI app built by multimind.server.platform.create_platform_app() that serves every supported route: the gateway (chat, generate, compare, sessions, metrics, compliance), the governance dashboard /api/* routes, the OpenAI-compatible guard proxy (/v1/chat/completions, /v1/embeddings, /v1/guardrails/reload) and, when the rag extra is installed, the RAG routes. Set API_KEYS to require an X-API-Key header. This is the recommended way to run the HTTP API.
• Gateway API — chat, generate, compare, sessions, metrics, and compliance endpoints (multimind.gateway.api)
• RAG API — document management, semantic search, and generation endpoints (multimind.gateway.rag_api; step-by-step server setup)

The guard proxy on its own is started with multimind serve (default port 8400); see cli.md.

Generated OpenAPI specs

Machine-readable specs generated from the FastAPI apps (app.openapi()) live in docs/api/:

• openapi-platform.json — the combined MultiMind API (multimind backend)
• openapi-gateway.json — multimind.gateway.api
• openapi-rag.json — multimind.gateway.rag_api
• openapi-multi-model.json — multimind.api.multi_model_api (/generate, /chat, /embeddings over MultiModelWrapper)
• openapi-unified.json — multimind.api.unified_api (/v1/process multi-modal workflows)

Every running server also serves live interactive Swagger docs at /docs (ReDoc at /redoc, raw spec at /openapi.json) on whatever port it is bound to (8080 for multimind backend, 8000 for the standalone apps).

Python API

• RAG pipeline guide
• Memory systems
• CLI reference
• Configuration