Assistants API status

Under reviewUpdated 2026-05-26

Meridian does not expose an Assistants API endpoint. Our architecture is purpose-built around stateless chat completions — each request is self-contained, with no server-side thread persistence between calls.

What we ship today

/v1/chat/completions — OpenAI-compatible, stateless
/v1/models — model listing
Streaming via SSE
Tool calling (function calling)

Why no Assistants API?

The Assistants pattern requires persistent threads, managed state, and server-side run orchestration. Meridian keeps the API surface minimal — you own your conversation state. This avoids vendor lock-in and keeps latency predictable.

Roadmap

Persistent threads are on the roadmap for a future release. When they land, they will be opt-in and designed for portability — not a clone of the OpenAI Assistants shape. Subscribe to the changelog for updates.

Recommendation: Build your own thread store client-side or in your backend. Pass the full message array to /v1/chat/completions each request. This gives you full control over truncation, summarization, and storage.