Skip to main content
Prism exposes a set of HTTP API routes under the /api path. These are Next.js Route Handlers that run on the server. All endpoints are co-located with the application — there is no separate API service.
Prism uses Appwrite for authentication. All endpoints rely on a valid Appwrite session. There are no API keys or bearer tokens — the browser sets a session cookie after login, and the server verifies it server-side. Automated clients must replicate the Appwrite session cookie.

​
Base URL

All paths below are relative to your Prism deployment origin. 
https://<your-prism-domain>
For local development the origin is http://localhost:3000.

​
Request format

  • POST requests must send a JSON body with Content-Type: application/json.
  • GET requests pass parameters as URL query strings.

​
Response format

Most endpoints return JSON. The chat endpoint returns a Server-Sent Events (SSE) stream.

​
Errors

All errors follow the same shape: 
{ "error": "Human-readable error message" }
Common HTTP status codes:
StatusMeaning
400Bad request — a required field is missing or invalid
500Internal server error — upstream service (Gemini, Qdrant, Appwrite) failed

​
Rate limits

Prism does not impose its own rate limits. Upstream services (Google Gemini, Qdrant) apply their own quotas.

​
Endpoint groups

Chat

Stream AI responses with RAG context, manage chat sessions, and generate session titles.

Documents

Index, delete, rename, and retrieve document content. Fetch similar-document recommendations.

Search

Run semantic vector searches across your document library.

Analytics

Retrieve UMAP-projected vector graph data for visualising document relationships.

Build docs developers (and LLMs) love

Get started for freeTalk to us