API Reference | PrivateGPT

PrivateGPT exposes a REST API that follows the Claude API model. All endpoints are available at http://localhost:8080 by default.

The interactive reference below is generated from the OpenAPI schema. Use it to explore request bodies, response shapes, and try requests directly.

Endpoint groups

Group	Endpoints	Description
Messages	`POST /v1/messages` and variants	Chat, streaming, token counting, validation, async
Models	`GET /v1/models`	List and inspect available models
Artifacts	`/v1/artifacts/*`	Ingest, list, retrieve, and delete documents
Embeddings	`POST /v1/embeddings`	Generate text embeddings
Tools	`/v1/tools/*`	Semantic search, web search, web fetch, database query, tabular analysis
Primitives	`POST /v1/primitives/search`	Low-level chunk retrieval
Skills	`/v1/skills/*`	Create and manage reusable instruction sets

When auth is enabled, pass a bearer token on every request:

curl http://localhost:8080/v1/messages \
  -H "Authorization: Bearer <your-token>" \
  -H "Content-Type: application/json" \
  -d '{...}'

Auth is disabled by default. Enable it in settings.yaml:

server:
  auth:
    enabled: true
    secret: "Basic <base64-encoded-user:pass>"