Skip to content

GenAI

AI Hub GenAI API on this instance: Swagger · ReDoc · Live OpenAPI JSON

Genai HTTP reference

API version

GenAI API: 1.4.14

The GenAI sub-application exposes authenticated model generation through an OpenAI-compatible Chat Completions contract. It is mounted at /genai and uses the shared AIHub provider configuration and connectors.

Endpoints

Operation Mounted route Purpose
POST /genai/api/v1/chat/completions Generate a response, request JSON/schema output, use function tools or stream supported responses with SSE
GET /genai/api/v1/catalogue Public shared providers, generation models, capabilities and database profiles; no API key or Client-ID
GET /genai/api/v1/models Discover configured providers, generation models, capabilities and effective profile mappings with provenance and code fallbacks
Swagger /genai/docs Inspect contracts and select concrete request examples
ReDoc /genai/redocs Browse request and response schemas
OpenAPI /genai/openapi.json Retrieve the machine-readable API schema

POST /chat/completions and GET /models require an AIHub X-API-Key and, where applicable, the authenticated client context. GET /catalogue is public and always uses shared provider settings, without client-specific overrides. The local SDK base URL is http://127.0.0.1:8086/genai/api/v1.

Start here

Read the Chat Completions and model-discovery guide for authentication, complete payloads, response formats, errors and SDK usage. It covers:

  • optional provider, model and profile selectors, their precedence and configured defaults;
  • reasoning, JSON/JSON Schema, prompt caching, sampling, token budgets and tools;
  • SSE streaming, usage chunks and assembling JSON output;
  • inline document inputs and extraction constraints;
  • the automatically loaded public catalogue, shared profile provenance and independent code fallbacks, refreshed from MongoDB or PostgreSQL.

The Swagger examples show supported combinations of controls. Use the live catalogue to adapt those examples to the configured models and capabilities of your instance.

Shared configuration and architecture

Provider configuration and semantic profiles are shared with other AIHub applications. See LLM providers, models and tasks for configuration precedence and provider-specific behavior, and Applications and APIs for mounting and ownership. Non-stream generation reuses call_llm_async(); supported OpenAI/Azure streaming retains the incremental SSE transport described in the endpoint guide.