gaise-api

Axum HTTP server for GAISe — exposes all GenAI providers behind a unified REST API with SSE streaming.

Endpoints

Method	Path	Description
`POST`	`/v1/instruct`	Non-streaming instruct request
`POST`	`/v1/instruct/stream`	Server-Sent Events streaming
`POST`	`/v1/embeddings`	Generate embedding vectors

Quick Start

# Set provider keys
export OPENAI_API_KEY="sk-..."
export ANTHROPIC_API_KEY="sk-ant-..."
export GEMINI_API_KEY="AIza..."

# Run the server
cargo run -p gaise-api
# Listening on 0.0.0.0:3000

# Call any provider via the same endpoint
curl -X POST http://localhost:3000/v1/instruct \
  -H "Content-Type: application/json" \
  -d '{
    "model": "openai::gpt-4o",
    "input": {
      "role": "user",
      "content": {"type": "text", "text": "Hello!"}
    }
  }'

Change "model" to "anthropic::claude-sonnet-4-6" or "gemini::gemini-2.5-flash" — same endpoint, same format.

Environment Variables

Variable	Description
`GAISE_PORT`	Server port (default: `3000`)
`OLLAMA_URL`	Ollama API URL
`OPENAI_API_URL` / `OPENAI_API_KEY`	OpenAI credentials
`ANTHROPIC_API_URL` / `ANTHROPIC_API_KEY`	Anthropic credentials
`GEMINI_API_URL` / `GEMINI_API_KEY`	Gemini credentials
`VERTEXAI_API_URL` / `VERTEXAI_SA_PATH`	Vertex AI credentials
`BEDROCK_REGION`	AWS Bedrock region

As a Library

use std::sync::Arc;
use gaise_api::{create_app, AppState};
use gaise_client::{GaiseClientService, GaiseClientConfig};

let config = GaiseClientConfig { /* ... */ ..Default::default() };
let state = Arc::new(AppState {
    client_service: GaiseClientService::new(config),
});

let app = create_app(state);
// Mount into your own Axum server

Part of GAISe