README.md

---
title: Respite API
emoji: 🤖
colorFrom: purple
colorTo: blue
sdk: gradio
sdk_version: 6.3.0
app_file: app.py
pinned: false
---

# Respite API

A Gradio/ZeroGPU Hugging Face Space for the Respite API, with audio generation, web search, and CPU-hosted TinyLlama inference.

## Features

- **Audio model**: `small-music` (433M parameters)
- **Text model**: TinyLlama 1.1B, explicitly loaded on CPU
- **Duration**: 1–120 seconds
- **Sample rate**: 44.1 kHz stereo
- Gradio 6.3.0 interface

## API discovery endpoints

Respite API exposes machine-readable API metadata alongside the Gradio UI:

- `GET /respite/resources` lists callable capabilities, including each resource's method, endpoint, inputs, and output.
- `GET /respite/specs` lists server-wide specifications such as version, supported content types, limits, discovery URLs, CPU cores, total RAM, and storage total/used/free space.

Both endpoints return JSON and currently require no authentication. The metadata is designed to grow as additional AI capabilities are added.

## Usage

1. Enter a text prompt describing the music you want
2. Adjust duration, steps, CFG scale, and seed
3. Click **Generate** and wait for the audio

## Setup

```bash
pip install -r requirements.txt
python app.py
```

## License

This project uses the Stable Audio 3 model. Please refer to Stability AI's [license](https://huggingface.co/stabilityai/stable-audio-3-small) for usage terms.
# rebuild trigger 1790687896
# rebuild probe 1703