README.md

metadata

title: TTS
emoji: 🤖
colorFrom: blue
colorTo: purple
sdk: docker
app_port: 7860
models:
  - unsloth/Qwen3-8B-GGUF

Qwen3-8B-GGUF API Space

A Hugging Face Space that exposes the unsloth/Qwen3-8B-GGUF model (8-bit) as a REST API using FastAPI and llama-cpp-python.

Endpoints

Method	Path	Description
GET	/	HTML documentation page
POST	/generate	Send a message and receive an AI response
GET	/health	Health check (model status)
GET	/docs	Interactive Swagger UI
GET	/redoc	ReDoc documentation

POST /generate — Request Format

{
  "instructions": "optional system instructions",
  "message": "required user message",
  "user": "optional username"
}

Field	Type	Required	Description
instructions	string	No	System-level instructions
message	string	Yes	The message to process
user	string	No	Username identifier

Prompt Construction

The prompt sent to the model is built as:

{instructions}
{user} said {message}

Example input:

{"instructions": "idk", "message": "thisissupermario", "user": "admin"}

Prompt sent to model:

idk
admin said thisissupermario

Example Usage

Python

import requests

response = requests.post(
    "https://dreamlongyt-agent.hf.space/generate",
    json={
        "instructions": "idk",
        "message": "thisissupermario",
        "user": "admin"
    }
)
print(response.json())

cURL

curl -X POST "https://dreamlongyt-agent.hf.space/generate" \
     -H "Content-Type: application/json" \
     -d \'{"instructions":"idk","message":"thisissupermario","user":"admin"}\'

Deployment

1. Go to huggingface.co/new-space
2. Name your Space and select Docker as the SDK
3. Clone the Space repo and push these files:
   • app.py
   • Dockerfile
   • requirements.txt
   • README.md
4. The Space will build automatically and be available at https://YOUR-USERNAME-YOUR-SPACE-NAME.hf.space

Note: The first request may be slow as the model (~4.7 GB) is downloaded from Hugging Face Hub on startup. The llama-cpp-python library is now installed using pre-compiled wheels for faster setup. Consider upgrading to a paid hardware tier for faster inference.