README.md

---
title: SmolLM2 Backend
emoji: 📊
colorFrom: yellow
colorTo: red
sdk: docker
pinned: false
license: apache-2.0
short_description: Backend of SmolLM2 chat
app_port: 7860
---

# SmolLM2 Backend

This project implements a FastAPI API that uses LangChain and LangGraph to generate text with the Qwen2.5-72B-Instruct model from HuggingFace.

## Configuration

### In HuggingFace Spaces

This project is designed to run in HuggingFace Spaces. To configure it:

1. Create a new Space in HuggingFace with SDK Docker
2. Configure the `HUGGINGFACE_TOKEN` or `HF_TOKEN` environment variable in the Space configuration:
   - Go to the "Settings" tab of your Space
   - Scroll down to the "Repository secrets" section
   - Add a new variable with the name `HUGGINGFACE_TOKEN` and your token as the value
   - Save the changes

### Local development

For local development:

1. Clone this repository
2. Create a `.env` file in the project root with your HuggingFace token:
   ```
   HUGGINGFACE_TOKEN=your_token_here
   ```
3. Install the dependencies:
   ```
   pip install -r requirements.txt
   ```

## Local execution

```bash
uvicorn app:app --reload
```

The API will be available at `http://localhost:7860`.

## Endpoints

### GET `/`

Welcome endpoint that returns a greeting message.

### POST `/generate`

Endpoint to generate text using the language model.

**Request parameters:**
```json
{
  "query": "Your question here",
  "thread_id": "optional_thread_identifier"
}
```

**Response:**
```json
{
  "generated_text": "Generated text by the model",
  "thread_id": "thread identifier"
}
```

## Docker

To run the application in a Docker container:

```bash
# Build the image
docker build -t smollm2-backend .

# Run the container
docker run -p 7860:7860 --env-file .env smollm2-backend
```

## API documentation

The interactive API documentation is available at:
- Swagger UI: `http://localhost:7860/docs`
- ReDoc: `http://localhost:7860/redoc`