---
title: Technical Description Assistant
emoji: 📝
colorFrom: blue
colorTo: purple
sdk: gradio
sdk_version: 4.29.0
app_file: app.py
pinned: false
license: mit
short_description: rewrite rough project descriptions into technical summaries
---

# Technical Description Assistant

A zero-shot Hugging Face Space that uses LLM to help developers instantly rewrite
rough project descriptions into clean, professional technical summaries.

## ⚠️ Required: Hugging Face Token

This Space **requires** a Hugging Face access token to function. The application uses only LLM-based processing with no fallback mechanisms.

1.  Go to your Hugging Face profile and create a new access token with **`read`** permissions: [**huggingface.co/settings/tokens**](https://huggingface.co/settings/tokens)
2.  Go to your Space settings: [**huggingface.co/spaces/Snaseem2026/technical-description-assistant/settings**](https://huggingface.co/spaces/Snaseem2026/technical-description-assistant/settings)
3.  Scroll down to **"Secrets and variables"** and click **"New secret"**.
4.  For the **Name**, enter `HUGGING_FACE_HUB_TOKEN`.
5.  For the **Value**, paste your `hf_...` access token.
6.  Click **"Save secret"** and restart the Space from the settings menu.

## How It Works


This application uses the Hugging Face OpenAI-compatible API (`https://router.huggingface.co/v1/chat/completions`) to call a pre-trained large language model (`meta-llama/Llama-3.2-3B-Instruct`). It requires no local model download and no rule-based fallbacks, relying entirely on the LLM for professional rewriting.

1.  **Input**: A developer enters a casual or rough description of their project.
2.  **Processing**: The description is sent to the LLM with a system prompt instructing it to act as an expert technical writer. The output is always concise and limited to a maximum of 3 sentences.
3.  **Output**: The model returns a polished, professional, and technically accurate summary (never more than 3 sentences).

## Tech Stack

-   **UI**: Gradio
-   **API**: OpenAI-compatible Chat API (`requests` library)
-   **Language Model**: `meta-llama/Llama-3.2-3B-Instruct` (LLM-only, no rule-based fallback)
-   **Hosting**: Hugging Face Spaces

## Running Locally

To run this project on your own machine:

1.  **Clone the repository:**
    ```bash
    git clone https://huggingface.co/spaces/Snaseem2026/technical-description-assistant
    cd technical-description-assistant
    ```

2.  **Install dependencies:**
    ```bash
    pip install -r requirements.txt
    ```

    (The app now requires the `requests` library for direct API calls.)

3.  **Set your Hugging Face token (required):**
    The application requires a valid token to access the Inference API.
    ```bash
    export HUGGING_FACE_HUB_TOKEN="hf_YOUR_TOKEN_HERE"
    ```

4.  **Run the application:**
    ```bash
    python app.py
    ```