SafeLLM API 🚀

This is a Flask-based API that processes user queries, ensuring they are non-toxic and appropriate. It uses pre-trained NLP models for query classification, toxicity detection, and query rewriting.

Features ✨

• 🛡️ Detects and blocks toxic or inappropriate content.
• 🧠 Classifies query intents and restricts certain types of queries.
• ✍️ Rewrites queries for better clarity.
• 🤖 Generates responses to rewritten queries.

Tech Stack 🛠️

• Flask: Backend framework for the API.
• Transformers Library: For text classification and query rewriting.
• Detoxify: For toxicity detection in text.

Endpoints 🔗

/query (Method: POST)

Processes user queries by:

1. Checking for toxicity.
2. Classifying intent.
3. Rewriting queries for clarity.
4. Generating a response.

Request Body (JSON): 📤

{
  "query": "Your input query here"
}

Response: 📥

🚫 If query contains toxicity:

{
  "query": "Query contains inappropriate content."
}

Status Code: 403

Response:

☣️ If query is restricted:

{
  "query": "Restricted query detected."
}

Status Code: 403

Response:

✅ If query is valid:

{
  "query": "LLM Response to: Your clarified query"
}

Status Code: 200

Installation 🛠️

Prerequisites

• Python 3.8 or later
• pip package manager

Setup ⚡

1. Clone the repository:

   git clone git@github.com:Clecotech/SafeLLM.git
   cd SafeLLM

2. Install dependencies:

  pip install -r requirements.txt

3. Run the application:

   python app.py
   

4. The app will run on http://localhost:5001 🌐

Environment Configuration 🌱

You can set up a virtual environment for this project:

python -m venv venv  
source venv/bin/activate  # On Windows: venv\Scripts\activate  
pip install -r requirements.txt  

Example Request 📡

cURL:

curl -X POST http://localhost:5001/query \
-H "Content-Type: application/json" \
-d '{"query": "Explain the process of natural language processing."}'