LLM Emergency Detection Server

A simple HTTP server that processes text through Large Language Models (LLM) to detect emergency situations. Currently supports LM Studio and is prepared for Ollama integration.

Features

• HTTP server implementation for emergency text analysis
• Support for LM Studio integration
• JSON-based API interface
• Configurable port and model settings
• Platform override capabilities

Prerequisites

• Python 3.x
• LM Studio installed and running locally
• Required Python packages:
  • http.server
  • json
  • Custom TBED_LMstudio module

Installation

1. Clone the repository:

git clone [repository-url]

2. Install required dependencies:

pip install -r requirements.txt

Configuration

Default settings in the code:

server_port = 9111
platform = "LMstudio"
port = 1234
url = 'http://localhost:1234/v1/chat/completions'
model = "llama3"

Usage

1. Start the server:

python server.py

2. Send POST requests to http://localhost:9111 with JSON payload:

Basic request format:

{
    "message": "your text here"
}

Advanced request with overrides:

{
    "message": "your text here",
    "platform_override": "LMstudio",
    "port_override": 1234
}

API Reference

Endpoints

POST /

Request Body Parameters

Parameter	Type	Description
message	string	Text to analyze for emergency situations
platform_override	string	(Optional) Override default platform
port_override	integer	(Optional) Override default port

Response Format

Success Response:

{
    "status": "success",
    "message": "Data received successfully",
    "emergency": "response data"
}

Error Response:

{
    "status": "error",
    "message": "error description"
}

Error Handling

The server handles various error scenarios:

• Invalid JSON format (400)
• Server processing errors (500)
• Platform connection issues

Contributing

1. Fork the repository
2. Create your feature branch
3. Commit your changes
4. Push to the branch
5. Create a new Pull Request

License

[Your License Here]

Authors

[Your Name]

Acknowledgments

• LM Studio team
• Contributors to the project