cnh-clm: Large Language Model

A modular, containerized pipeline that combines Large Language Models (LLMs), OCR, CVE parsing, and retrieval-based generation into a unified chatbot interface for cybersecurity red teaming support.

Features

• Web-based Chatbot Interface using FastAPI + WebSockets

• LLM Integration with customizable prompt & generation backends

• OCR Extraction to parse text from cybersecurity documents

• CVE/ExploitDB Preprocessing for use in retrieval-augmented generation

• Vector Store Retrieval using custom or pretrained embedding models

• Interactive Notebooks for experimentation and finetuning

• Docker Support for seamless containerized deployment

Getting Started

1. Clone the Repository git clone https://github.com/iamasr9999/cnh-clm.git cd cnh-clm

2. Install Dependencies

   • pip install -r requirements.txt

   Or use Docker

   — command to build docker

   • docker build -t tag_name .

   — Push to docker hub

   • docker login

   — Tag your image

   • docker tag tag_name docker_username/repository:tag

   — push it

   • docker push docker_username/repository:tag

3. Run the Application on server with uvicorn

   • While starting the job make sure to open port 8000 and 22
   • This command "/src/venv/bin/uvicorn", "api_server:app", "--host", "0.0.0.0", "--port", "8000" is already in the dockerfile so when you start a job the process stats.
   • Visit http://localhost:8000 to access UI

LLM Architecture

Embedding Store: Converts input data into vector form

Retriever: Selects relevant context based on input query

Generator: Produces final LLM response

Supports plug-and-play with other models

Chatbot UI

FastAPI serves chatbot-ui.html, allowing:

Prompt submission and streaming response

Resume/edit previous messages

Visual context (images from static/)

Logs

Logs for server activity are saved to:
- logs/server.log

• cnh-clm: Large Language Model
  • Features
  • Getting Started
  • LLM Architecture
  • Chatbot UI
  • Logs