🔐 CVE Security Agent (RAG Chatbot)

A FastAPI-based cybersecurity agent that fetches, embeds, and queries CVE data to provide exploit insights, attack scenarios, mitigations, patch suggestions, and CVSS verification using a retrieval-augmented generation (RAG) approach.

---

Features

• ✅ Automated CVE collection from NVD API

• ✅ Embedding and vectorization using ChromaDB + all-MiniLM-L6-v2

• ✅ Fast retrieval of CVE details (description, severity, attack scenario, affected products, references, mitigation)

• ✅ Summarization and patch suggestions via LLM

• ✅ Verify CVSS scores programmatically

• ✅ LangChain Agent Support – multi-step reasoning:

  • Check CVE → fetch exploit → summarize → suggest patch → verify CVSS

• ✅ FastAPI endpoint for programmatic queries

• ✅ Docker-ready deployment for Hugging Face Spaces or local servers

---

Project Structure

My_CVE_Chatbot/
│
├─ src/
│   ├─ datacollection.py      
│   ├─ ingestion.py            
│   ├─ retrieval.py                          
│   └─ agent.py
|   └─ agent_tools.py           
├─ app.py 
├─ requirements.txt
├─ Dockerfile
└─ README.md

---

Setup & Deployment

1. Clone the repository

git clone https://huggingface.co/spaces/<username>/<repo-name>
cd <repo-name>

2. Install dependencies

pip install -r requirements.txt

3. Fetch CVE data and build the vector database

python src/datacollection.py
python src/ingestion.py

4. Run FastAPI server locally

uvicorn src.app:app --host 0.0.0.0 --port 8000

• Open http://localhost:8000/docs to interact with the API using Swagger UI.

5. Example API Request

curl -X POST http://localhost:8000/ask \
     -H "Content-Type: application/json" \
     -d '{"question":"Explain CVE-2023-12345 and suggest mitigation."}'

---

LangChain Multi-Step Agent Workflow

The project also includes a LangChain-based agent that can handle:

1. Check CVE: Retrieve CVE details from the vector database
2. Fetch Exploit: Search known exploits for the CVE
3. Summarize: Provide a concise summary of the vulnerability
4. Suggest Patch: Recommend remediation/patches
5. Verify CVSS: Cross-check CVSS score for accuracy

This allows a full end-to-end security reasoning pipeline in a single agent workflow.

---

Deployment on Hugging Face Spaces

1. Make sure Dockerfile and requirements.txt are present.
2. Push to your Space:

git add .
git commit -m "Deploy FastAPI CVE security agent with LangChain agent"
git push

• Hugging Face will automatically:

  • Build the Docker image
  • Install dependencies
  • Launch the FastAPI app

---

Technologies Used

• FastAPI – lightweight web API framework
• ChromaDB – vector database for CVE embeddings
• Sentence Transformers – generate embeddings for semantic search
• Transformers / Mistral-7B-Instruct – LLM for summarization & patch suggestions
• LangChain – multi-step agent workflow
• NVD API – source of CVE information
• Docker – containerized deployment

---

Future Enhancements

• 🔹 Integrate real-time exploit feeds
• 🔹 Add interactive dashboards for CVE trends and analytics
• 🔹 Extend LangChain agent reasoning for automated CVSS validation
• 🔹 Add automated email alerts for high-severity CVEs

---

=======

new-rag-render

75e7d9bca3f14a67a62059d70433ec9eb8875594