add app.py, Dockerfile, README for HF Spaces

README.md

@@ -1,7 +1,82 @@
-# AI Builders 2025 : CodeSentinel | CWE Classifier for C/C++
+---
+title: CodeSentinel
+emoji: 🛡️
+colorFrom: green
+colorTo: gray
+sdk: docker
+app_port: 7860
+pinned: false
+---
 
-## Overview:
-CodeSentinel is an AI-based system designed to automatically detect and classify software vulnerabilities in C and C++ code using the Common Weakness Enumeration (CWE) standard. By utilizing Machine Learning (ML) and Natural Language Processing (NLP) techniques, this project aims to automate the identification of common security vulnerabilities in source code, significantly reducing the time and effort required for manual analysis.
+# CodeSentinel
+
+Vulnerability classification tool combining fine-tuned ML models with MITRE framework coverage.
+
+Paste a **code snippet**, **CVE description**, or **bug report** — CodeSentinel identifies the vulnerability type, severity, and (for AI/ML inputs) the relevant ATLAS attack technique.
+
+## What it does
+
+- **Code input** → Qwen2.5-Coder 7B analyzes the code → RoBERTa classifies the CWE
+- **Text input** → RoBERTa classifies directly from the description
+- **AI/ML input** → ATLAS pattern matcher identifies the relevant attack technique
+
+## Models
+
+| Model | Purpose | Accuracy |
+|-------|---------|----------|
+| [`martynattakit/vuln-classifier-roberta`](https://huggingface.co/martynattakit/vuln-classifier-roberta) | CWE classification from text | Macro F1: 0.850 |
+| [`martynattakit/vuln-analyzer-qwen-lora`](https://huggingface.co/martynattakit/vuln-analyzer-qwen-lora) | Code → vulnerability description | Eval loss: — |
+
+## Coverage
+
+**CWE Top 25** (MITRE 2024):
+CWE-787, CWE-79, CWE-89, CWE-416, CWE-78, CWE-20, CWE-125, CWE-22, CWE-352, CWE-434, CWE-862, CWE-476, CWE-287, CWE-190, CWE-502, CWE-77, CWE-119, CWE-798, CWE-918, CWE-306, CWE-362, CWE-269, CWE-94, CWE-863, CWE-276
+
+**MITRE ATLAS** (25 techniques):
+Prompt injection, data poisoning, model extraction, membership inference, adversarial examples, jailbreaking, and more.
+
+## Known limitations
+
+- **CWE-77**: 0 F1 — insufficient training samples. Predictions for this class are unreliable.
+- **CWE-863**: F1 0.60 — semantic overlap with CWE-862 makes these hard to distinguish.
+- **ATLAS matching** uses keyword signals + retrieval, not a fine-tuned classifier. Confidence scores reflect signal overlap, not ground-truth accuracy. No labeled ATLAS dataset exists yet.
+- **Code analysis** training data is primarily C/C++ (BigVul). Python/JS/Go descriptions may be less precise.
+
+## Stack
+
+```
+RoBERTa-base        fine-tuned on 165k CVE→CWE pairs (xamxte/cve-to-cwe)
+Qwen2.5-Coder-7B    QLoRA fine-tuned on BigVul (1,596 samples)
+ATLAS matcher       keyword RAG over 25 hand-crafted MITRE case studies
+FastAPI             REST API backend
+```
+
+## Local development
+
+```bash
+pip install -r requirements.txt
+python app.py
+# → http://localhost:7860
+```
+
+## Project structure
+
+```
+pipeline/
+    classifier.py      RoBERTa inference wrapper
+    code_analyzer.py   Qwen inference wrapper  
+    atlas_matcher.py   ATLAS pattern matcher
+    router.py          Input routing + output card
+api/
+    main.py            FastAPI endpoints
+frontend/
+    index.html         Web UI
+data/
+    atlas_cases.json   25 MITRE ATLAS techniques (hand-crafted)
+notebooks/
+    01_roberta_finetune.ipynb
+    02_qwen_qlora.ipynb
+```
 
 ## Links
 - [Try the application here!](https://huggingface.co/spaces/martynattakit/CodeSentinel-CWE_Classification)

app.py

@@ -0,0 +1,30 @@
+"""
+app.py
+HF Spaces entry point — serves both the FastAPI backend and the frontend UI.
+HF Spaces runs this file directly with: python app.py
+"""
+
+import uvicorn
+from fastapi.staticfiles import StaticFiles
+from fastapi.responses import FileResponse
+from pathlib import Path
+
+from api.main import app
+
+# ── Serve frontend ────────────────────────────────────────────────────────────
+# Mount the frontend folder so index.html is served at "/"
+
+FRONTEND_DIR = Path(__file__).parent / "frontend"
+
+@app.get("/", include_in_schema=False)
+async def serve_frontend():
+    return FileResponse(FRONTEND_DIR / "index.html")
+
+# ── Run ───────────────────────────────────────────────────────────────────────
+if __name__ == "__main__":
+    uvicorn.run(
+        "app:app",
+        host="0.0.0.0",
+        port=7860,      # HF Spaces default port
+        reload=False,
+    )

dockerfile

@@ -0,0 +1,24 @@
+FROM python:3.10-slim
+
+WORKDIR /app
+
+# Install system dependencies
+RUN apt-get update && apt-get install -y \
+    git \
+    curl \
+    && rm -rf /var/lib/apt/lists/*
+
+# Copy requirements first for layer caching
+COPY requirements.txt .
+
+# Install Python dependencies
+RUN pip install --no-cache-dir -r requirements.txt
+
+# Copy project files
+COPY . .
+
+# HF Spaces runs on port 7860
+EXPOSE 7860
+
+# Start the app
+CMD ["python", "app.py"]