File size: 997 Bytes
311aeb0
c8d30bc
a555029
d9305b4
c8d30bc
 
 
 
311aeb0
d9305b4
311aeb0
 
d9305b4
 
a555029
c8d30bc
 
 
d9305b4
 
 
 
 
c8d30bc
 
 
a555029
d9305b4
c8d30bc
 
 
 
d9305b4
 
a555029
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
---
title: ThreatHunter
emoji: 🛡
colorFrom: blue
colorTo: red
sdk: docker
app_port: 7860
pinned: true
license: mit
short_description: AI Security Intelligence on AMD MI300X
---

# ThreatHunter

AI Multi-Agent Security Intelligence Platform powered by AMD Instinct MI300X + Qwen2.5-32B-Instruct

## Features

- Multi-Agent Pipeline: Orchestrator, Security Guard, Scout, Intel Fusion, Analyst, Critic, Advisor
- Qwen2.5 on AMD MI300X: ROCm-optimized vLLM inference with 32K context window
- Real-time Monitoring: Live SSE streaming of agent work logs
- Harness Engineering: 3-pillar safety architecture
- Dual-layer Memory: JSON + LlamaIndex persistent agent memory

## Usage

1. Enter package names (e.g. Django 4.2, Redis 7.0) or drag a .py file
2. Click Scan
3. Watch the multi-agent pipeline analyze vulnerabilities in real-time

## AMD Developer Hackathon 2026

- GPU: AMD Instinct MI300X (192GB HBM3)
- Model: Qwen/Qwen2.5-32B-Instruct via ROCm vLLM
- Framework: CrewAI + LiteLLM + FastAPI