| --- |
| 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 |
|
|
