File size: 2,681 Bytes
af6cd33 72d96c1 af6cd33 72d96c1 af6cd33 72d96c1 af6cd33 72d96c1 af6cd33 72d96c1 af6cd33 72d96c1 af6cd33 72d96c1 af6cd33 72d96c1 af6cd33 72d96c1 | 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 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 | ---
title: ForgeSight
emoji: 🔍
colorFrom: red
colorTo: gray
sdk: gradio
sdk_version: 5.29.1
app_file: app.py
pinned: true
license: mit
short_description: "Multimodal QC Copilot on AMD MI300X + ROCm"
tags:
- amd
- rocm
- mi300x
- qwen
- vllm
- quality-control
- agents
---
# 🔍 ForgeSight — Multimodal QC Copilot on AMD Instinct™ MI300X
ForgeSight is a production-ready **Agentic Quality Control (QC) Pipeline** designed for high-throughput manufacturing environments. Built exclusively for the **AMD + lablab.ai Developer Hackathon**, it leverages the massive 192GB VRAM of the **AMD Instinct MI300X** to run a state-of-the-art multimodal multi-agent workflow.
## 🚀 Key Features
* **Multimodal Reasoning**: Uses **Qwen2-VL-7B** to "see" and understand complex assembly line defects in a single forward pass.
* **4-Agent Pipeline**: Chained reasoning workflow:
1. **Inspector** — Identifies surface defects, anomalies, and violations.
2. **Diagnostician** — Performs industry-literate root-cause analysis.
3. **Action** — Generates prioritized work orders and tool checklists.
4. **Reporter** — Summarizes findings into human-readable executive reports.
* **MI300X Optimized**: Served via **vLLM on ROCm**, utilizing continuous batching and paged attention for near-instant inference.
* **Audit-Ready**: Generates downloadable **PDF QC Audit Reports** for every inspection.
* **Persistent Data**: Integrated with **MongoDB Atlas** for long-term defect tracking and telemetry history.
## 🏗️ Technical Architecture
```mermaid
graph TD
A[React Dashboard] --> B[FastAPI Gateway]
B --> C[Gradio Admin Console]
B --> D[4-Agent Pipeline]
D --> E[AMD MI300X Inference Server]
E --> F[vLLM / ROCm]
F --> G[Qwen2-VL-7B-Instruct]
B --> H[MongoDB Atlas]
B --> I[PDF Generator]
```
### Stack
- **Hardware**: AMD Instinct MI300X (192GB HBM3)
- **Software**: ROCm 6.2, PyTorch 2.4, vLLM
- **Frontend**: React 18, Tailwind CSS, Recharts
- **Backend**: FastAPI, Gradio, Python 3.10
## 🛠️ Installation & Setup
1. **Clone the Repo**: `git clone https://github.com/rasali535/hans.git`
2. **Install Deps**: `pip install -r requirements.txt`
3. **Configure Environment**: Set `AMD_INFERENCE_URL` and `AMD_INFERENCE_TOKEN` in your `.env`.
4. **Launch**: `python hf_space/app.py`
## 📊 Performance on AMD
The MI300X's 5.3 TB/s bandwidth allows ForgeSight to maintain **>2500 tokens/sec** throughput, enabling real-time visual inspection of high-speed manufacturing lines without the latency typical of cloud-based VLM APIs.
---
Built by **Hans** for the **AMD Developer Hackathon**.
|