The Premier Autonomous Agent for P. falciparum Diagnostics.

MicroSmart PF is a high-performance diagnostic interface that bridges Computer Vision and Clinical Reasoning. It automates the detection of Malaria parasites in thin blood smears and generates WHO-compliant pathology reports in real-time.

Designed as a professional "Cockpit" for lab technicians and pathologists, it prioritizes speed, accuracy, and dark-mode ergonomics.

MicroSmart PF is the specialized malaria node of the larger MicroSmart Project. We are building a constellation of autonomous agents for hematology and cytology.

• MicroSmart PF: P. falciparum Malaria (Active)
• MicroSmart Heme: Hematology & CBC Analysis (In Development)
• MicroSmart Cyto: Cervical Cancer Screening (R&D)

🌐 Explore the Parent Project

• The Eye (Vision Agent): Powered by YOLOv8. Scans slides at ~40ms/frame to detect Trophozoites, Gametocytes, and WBCs with pixel-perfect bounding boxes.
• The Brain (Reasoning Agent): Powered by Llama 3.3 (via Cerebras). Interprets raw cellular counts, calculates parasitemia levels, and acts as a virtual pathologist to write the final report.

• Professional Workbench: A collapsible, 3-pane dashboard designed for high-throughput screening.
• Batch Processing: Queue multiple slides and process them sequentially without blocking the UI.
• Smart Viewer: High-fidelity deep zoom with AI/RAW toggles (Spacebar shortcut).
• Filmstrip Navigation: Rapidly switch between patient samples using Arrow Keys.
• Zero-Latency UX: Local-first state management with optimized React rendering.

Frontend

Backend

git clone [https://github.com/ujpm/microsmart_pf.git](https://github.com/ujpm/microsmart_pf.git)
cd microsmart_pf

The backend handles image processing and AI inference.

1. Create a virtual environment:

cd backend
python -m venv venv
source venv/bin/activate  # Windows: venv\Scripts\activate

2. Install dependencies:

pip install -r requirements.txt

3. Configure API Keys: Create a .env file in backend/ and add your Cerebras key:

CEREBRAS_API_KEY="csk-REPLACE_WITH_YOUR_KEY"

4. Launch the Server: Note: We bind to 0.0.0.0 to ensure access from cloud IDEs (Codespaces/Gitpod).

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

The frontend is the interactive cockpit.

1. Open a new terminal.
2. Setup and run:

cd frontend
npm install
npm run dev

3. Access the workbench at: http://localhost:5173

graph LR
    User[Lab Technician] -->|Uploads Batch| Cockpit[React Workbench]
    Cockpit -->|Queue Management| Backend[FastAPI Server]
    
    subgraph "The Double-Engine"
        Backend -->|Raw Slide| Vision[YOLOv8 'The Eye']
        Vision -->|Bounding Boxes + Counts| Backend
        Backend -->|Aggregated Data| Brain[Llama 3.3 'The Brain']
        Brain -->|Clinical Pathology Report| Backend
    end
    
    Backend -->|JSON Analysis| Cockpit
    Cockpit -->|Visualizes| Display[Smart Viewer & Console]

Architecture & Development Designed by UJPM

License This project is open source under the MIT License. See LICENSE for details.