# Pathora Colposcopy API Documentation

## Overview
FastAPI backend for Pathora Colposcopy Assistant with AI model inference and LLM capabilities.

## Base URL
- Local: `http://localhost:8000`
- Production: `https://huggingface.co/spaces/ManalifeAI/Pathora_Colposcopy_Assistant`

---

## Endpoints

### Health Check
**GET** `/health`

Check API health status and verify AI models and LLM availability.

**Response:**
```json
{
  "status": "healthy",
  "service": "Pathora Colposcopy API",
  "ai_models": {
    "acetowhite_model": "loaded",
    "cervix_model": "loaded"
  },
  "llm": {
    "gemini_available": true,
    "api_key_configured": true
  }
}
```

---

## AI Model Endpoints

### 1. Acetowhite Contour Detection
**POST** `/api/infer-aw-contour`

Detect acetowhite lesions and generate contour overlays.

**Parameters:**
- `file` (UploadFile): Image file
- `conf_threshold` (float, optional): Confidence threshold (0.0-1.0, default: 0.4)

**Response:**
```json
{
  "status": "success",
  "message": "Inference completed successfully",
  "result_image": "base64_encoded_image",
  "contours": [
    {
      "points": [[x1, y1], [x2, y2], ...],
      "area": 1234.5,
      "confidence": 0.85
    }
  ],
  "detections": 2,
  "confidence_threshold": 0.4
}
```

### 2. Cervix Bounding Box Detection
**POST** `/api/infer-cervix-bbox`

Detect cervix location and return bounding boxes.

**Parameters:**
- `file` (UploadFile): Image file
- `conf_threshold` (float, optional): Confidence threshold (0.0-1.0, default: 0.4)

**Response:**
```json
{
  "status": "success",
  "message": "Cervix bounding box detection completed",
  "result_image": "base64_encoded_image",
  "bounding_boxes": [
    {
      "x1": 100,
      "y1": 150,
      "x2": 400,
      "y2": 450,
      "confidence": 0.92,
      "class": "cervix"
    }
  ],
  "detections": 1,
  "frame_width": 1920,
  "frame_height": 1080,
  "confidence_threshold": 0.4
}
```

### 3. Batch Image Inference
**POST** `/api/batch-infer`

Process multiple images for acetowhite detection in one request.

**Parameters:**
- `files` (List[UploadFile]): Multiple image files
- `conf_threshold` (float, optional): Confidence threshold (default: 0.4)

**Response:**
```json
{
  "status": "completed",
  "total_files": 3,
  "results": [
    {
      "filename": "image1.jpg",
      "status": "success",
      "result_image": "base64...",
      "contours": [...],
      "detections": 2
    }
  ]
}
```

### 4. Single Frame Analysis
**POST** `/infer/image`

Analyze single image for cervix quality assessment.

**Parameters:**
- `file` (UploadFile): Image file

**Response:**
```json
{
  "status": "Excellent",
  "quality_percent": 95,
  "cervix_detected": true,
  "focus_score": 0.89,
  "brightness_score": 0.92
}
```

### 5. Video Frame Analysis
**POST** `/infer/video`

Process video frames for quality assessment.

**Parameters:**
- `file` (UploadFile): Video file

**Response:**
```json
{
  "total_frames": 150,
  "results": [
    {
      "frame": 0,
      "status": "Excellent",
      "quality_percent": 95
    }
  ]
}
```

---

## LLM Endpoints

### 6. Chat with AI Assistant
**POST** `/api/chat`

Conversational AI endpoint for colposcopy guidance.

**Request Body:**
```json
{
  "message": "What are the signs of high-grade lesions?",
  "history": [
    {
      "role": "user",
      "text": "Hello"
    },
    {
      "role": "bot",
      "text": "Hello! I'm Pathora AI..."
    }
  ],
  "system_prompt": "Optional custom system prompt"
}
```

**Response:**
```json
{
  "status": "success",
  "response": "High-grade lesions typically show...",
  "model": "gemini-1.5-flash"
}
```

### 7. Generate Colposcopy Report
**POST** `/api/generate-report`

Generate comprehensive colposcopy report based on patient data and findings.

**Request Body:**
```json
{
  "patient_data": {
    "age": 35,
    "gravida": 2,
    "para": 2,
    "lmp": "2024-02-01",
    "indication": "Abnormal Pap smear"
  },
  "exam_findings": {
    "native": {
      "cervix_visible": true,
      "transformation_zone": "Type 1"
    },
    "acetic_acid": {
      "acetowhite_lesions": true,
      "location": "6-9 o'clock"
    },
    "green_filter": {
      "vascular_patterns": "Punctation"
    },
    "lugol": {
      "iodine_uptake": "Partial"
    }
  },
  "images": [],
  "system_prompt": "Optional custom prompt"
}
```

**Response:**
```json
{
  "status": "success",
  "report": "COLPOSCOPY REPORT\n\nCLINICAL SUMMARY:\n...",
  "model": "gemini-1.5-flash"
}
```

---

## Environment Variables

Required for LLM functionality:

```bash
GEMINI_API_KEY=your_api_key_here
VITE_GEMINI_API_KEY=your_api_key_here  # For frontend compatibility
```

Get your API key from: https://makersuite.google.com/app/apikey

---

## Error Responses

All endpoints return standardized error responses:

```json
{
  "detail": "Error message description"
}
```

**Common HTTP Status Codes:**
- `400`: Bad Request (invalid file, parameters)
- `500`: Internal Server Error (AI model error, processing failure)
- `503`: Service Unavailable (LLM not configured, API key missing)

---

## Model Information

### AI Models
- **Acetowhite Detection**: YOLO-based segmentation model (`AW_yolo.pt`)
- **Cervix Detection**: YOLO-based object detection model (`cervix_yolo.pt`)

### LLM Model
- **Gemini 1.5 Flash**: Google's generative AI for chat and report generation
  - Temperature: 0.4 (balanced between creativity and consistency)
  - Max Output Tokens: 2048

---

## Usage Examples

### Python
```python
import requests

# AI Model Inference
with open('image.jpg', 'rb') as f:
    response = requests.post(
        'http://localhost:8000/api/infer-aw-contour',
        files={'file': f},
        data={'conf_threshold': 0.5}
    )
    result = response.json()

# Chat
response = requests.post(
    'http://localhost:8000/api/chat',
    json={
        'message': 'What is Reid colposcopic index?',
        'history': []
    }
)
chat_result = response.json()
```

### JavaScript/TypeScript
```typescript
// AI Model Inference
const formData = new FormData();
formData.append('file', imageFile);
formData.append('conf_threshold', '0.5');

const response = await fetch('/api/infer-aw-contour', {
  method: 'POST',
  body: formData
});
const result = await response.json();

// Chat
const chatResponse = await fetch('/api/chat', {
  method: 'POST',
  headers: { 'Content-Type': 'application/json' },
  body: JSON.stringify({
    message: 'Explain transformation zone types',
    history: []
  })
});
const chatResult = await chatResponse.json();
```

---

## Development

### Running Locally
```bash
# Install dependencies
cd backend
pip install -r requirements.txt

# Set environment variables
export GEMINI_API_KEY=your_key

# Run server
uvicorn backend.app:app --reload --host 0.0.0.0 --port 8000
```

### Building with Docker
```bash
docker build -t pathora-colpo .
docker run -p 7860:7860 -e GEMINI_API_KEY=your_key pathora-colpo
```