Resolve merge conflict in requirements.txt

.dockerignore

@@ -10,7 +10,6 @@ build/
 venv/
 env/
 server/
-.git
 .gitignore
 .env
 .env.local

Dockerfile

@@ -1,11 +1,11 @@
-# Use official Python runtime as base image
 FROM python:3.11-slim
 
-# Set working directory in container
 WORKDIR /app
 
-# Install system dependencies (OpenCV, HDF5 support)
+# --- System deps ---
 RUN apt-get update && apt-get install -y \
+    git \
+    git-lfs \
     libsm6 \
     libxext6 \
     libxrender-dev \
@@ -13,35 +13,20 @@ RUN apt-get update && apt-get install -y \
     pkg-config \
     && rm -rf /var/lib/apt/lists/*
 
-# Copy requirements file
-COPY requirements.txt .
+# --- Enable Git LFS ---
+RUN git lfs install
 
-# Install Python dependencies
-RUN pip install --no-cache-dir -r requirements.txt
-
-# Copy application code
-COPY main.py .
-COPY model.py .
-COPY schemas.py .
-COPY inference/ ./inference/
+# --- Copy repo ---
+COPY . .
 
-# Copy pre-trained model (use .keras format only)
-COPY models/Forest_Segmentation_Best.keras ./models/
+# --- Pull LFS files INSIDE container ---
+RUN git lfs pull
 
-# Create logs directory
-RUN mkdir -p logs
+# --- Install Python deps ---
+RUN pip install --no-cache-dir -r requirements.txt
 
-# Expose port (Hugging Face Spaces default)
+# --- Runtime ---
 EXPOSE 7860
-
-# Set environment for production
 ENV PYTHONUNBUFFERED=1
-ENV PORT=7860
-ENV HOST=0.0.0.0
-
-# Health check
-HEALTHCHECK --interval=30s --timeout=10s --start-period=40s --retries=3 \
-    CMD python -c "import requests; requests.get('http://localhost:7860/health')" || exit 1
 
-# Run FastAPI application with uvicorn
 CMD ["uvicorn", "main:app", "--host", "0.0.0.0", "--port", "7860"]

main.py

@@ -1,5 +1,5 @@
 # main.py
-
+from huggingface_hub import hf_hub_download
 from fastapi import FastAPI, HTTPException, Request
 from fastapi.responses import JSONResponse, HTMLResponse
 import tensorflow as tf
@@ -207,11 +207,21 @@ def predict(payload: PredictRequest):
         # ---------------------------------------------------------------------
         # Run model (raw confidence)
         # ---------------------------------------------------------------------
+        MODEL_REPO = "prshntdxt/Forest_Segmentation_Best"
+        MODEL_FILE = "Forest_Segmentation_Best.keras"
+        
+        MODEL_PATH = hf_hub_download(
+            repo_id=MODEL_REPO,
+            filename=MODEL_FILE,
+        )
+        
         model = tf.keras.models.load_model(
-            "models/Forest_Segmentation_Best.keras",
+            MODEL_PATH,
             compile=False
         )
 
+
+
         confidence_map = model.predict(
             input_tensor, verbose=0
         )[0, :, :, 0]

schemas.py

@@ -1,41 +1,71 @@
 from pydantic import BaseModel
-from typing import Dict, List, Optional, Union
+from typing import Dict, List, Optional, Union, Any
+
+# =============================================================================
+# REQUEST SCHEMA
+# =============================================================================
 
 class PredictRequest(BaseModel):
     """
-    Forest segmentation prediction request
-    
-    Supports Landsat 8 Collection 2 Level 2 data format.
-    Each band can be provided as:
-    - Base64-encoded float32 data (for remote API calls)
-    - Array/list format (for direct server calls)
-    
-    Required bands:
-      - Blue, Green, Red: Optical bands
-      - NIR, SWIR1, SWIR2: Infrared bands
-      - NDVI, NDWI, NBR: Spectral indices (or server will compute them)
-    
-    Optional special keys:
-      - _invert_mask: Set to true to invert forest/non-forest in response
-    
-    Value range expectations:
-      - Optical bands: [-0.2, 0.6]
-      - Indices: [-1, 1]
+    Forest segmentation prediction request.
+
+    This schema is intentionally flexible to support:
+    - Supabase Edge Functions
+    - Hugging Face remote inference
+    - Local inference scripts
+
+    Required:
+    - bands: Dict[str, Union[str, List[float]]]
+
+    Allowed extra fields (sent by Supabase):
+    - width, height
+    - bbox
+    - band_names
+    - preprocessing
+    - model_name, model_version
     """
+
     model_name: str = "forest_segmentation"
     model_version: str = "landsat8_v1"
-    bands: Dict[str, Union[str, List, int]]  # Band data as base64 or array
 
+    # Band data:
+    # - base64-encoded float32 (remote calls)
+    # - list/array of floats (local calls)
+    bands: Dict[str, Union[str, List[float]]]
+
+    class Config:
+        # 🔑 CRITICAL FIX
+        # Allows Supabase to send extra metadata without 422 errors
+        extra = "allow"
 
-from typing import List, Dict, Any, Optional
-from pydantic import BaseModel
+
+# =============================================================================
+# RESPONSE SCHEMA
+# =============================================================================
 
 class PredictResponse(BaseModel):
-    mask: List[int]                    # 1D flat list ✓
-    inverted_mask: List[int]           # 1D flat list ✓
+    """
+    Forest segmentation prediction response.
+
+    Mask values are CONTINUOUS (0–255) — NOT binary.
+    """
+
+    # Flattened mask (length = width * height)
+    mask: List[int]
+
+    # Inverted mask (optional utility)
+    inverted_mask: List[int]
+
+    # Statistics
     forest_percentage: float
     forest_confidence: float
     mean_prediction: float
+
+    # Class mapping
     classes: Dict[str, int]
+
+    # Model metadata
     model_info: Dict[str, Any]
+
+    # Optional debug block
     debug: Optional[Dict[str, Any]] = None