phase 12

gpu/server.py

@@ -73,6 +73,17 @@ class OCRResponse(BaseModel):
     results: list[OCRTextResult]
 
 
+class PreprocessRequest(BaseModel):
+    image: str
+    binarize: bool = False
+    deskew: bool = False
+    contrast: bool = True
+
+
+class PreprocessResponse(BaseModel):
+    image: str  # base64 JPEG of processed image
+
+
 class VLMRequest(BaseModel):
     image: str
     prompt: str
@@ -183,6 +194,34 @@ def ocr(req: OCRRequest):
     return OCRResponse(results=[OCRTextResult(**r) for r in results])
 
 
+@app.post("/preprocess", response_model=PreprocessResponse)
+def preprocess_image(req: PreprocessRequest):
+    try:
+        image = _decode_image(req.image)
+    except Exception as e:
+        raise HTTPException(status_code=400, detail=f"Bad image: {e}")
+
+    try:
+        from models.preprocess import preprocess
+        processed = preprocess(
+            image,
+            do_binarize=req.binarize,
+            do_deskew=req.deskew,
+            do_contrast=req.contrast,
+        )
+    except Exception as e:
+        raise HTTPException(status_code=500, detail=f"Preprocessing failed: {e}")
+
+    from PIL import Image as PILImage
+    import io
+    img = PILImage.fromarray(processed)
+    buf = io.BytesIO()
+    img.save(buf, format="JPEG", quality=90)
+    result_b64 = base64.b64encode(buf.getvalue()).decode()
+
+    return PreprocessResponse(image=result_b64)
+
+
 @app.post("/vlm", response_model=VLMResponse)
 def vlm(req: VLMRequest):
     try: