phase 9

gpu/server.py

@@ -25,6 +25,7 @@ from config import get_config, get_device, update_config
 from models import registry
 from models.yolo import detect as yolo_detect
 from models.ocr import ocr as ocr_run
+from models.vlm import query as vlm_query
 
 logger = logging.getLogger(__name__)
 
@@ -72,6 +73,18 @@ class OCRResponse(BaseModel):
     results: list[OCRTextResult]
 
 
+class VLMRequest(BaseModel):
+    image: str
+    prompt: str
+    model: str | None = None
+
+
+class VLMResponse(BaseModel):
+    brand: str
+    confidence: float
+    reasoning: str
+
+
 class ConfigUpdate(BaseModel):
     device: str | None = None
     yolo_model: str | None = None
@@ -170,6 +183,21 @@ def ocr(req: OCRRequest):
     return OCRResponse(results=[OCRTextResult(**r) for r in results])
 
 
+@app.post("/vlm", response_model=VLMResponse)
+def vlm(req: VLMRequest):
+    try:
+        image = _decode_image(req.image)
+    except Exception as e:
+        raise HTTPException(status_code=400, detail=f"Bad image: {e}")
+
+    try:
+        result = vlm_query(image, req.prompt)
+    except Exception as e:
+        raise HTTPException(status_code=500, detail=f"VLM failed: {e}")
+
+    return VLMResponse(**result)
+
+
 if __name__ == "__main__":
     import uvicorn