Update app.py

app.py

@@ -1,20 +1,20 @@
+import os
+import requests
 from fastapi import FastAPI, File, UploadFile, Form
 from fastapi.responses import JSONResponse
-from transformers import pipeline
-from io import BytesIO
 from PIL import Image
-import os
-
-hf_token = os.environ.get("HF_TOKEN")
+from io import BytesIO
+import base64
 
 app = FastAPI()
 
-# Load pipeline once
-pipe = pipeline("image-text-to-text", model="google/medgemma-4b-pt",use_auth_token=hf_token)
+HF_API_URL = "https://api-inference.huggingface.co/models/google/medgemma-27b-it"
+HF_TOKEN = os.getenv("HF_TOKEN")
 
-@app.get("/")
-def root():
-    return {"message": "MedGemma Image-to-Text API running"}
+headers = {
+    "Authorization": f"Bearer {HF_TOKEN}",
+    "Content-Type": "application/json"
+}
 
 @app.post("/analyze/")
 async def analyze_image(prompt: str = Form(...), file: UploadFile = File(...)):
@@ -22,17 +22,28 @@ async def analyze_image(prompt: str = Form(...), file: UploadFile = File(...)):
         image_data = await file.read()
         image = Image.open(BytesIO(image_data)).convert("RGB")
 
-        response = pipe([
-            {
-                "role": "user",
-                "content": [
-                    {"type": "image", "image": image},
-                    {"type": "text", "text": prompt}
-                ]
-            }
-        ])
-
-        return {"result": response[0]['generated_text']}
+        buffered = BytesIO()
+        image.save(buffered, format="JPEG")
+        img_str = base64.b64encode(buffered.getvalue()).decode()
+
+        payload = {
+            "inputs": [
+                {
+                    "role": "user",
+                    "content": [
+                        {"type": "image", "image": img_str},
+                        {"type": "text", "text": prompt}
+                    ]
+                }
+            ]
+        }
+
+        response = requests.post(HF_API_URL, headers=headers, json=payload)
+
+        if response.status_code != 200:
+            return JSONResponse(status_code=response.status_code, content=response.json())
+
+        return {"result": response.json()[0]['generated_text']}
 
     except Exception as e:
         return JSONResponse(status_code=500, content={"error": str(e)})