Update app.py

app.py

@@ -6,10 +6,18 @@ import base64
 from PIL import Image
 import io
 
+
+def resize_image(image_path, max_size=(800, 800), quality=85):
+    with Image.open(image_path) as img:
+        img.thumbnail(max_size, Image.ANTIALIAS)
+        buffer = io.BytesIO()
+        img.save(buffer, format="JPEG", quality=quality)
+        return buffer.getvalue()
+        
 def filepath_to_base64(image_path):
-    with open(image_path, "rb") as image_file:
-        img_base64 = base64.b64encode(image_file.read())
-    return f"data:image/png;base64,{img_base64.decode('utf-8')}"
+    img_bytes = resize_image(image_path)
+    img_base64 = base64.b64encode(img_bytes)
+    return f"data:image/jpeg;base64,{img_base64.decode('utf-8')}"
 
 # A chave da API é lida de uma variável de ambiente para maior segurança.
 api_key = os.getenv('API_KEY')
@@ -28,12 +36,11 @@ def call_fuyu_8b_api(image_path, content, temperature=0.2, top_p=0.7, max_tokens
         "accept": "text/event-stream",
         "content-type": "application/json",
     }
-    # Gera uma seed aleatória se nenhuma for fornecida pelo usuário.
     if seed is None or seed == "":
         seed = random.randint(0, 18446744073709552000)
     else:
         try:
-            seed = int(seed)  # Garante que a seed fornecida seja um inteiro.
+            seed = int(seed)
         except ValueError:
             return "Seed must be an integer."
 
@@ -55,11 +62,9 @@ def call_fuyu_8b_api(image_path, content, temperature=0.2, top_p=0.7, max_tokens
     if response.status_code != 200:
         print(f"Erro na requisição: {response.status_code}")
         try:
-            # Tenta imprimir a mensagem de erro detalhada, se disponível
             error_details = response.json()
             print(error_details)
         except ValueError:
-            # Se a resposta de erro não for um JSON, imprime o texto bruto
             print(response.text)
     else:
         print(response)
@@ -69,7 +74,6 @@ def call_fuyu_8b_api(image_path, content, temperature=0.2, top_p=0.7, max_tokens
                 decoded_line = line.decode("utf-8")
                 if "content" in decoded_line:
                     response_text += decoded_line.split('"content":"')[1].split('","finish_reason')[0]
-        
         return response_text
 
 # Definindo os componentes da interface
@@ -78,7 +82,7 @@ image_input = gr.Image(type="filepath", label="Upload Image")
 temperature_input = gr.Slider(minimum=0, maximum=1, step=0.01, value=0.2, label="Temperature")
 top_p_input = gr.Slider(minimum=0, maximum=1, step=0.01, value=0.7, label="Top P")
 max_tokens_input = gr.Slider(minimum=1, maximum=1024, step=1, value=1024, label="Max Tokens")
-seed_input = gr.Textbox(label="Seed (optional)")  
+seed_input = gr.Textbox(label="Seed (optional)")
 
 # Criando a interface Gradio
 iface = gr.Interface(fn=call_fuyu_8b_api,
@@ -88,4 +92,4 @@ iface = gr.Interface(fn=call_fuyu_8b_api,
                      description="Explore the capabilities of Fuyu-8B multi-modal transformer.")
 
 # Executando a interface
-iface.launch()
+iface.launch()