Update app.py

app.py

@@ -104,12 +104,12 @@ def encode_local_image(image):
     pil_image.save(buffer, format="JPEG")
     return f"data:image/jpeg;base64,{base64.b64encode(buffer.getvalue()).decode('utf-8')}"
 
-# Describe image using the model
-def describe_image(image_path):
+# Describe image using the model with error handling
+def describe_image(image_path, query=''):
     image_string = encode_local_image(image_path)
 
     messages = [
-        {"role": "user", "content": [{"type": "image"}, {"type": "text", "text": os.getenv('USER_PROMPT')}]}
+        {"role": "user", "content": [{"type": "image"}, {"type": "text", "text": f"{os.getenv('USER_PROMPT')}\n{query}" }]}
     ]
 
     prompt_with_template = idefics_processor.apply_chat_template(
@@ -123,12 +123,24 @@ def describe_image(image_path):
         "parameters": {"return_full_text": False, "max_new_tokens": 2048},
     }
 
-    response = idefics_client.post(json=payload).decode()
-    return response
+    try:
+        response = idefics_client.post(json=payload)
+        # Check if the response is empty or not valid JSON
+        if response.status_code != 200 or not response.text:
+            raise ValueError(f"Invalid response: {response.status_code}, {response.text}")
+        return response.text
+    except Exception as e:
+        print(f"Error during image description: {e}")
+        return "Unable to describe the image due to an error."
 
 # Telegram bot client
 client = TelegramClient('bot', api_id, api_hash).start(bot_token=bot_token)
 
+# Get bot user ID for message filtering
+async def get_bot_id():
+    me = await client.get_me()
+    return me.id
+
 # Async function to get OpenAI completion
 async def get_completion(event, user_id, prompt):
     async with client.action(event.chat_id, 'typing'):
@@ -144,14 +156,14 @@ async def get_completion(event, user_id, prompt):
                 model=model,
                 messages=messages,
                 max_tokens=512,
-                temperature=0.5,
+                temperature=0.6,
                 top_p=1.0,
                 frequency_penalty=0.9,
                 presence_penalty=0.9,
             )
             message = response.choices[0].message.content
         except Exception as e:
-            message = f"Error: {str(e)}"
+            message = f"Whoops!"
             print(e)
 
         update_chat_history(user_id, "user", prompt)  # Update history
@@ -161,11 +173,11 @@ async def get_completion(event, user_id, prompt):
 # Telegram bot events
 @client.on(events.NewMessage(pattern='/start'))
 async def start(event):
-    await event.respond("Hello! I am your chat assistant.")
+    await event.respond("Hello!")
 
 @client.on(events.NewMessage(pattern='/help'))
 async def help(event):
-    await event.respond("Here is how I can help you:\n/start - Start the bot\n/help - Get help\n/reset - Reset chat history")
+    await event.respond("Here is how I can help you:\n/start - To check if I am alive\n/help - Show this message\n/reset - Reset chat history")
 
 @client.on(events.NewMessage(pattern='/reset'))
 async def reset(event):
@@ -177,16 +189,24 @@ async def reset(event):
 
 @client.on(events.NewMessage)
 async def handle_message(event):
+    bot_id = await get_bot_id()  # Get bot ID to avoid responding to itself
+
     try:
         user_id = event.chat_id  # Use chat_id to distinguish between users
 
+        # Ignore messages from the bot itself
+        if event.sender_id == bot_id:
+            return
+
+        user_message = event.raw_text
+
         if event.photo:
+            # If an image is sent, describe the image
             photo = await event.download_media()
-            image_description = describe_image(photo)
-            user_message = f"{event.raw_text}\n\nContent of the image: {image_description}"
-        else:
-            user_message = event.raw_text
+            image_description = describe_image(photo, user_message)
+            user_message += f"\n\nContent of the image: {image_description}"
 
+        # Ignore command messages to prevent double processing
         if user_message.startswith('/start') or user_message.startswith('/help') or user_message.startswith('/reset'):
             return
 
@@ -194,7 +214,7 @@ async def handle_message(event):
         await event.respond(response)
     except Exception as e:
         print(f"An error occurred: {e}")
-        await event.respond("Whoopsie 🤭")
+        await event.respond("Whoopsie!")
 
 # Keep-alive function to keep the bot running
 def keep_alive():
@@ -209,10 +229,8 @@ def keep_alive():
                 model=model,
                 messages=messages,
                 max_tokens=10,
-                temperature=0.6,
-                top_p=0.9,
-                frequency_penalty=0.2,
-                presence_penalty=0.6,
+                temperature=0.1,
+                top_p=0.1,
             )
             print(request)
         except Exception as e: