Update app.py

app.py

@@ -31,32 +31,34 @@ class GeminiLLM(LLM):
         return "google-gemini-llm"
     
     def _make_request(self, api_key: str, prompt: str) -> requests.Response:
-        url = f"https://generativelanguage.googleapis.com/v1beta/models/{self.model_name}:generateContent"
-        headers = {
-            "Content-Type": "application/json",
-            "X-goog-api-key": api_key
-        }
-        data = {
-            "contents": [
-                {
-                    "role": "system",
-                    "parts": [
-                        {"text": "You are an agent. Please respond concisely only with the answer, no extra explanations."}
-                    ]
-                },
-                {
-                    "role": "user",
-                    "parts": [
-                        {"text": prompt}
-                    ]
-                }
-            ],
-            "generationConfig": {
-                "temperature": self.temperature
+    url = f"https://generativelanguage.googleapis.com/v1beta/models/{self.model_name}:generateContent"
+    headers = {
+        "Content-Type": "application/json",
+        "X-goog-api-key": api_key
+    }
+    
+    full_prompt = (
+        "You are an agent. Please respond concisely only with the answer, "
+        "no extra explanations.\n\n"
+        f"Question: {prompt}"
+    )
+
+    data = {
+        "contents": [
+            {
+                "role": "user",
+                "parts": [
+                    {"text": full_prompt}
+                ]
             }
+        ],
+        "generationConfig": {
+            "temperature": self.temperature
         }
-    
-        return requests.post(url, headers=headers, json=data)
+    }
+
+    return requests.post(url, headers=headers, json=data)
+
 
 
     def _call(self, prompt: str, stop: Optional[List[str]] = None) -> str:
@@ -238,59 +240,6 @@ def square_root(a: float) -> float:
         return a ** 0.5
     return sqrt(a)
 
-@tool
-def ask_youtube_video(url: str, objeto: str = "bird") -> str:
-    """
-    Analyzes a YouTube video and estimates how many objects of the requested type appear in it.
-    Uses a YOLOv8 object detection model to count appearances in selected frames.
-
-    Args:
-        url (str): The YouTube video URL.
-        objeto (str): Object to search for (default is "bird").
-
-    Returns:
-        str: Approximate number of detected objects in the video or an error message.
-    """
-    import tempfile
-    import cv2
-    from ultralytics import YOLO
-    from pytube import YouTube
-
-    try:
-        # Download video temporarily
-        yt = YouTube(url)
-        video_path = tempfile.NamedTemporaryFile(suffix=".mp4", delete=False).name
-        yt.streams.filter(file_extension="mp4", progressive=True).first().download(filename=video_path)
-
-        # Load pre-trained YOLO model (lightweight for faster analysis)
-        model = YOLO("yolov8s.pt")
-
-        # Process video
-        cap = cv2.VideoCapture(video_path)
-        frame_count, total_detected = 0, 0
-
-        while True:
-            ret, frame = cap.read()
-            if not ret:
-                break
-            frame_count += 1
-
-            # Analyze 1 out of every 30 frames
-            if frame_count % 30 == 0:
-                results = model(frame, verbose=False)
-                for r in results:
-                    for c in r.boxes.cls:
-                        label = model.names[int(c)]
-                        if objeto.lower() in label.lower():
-                            total_detected += 1
-
-        cap.release()
-        return f"Approximately {total_detected} {objeto}(s) detected in the video."
-
-    except Exception as e:
-        return f"Error analyzing video: {str(e)}"
-
-
 # --- Basic Agent Definition ---
 # ----- THIS IS WERE YOU CAN BUILD WHAT YOU WANT ------
 from langchain_core.prompts import PromptTemplate