video update

agent.py

@@ -281,7 +281,23 @@ def analyze_video(video_path: str, question: str) -> str:
     """
     if cv2 is None:
         return "Error: cv2 is not installed. Please install opencv-python."
+    
+    temp_dir = tempfile.gettempdir()
+    downloaded_video = None
+    
     try:
+        # Check if video_path is a URL
+        if video_path.startswith("http"):
+            print(f"Downloading video from URL: {video_path}")
+            downloaded_video = os.path.join(temp_dir, f"video_{int(time.time())}.mp4")
+            try:
+                # Use yt-dlp to download the video
+                # Note: --ffmpeg-location could be used if we knew where it was, but we assume it's in path or missing
+                subprocess.run(["yt-dlp", "-f", "best[ext=mp4]/mp4", "-o", downloaded_video, video_path], check=True, timeout=120)
+                video_path = downloaded_video
+            except Exception as e:
+                return f"Error downloading video from URL: {str(e)}. Tip: Check if yt-dlp is installed and the URL is valid."
+
         # 1. Extract frames evenly spaced throughout the video
         cap = cv2.VideoCapture(video_path)
         total_frames = int(cap.get(cv2.CAP_PROP_FRAME_COUNT))
@@ -341,6 +357,12 @@ def analyze_video(video_path: str, question: str) -> str:
         return f"Video Summary based on extracted frames and audio:\n{video_context}"
     except Exception as e:
         return f"Error analyzing video: {str(e)}"
+    finally:
+        if downloaded_video and os.path.exists(downloaded_video):
+            try:
+                os.remove(downloaded_video)
+            except:
+                pass
 
 @tool
 def read_url(url: str) -> str:
@@ -469,28 +491,16 @@ def answer_message(state: AgentState) -> AgentState:
     current_date = datetime.datetime.now().strftime("%Y-%m-%d")
     
     prompt = [SystemMessage(f"""
-    You are a GAIA question answering expert. 
-    Your task is to provide an answer to a question. 
-    Think carefully before answering the question. 
+    You are a master of the GAIA benchmark, a general AI assistant designed to solve complex multi-step tasks.
+    Think carefully and logically. Use your tools effectively. Use your internal monologue to plan your steps.
     
     TODAY'S EXACT DATE is {current_date}. Keep this in mind for all time-sensitive queries.
 
-    CRITICAL RULES FOR FILES & TOOLS:
-    1. If a message contains a path like `[Attached File Local Path: ...]` followed by an image (.png, .jpg, .jpeg), video, or audio file, YOU MUST USE THE CORRESPONDING TOOL (analyze_image, analyze_video, analyze_audio).
-    2. YOU ARE NOT BLIND. NEVER say "I cannot see images" or "I don't have access to files". Use your tools to see and hear for you!
-    3. If you see a file path, invoke the tool IMMEDIATELY in your first ReAct step.
-    4. For text/data files, use `read_document` or `run_python_script` (especially for .xlsx or .pdf).
-    5. Be thorough. If one tool doesn't give enough info, use another (e.g., search the web for context).
-
-    Example of Tool Trigger:
-    User: "What is in this image? [Attached File Local Path: /path/to/image.png]"
-    Your Thought: "I need to see the image to answer. I will use the analyze_image tool."
-    Your Action: Call `analyze_image(image_path='/path/to/image.png', question='What is in this image?')`
-
-    YOUR FINAL ANSWER should be a number OR as few words as possible OR a comma separated list of numbers and/or strings. 
-    If you are asked for a number, don't use comma to write your number, and don't use units such as $ or percent sign unless specified otherwise. 
-    If you are asked for a string, don't use articles, neither abbreviations (e.g. for cities), and write the digits in plain text unless specified otherwise. 
-    If you are asked for a comma separated list, apply the above rules depending of whether the element to be put in the list is a number or a string.
+    CRITICAL RULES:
+    1. If you see a path like `[Attached File Local Path: ...]` followed by an image, video, or audio file, YOU MUST USE THE CORRESPONDING TOOL (analyze_image, analyze_video, analyze_audio) IMMEDIATELY in your next step.
+    2. Plan your steps ahead. 8 steps is your LIMIT for the reasoning loop, so make every step count.
+    3. If a tool fails (e.g., 429 or 402), the system will automatically try another model for you, so just keep going!
+    4. Be concise and accurate. YOUR FINAL ANSWER should be a number OR as few words as possible OR a comma separated list.
     """)]
     messages = prompt + messages
 
@@ -549,9 +559,10 @@ def answer_message(state: AgentState) -> AgentState:
             "You are a strict output formatter for the GAIA benchmark. "
             "Given a verbose draft answer, extract ONLY the final exact answer required. "
             "Return nothing else. DO NOT include prefixes like 'The answer is'. "
-            "Strip all punctuation points at the end and quotes. "
-            "If the answer is a number, just return the number without commas or units unless specified. "
-            "If it is a name or word, just return the exact string. If a list, return only the comma-separated list."
+            "Strip trailing punctuation like periods and quotes. "
+            "If the answer is a number, just return the number. "
+            "If the answer is a list or set of elements, return them as a COMMA-SEPARATED list (e.g., 'a, b, c'). "
+            "DO NOT strip commas that separate list items."
         )
     )
     final_response, _ = smart_invoke([formatting_sys, HumanMessage(content=extract_text_from_content(draft_response.content))], use_tools=False, start_tier=current_tier)

app copy.py

@@ -57,7 +57,9 @@ questions_url = f"{DEFAULT_API_URL}/questions"
 response = requests.get(questions_url, timeout=15)
 response.raise_for_status()
 questions_data = response.json()
-for item in questions_data[3:4]:
+import time
+print(f"Running agent on {len(questions_data)} questions sequentially to avoid 429 errors...")
+for item in questions_data[6:7]:
     question_text = item.get("question")
     if question_text is None:
         continue
@@ -73,11 +75,13 @@ for item in questions_data[3:4]:
         else:
             question_text += f"\n\n[Attached File: {file_name} (Download Failed)]"
             
-    print(files_text, task_id)
+    print(f"Processing Task ID: {task_id}")
     output = agent(question_text)
     print("Q:", question_text)
     print("A:", output)
     print("-" * 40)
+    # Stagger requests to refill Token bucket and provide space for other concurrent tasks if any
+    time.sleep(5)
 
 
 

requirements.txt

@@ -27,3 +27,5 @@ PyPDF2
 openai-whisper
 langchain-openai
 langchain-google-genai
+yt-dlp
+ffmpeg