Update agent prompt

agent.py

@@ -9,8 +9,9 @@ from tools import (
     excel_reader,
     math_calculator,
     search_engine,
-    image_transcriber,
+    image_analyzer,
     YoutubeTranscriberTool,
+    YoutubeVideoDescriptorTool,
 )
 from utils import OPENAI_MODEL_ID
 
@@ -29,6 +30,12 @@ Each question will contain the question, the task ID, and if available, a file n
 Key points:
 - Do not add any prefix or suffix to YOUR FINAL ANSWER.
 - Answer shortly and concisely, following the template strictly.
+- Again, YOUR FINAL ANSWER should be a number OR as few words as possible OR a comma separated list of numbers and/or strings.
+
+Example:
+[Final Answer: 42]
+Good example: "42"
+Bad example: "The answer is 42"
 """
 
 
@@ -47,8 +54,9 @@ class MainAgent:
                 txt_reader,
                 pdf_reader,
                 excel_reader,
-                image_transcriber,
+                image_analyzer,
                 YoutubeTranscriberTool(),
+                YoutubeVideoDescriptorTool(),
             ],
             max_steps=5,
             model=model,

tools.py

@@ -308,9 +308,8 @@ class YoutubeVideoDescriptorTool(Tool):  # pylint: disable=C0115
             if not ret:
                 break
 
-            # Keep every Nth frame to reduce to target_fps
             if i % frame_interval == 0:
-                frames.append(cv2.cvtColor(frame, cv2.COLOR_BGR2RGB))  # store frame in memory (RGB)
+                frames.append(cv2.cvtColor(frame, cv2.COLOR_BGR2RGB))
 
             i += 1
 
@@ -321,9 +320,13 @@ class YoutubeVideoDescriptorTool(Tool):  # pylint: disable=C0115
             _, buffer = cv2.imencode(".jpg", frame)
             encoded_buffer = base64.b64encode(buffer).decode("utf-8")
             base64_frames.append(encoded_buffer)
-            if len(base64_frames) == 20:  # yield every 20 frames
+            if len(base64_frames) == 20:
+                # yield every 20 frames
                 yield base64_frames
                 base64_frames = []
+        # yield any remaining frames
+        if len(frames) < 20:
+            yield base64_frames
 
     def forward(self, task: str, youtube_url: str) -> str:  # pylint: disable=W0221
         file_name = f"{uuid4()}.mp4"