clean wiki, stt, model upgrade

agent.py

@@ -1,7 +1,7 @@
 from dotenv import load_dotenv
 from smolagents import CodeAgent
 from smolagents import OpenAIServerModel
-from tool import fetch_webpage, read_file_tool, get_youtube_transcript
+from tool import fetch_webpage, read_file_tool, get_youtube_transcript, transcribe_audio
 
 from smolagents import VisitWebpageTool, WikipediaSearchTool, PythonInterpreterTool, DuckDuckGoSearchTool, WebSearchTool, SpeechToTextTool
 
@@ -10,15 +10,15 @@ from prompt import gaia_prompt
 load_dotenv()
 
 openai_nano_model = OpenAIServerModel(
-    model_id="gpt-4.1-nano-2025-04-14",
+    # model_id="gpt-4.1-nano-2025-04-14",
+    model_id="gpt-4.1-mini-2025-04-14",
     # model_id="o3-mini-2025-01-31",
 )
 
 gaia_agent = CodeAgent(
     model=openai_nano_model,
-    tools=[fetch_webpage, DuckDuckGoSearchTool(), PythonInterpreterTool(), read_file_tool, get_youtube_transcript], # WikipediaSearchTool(), VisitWebpageTool(max_output_length=60000)
+    tools=[fetch_webpage, DuckDuckGoSearchTool(), PythonInterpreterTool(), read_file_tool, get_youtube_transcript, transcribe_audio], # WikipediaSearchTool(), VisitWebpageTool(max_output_length=60000)
     max_steps=7,
-    # verbosity_level=2,
     additional_authorized_imports=["requests", "bs4", "pandas", "numpy", "markdownify"]
 )
 
@@ -34,6 +34,7 @@ class GAIA_Agent:
             full_context = self.system_prompt + "\nTHE QUESTION:\n" + question
 
             final_answer = self.agent.run(full_context)
+            print(f"final answer returned by agent ----> {final_answer}")
             return final_answer
         except Exception as e:
             error = f"An error occurred while processing the question: {e}"
@@ -45,25 +46,10 @@ class GAIA_Agent:
 # clean answer function
 
 if __name__ == "__main__":
-    pass
-    # gaia_agent.run("What is the weather in Mumbai?")
-#     answer = gaia_agent.run(
-# f"""
-# You are a general AI assistant. I will ask you a question. You can answer with the following template:[YOUR FINAL ANSWER]. 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 neither 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. Remember: GAIA requires exact answer matching. Just provide the factual answer.
+    answer = gaia_agent.run(
+        """
+        You are a general AI assistant. I will ask you a question. You can answer with the following template:[YOUR FINAL ANSWER]. 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 neither 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. Remember: GAIA requires exact answer matching. Just provide the factual answer.
 
-# How many studio albums were published by Mercedes Sosa between 2000 and 2009 (included)? You can use the latest 2022 version of english wikipedia.
-# """
-#     )
-
-    # print(gaia_prompt)
-#     answer = gaia_agent.run("""
-#         You are a general AI assistant. I will ask you a question. Report your thoughts, and finish your answer with the following template: FINAL ANSWER: [YOUR FINAL ANSWER]. 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 neither 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.
-#         You can search for results and then visit a webpage to get more information. Break down the problem into smaller sub-problems and solve them one by one.
-#         Think like a human.
-
-#         What is the final numeric output from the attached Python code?
-                            
-#         ----
-
-# """)
-#     print(f"this is the final answer the gaia agent gave ---> {answer}")
+        How many studio albums were published by Mercedes Sosa between 2000 and 2009 (included)? You can use the latest 2022 version of english wikipedia."""
+    )
+    pass

app.py

@@ -77,7 +77,7 @@ def run_and_submit_all( profile: gr.OAuthProfile | None):
     results_log = []
     answers_payload = []
     print(f"Running agent on {len(questions_data)} questions...")
-    to_answer = [1, 3, 5, 6, 7, 8, 9, 10, 11, 12, 13, 15, 16, 17, 18, 19, 20]
+    to_answer = [1, 3, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 19, 20]
     for index, item in enumerate(questions_data):
 
         if index + 1 in to_answer:

requirements.txt

@@ -9,4 +9,6 @@ smolagents[openai]
 youtube-transcript-api
 beautifulsoup4
 markdownify
-duckduckgo-search
+duckduckgo-search
+openpyxl
+wikipedia-api

tool.py

@@ -1,5 +1,7 @@
 from smolagents import Tool, tool
 from youtube_transcript_api import YouTubeTranscriptApi
+from transformers import AutoModelForSpeechSeq2Seq, AutoProcessor, pipeline
+import torch
 
 @tool
 def fetch_webpage(url: str, convert_to_markdown: bool = True) -> str:
@@ -31,6 +33,46 @@ def fetch_webpage(url: str, convert_to_markdown: bool = True) -> str:
 
         # for wikipedia only keep the main content
         if "wikipedia.org" in url:
+
+            elements_to_remove = [
+                # Navigation and reference elements
+                {'class': 'navbox'},
+                {'class': 'navbox-group'},
+                {'class': 'reflist'},
+                {'class': 'navigation-box'},
+                {'class': 'sister-project'},
+                {'class': 'metadata'},
+                {'class': 'interlanguage-link'},
+                {'class': 'catlinks'},
+                {'id': 'References'},
+                {'id': 'External_links'},
+                {'id': 'Further_reading'},
+                {'id': 'See_also'},
+                {'id': 'Notes'},
+            ]
+            
+            for selector in elements_to_remove:
+                elements = soup.find_all(attrs=selector)
+                for element in elements:
+                    # For ID-based elements, remove the parent section
+                    if 'id' in selector:
+                        parent = element.parent
+                        if parent and parent.name in ['h2', 'h3', 'h4']:
+                            # Remove heading and all content until next heading
+                            current = parent
+                            while current and current.next_sibling:
+                                next_elem = current.next_sibling
+                                if (hasattr(next_elem, 'name') and 
+                                    next_elem.name in ['h2', 'h3', 'h4']):
+                                    break
+                                if hasattr(next_elem, 'decompose'):
+                                    next_elem.decompose()
+                                else:
+                                    current = next_elem
+                            parent.decompose()
+                    else:
+                        element.decompose()
+
             main_content = soup.find("main",{"id":"content"})
             if main_content:
                 content = md(str(main_content),strip=['script', 'style'], heading_style="ATX").strip()
@@ -79,3 +121,41 @@ def get_youtube_transcript(video_id: str) -> str:
     # raw data is in the form of [{        'text': 'Hey there',        'start': 0.0,        'duration': 1.54    },    {        'text': 'how are you',,        'start': 1.54,        'duration': 4.16    },   ... ] we will return ony the text element as lines
     transcript = "\n".join([item['text'] for item in raw_data])
     return transcript
+
+
+@tool
+def transcribe_audio(audio_path: str) -> str:
+    """
+    Speech to Text - transcribes audio file and returns the text
+
+    Args:
+        audio_path (str): Local file path to the audio
+
+    Returns:
+        str: The transcript of the audio file
+    """
+    
+    device = "cuda:0" if torch.cuda.is_available() else "cpu"
+    torch_dtype = torch.float16 if torch.cuda.is_available() else torch.float32
+    
+    model_id = "openai/whisper-small"
+    
+    model = AutoModelForSpeechSeq2Seq.from_pretrained(
+        model_id, torch_dtype=torch_dtype, low_cpu_mem_usage=True, use_safetensors=True
+    )
+    model.to(device)
+    
+    processor = AutoProcessor.from_pretrained(model_id)
+    
+    pipe = pipeline(
+        "automatic-speech-recognition",
+        model=model,
+        tokenizer=processor.tokenizer,
+        feature_extractor=processor.feature_extractor,
+        torch_dtype=torch_dtype,
+        device=device,
+        chunk_length_s=30,
+    )
+    
+    result = pipe(audio_path)
+    return result