more tools

agent.py

@@ -1,17 +1,25 @@
 import re
 from dotenv import load_dotenv
+import requests
 
+# langchain imports
 from langchain_core.tools import Tool, tool
-from langchain_core.messages import AnyMessage, HumanMessage, SystemMessage
+from langchain_core.messages import HumanMessage
 from langgraph.prebuilt import create_react_agent
 #from langgraph.graph import START, StateGraph
-#from langgraph.prebuilt import tools_condition, create_react_agent
+#from langgraph.prebuilt import tools_condition
 from langchain_community.tools import DuckDuckGoSearchRun
+from langchain_community.retrievers import WikipediaRetriever
 from langchain_experimental.utilities import PythonREPL
 from langchain_google_genai import ChatGoogleGenerativeAI
 #from langchain_ollama import ChatOllama
 from langfuse.langchain import CallbackHandler
 
+# tool imports
+import pandas as pd
+import whisper
+from youtube_transcript_api import YouTubeTranscriptApi
+
 load_dotenv()
 langfuse_handler = CallbackHandler()
 
@@ -25,6 +33,65 @@ with open('system_prompt.txt', 'r', encoding='utf-8') as f:
 
 # --- Tools ---
 
+# python REPL tool
+python_repl = PythonREPL()
+execute_python = Tool(
+    name="execute_python",
+    description="A Python shell. Use this tool to execute python commands. "
+                "Input should be valid python code. "
+                "If you want to see the output of a value, you should print it out with `print(...)`.",
+    func=python_repl.run,
+)
+
+@tool
+def get_youtube_transcript(url: str) -> str:
+    """
+    Retrieve the text transcript of a YouTube video
+
+    Args:
+        url (str): link to the YouTube video
+
+    Returns:
+        str: text transcript
+    """
+    def extract_video_id(url: str) -> str:
+        # extracts video id from youtube url
+        patterns = [
+            r"v=([a-zA-Z0-9_-]{11})",                  # regular link
+            r"youtu\.be/([a-zA-Z0-9_-]{11})",          # shortened link
+            r"youtube\.com/embed/([a-zA-Z0-9_-]{11})", # embed link
+        ]
+        for pattern in patterns:
+            match = re.search(pattern, url)
+            if match:
+                return match.group(1)
+        raise ValueError("Invalid YouTube URL")
+
+    try:
+        video_id = extract_video_id(url)
+        api = YouTubeTranscriptApi()
+        transcript = api.fetch(video_id)
+        txt = '\n'.join([s.text for s in transcript.snippets])
+        return txt
+    except Exception as e:
+        return f"An error occured using get_youtube_transcript tool: {e}"
+
+@tool
+def reverse_string(text: str) -> str:
+    """
+    A tool to reverse the order of characters in a text string
+
+    Args:
+        text (str): text string to reverse
+
+    Returns:
+        str: reversed text string
+    """
+    try:
+        return text[::-1]
+    except Exception as e:
+        return f"An error occured using reverse_string tool: {e}"
+
 @tool
 def search_web(query: str) -> str:
     """
@@ -38,38 +105,105 @@ def search_web(query: str) -> str:
     """
     try:
         search = DuckDuckGoSearchRun()
-        search.invoke(query)
+        return search.invoke(query)
     except Exception as e:
         return f"An error occured using search_web tool: {e}"
 
-# python REPL tool
-python_repl = PythonREPL()
-repl_tool = Tool(
-    name="python_repl",
-    description="A Python shell. Use this to execute python commands. "
-                "Input should be a valid python command. "
-                "If you want to see the output of a value, you should print it out with `print(...)`.",
-    func=python_repl.run,
-)
+@tool
+def search_wikipedia(query: str) -> str:
+    """
+    A tool to perform a search for a query using Wikipedia
+    
+    Args:
+        query (str): query to search on Wikipedia
 
+    Returns:
+        str: wikipedia search result
+    """
+    try:
+        retriever = WikipediaRetriever()
+        return retriever.invoke(query)
+    except Exception as e:
+        return f"An error occured using search_wiki tool: {e}"
+    
 @tool
-def reverse_string(text: str) -> str:
+def transcribe_audio(url: str) -> str: 
     """
-    Reverse the order of characters in a text string
+    A tool to transcribe an audio file (.mp3) using an automatic speech recognition model
 
     Args:
-        text (str): text string to reverse
+        url (str): link to audio file (.mp3)
 
     Returns:
-        str: reversed text string
+        str: transcript of the audio file
     """
     try:
-        return text[::-1]
+        # fetch audio file
+        response = requests.get(url)
+        response.raise_for_status()
+
+        tmp = 'tmp_audio.mp3'
+        with open(tmp, "wb") as f:
+            f.write(response.content)
+
+        # transcribe
+        model = whisper.load_model('tiny')
+        result = model.transcribe(tmp)
+
+        return result['text']
     except Exception as e:
-        return f"An error occured using reverse_string tool: {e}"
+        return f"An error occured using transcribe_audio tool: {e}"
+
+def view_py_file(url: str) -> str:
+    """
+    A tool to view the contents of a python file (.py)
 
+    Args:
+        url (str): link to python file (.py)
+
+    Returns:
+        str: contents of python file
+    """
+    try:
+        # fetch python file
+        response = requests.get(url)
+        response.raise_for_status()
+
+        return response.text
+    except Exception as e:
+        return f"An error occured using view_py_file tool: {e}"
+    
+def view_xlsx_file(url: str) -> str:
+    """
+    A tool to view the contents of an excel file (.xlsx)
+
+    Args:
+        url (str): link to excel file (.xlsx)
+
+    Returns:
+        str: contents of excel file
+    """
+    try:
+        # fetch python file
+        response = requests.get(url)
+        response.raise_for_status()
+
+        tmp = 'tmp.xlsx'
+        with open(tmp, "wb") as f:
+            f.write(response.content)
+
+        data = pd.read_excel('tmp.xlsx')
+
+        return data.to_string()
+    except Exception as e:
+        return f"An error occured using view_xlsx_file tool: {e}"
 
-tools = [search_web, reverse_string, repl_tool]
+# agent toolkit
+tools = [
+    execute_python, get_youtube_transcript,
+    reverse_string, search_web, transcribe_audio,
+    view_py_file, view_xlsx_file
+]
 
 # --- LangGraph ---
 agent = create_react_agent(
@@ -102,7 +236,7 @@ class GAIAAgent:
         if match:
             final_answer = match.group(0)
         else:
-            final_answer = f"I don't know\n'{final_message}'"
+            final_answer = final_message
 
         print(f"Agent returning answer: {final_answer}")
         return final_answer

app.py

@@ -83,15 +83,10 @@ def run_and_submit_all( profile: gr.OAuthProfile | None):
             continue
         if file_name:
             # if question has an attached file, add the URL to the question prompt
+            file_ext = file_name.split('.')[-1]
             file_url = f"{api_url}/files/{task_id}"
-            question_text += f'\nFile URL: "{file_url}"'
-
-            # add file type to prompt also
-            try:
-                ext = file_name.split('.')[-1]
-                question_text += f" (.{ext} file)"
-            except Exception as e:
-                print(f"Error getting file extension on {task_id}: {e}")
+            question_text += f'\nFILE URL ({file_ext}): "{file_url}"'
+
         try:
             submitted_answer = agent(question_text)
             answers_payload.append({"task_id": task_id, "submitted_answer": submitted_answer})

requirements.txt

@@ -8,4 +8,8 @@ langchain_core
 langchain_huggingface
 langchain_google_genai
 langchain_experimental
-langfuse
+langfuse
+pandas
+openpyxl
+openai-whisper
+youtube_transcript_api