improved tools

app.py

@@ -18,7 +18,8 @@ from tools import (
     video_analysis_tool,
     audio_processing_tool,
     file_type_detection_tool,
-    read_file_tool
+    read_file_tool,
+    wiki_search_tool
 )
 import re
 
@@ -30,6 +31,7 @@ DEFAULT_API_URL = "https://agents-course-unit4-scoring.hf.space"
 llm = ChatOpenAI(model="gpt-3.5-turbo", temperature=0)
 tools = [
     serp_search_tool,
+    wiki_search_tool,
     download_file,
     image_recognition_tool,
     reverse_text_tool,
@@ -52,10 +54,6 @@ class MyAgent(TypedDict):
 # File Handling Functions
 # =========================
 def process_question_with_files(question_data: dict) -> str:
-    """
-    Process a question that may have attached files.
-    Downloads and processes files, then combines with the question.
-    """
     question_text = question_data.get('question', '')
     file_name = question_data.get('file_name', '')
 
@@ -64,44 +62,40 @@ def process_question_with_files(question_data: dict) -> str:
 
     print(f"📎 Processing question with attached file: {file_name}")
     try:
-        # Download the file from the API
+        # Download the file
         file_url = f"{DEFAULT_API_URL}/files/{file_name}"
         local_file_path = f"/tmp/{file_name}"
-        print(f"📥 Downloading file from: {file_url}")
         download_result = download_file(file_url, local_file_path)
-        print(f"📥 Download result: {download_result}")
         if "Failed to download" in download_result:
             return f"{question_text}\n\n[Note: Could not download attached file {file_name}]"
-        # Detect file type
-        file_extension = file_name.lower().split('.')[-1]
-        if file_extension in ['png', 'jpg', 'jpeg', 'gif', 'bmp']:
-            print(f"🖼️ Processing image file: {file_name}")
-            image_result = image_recognition_tool.invoke(local_file_path)
-            enhanced_question = f"{question_text}\n\n[Image Analysis: {image_result}]"
-        elif file_extension in ['mp3', 'wav', 'm4a', 'flac', 'ogg']:
-            print(f"🎵 Processing audio file: {file_name}")
-            audio_result = audio_processing_tool.invoke(local_file_path)
-            enhanced_question = f"{question_text}\n\n[Audio Transcription: {audio_result}]"
-        elif file_extension in ['xls', 'xlsx', 'csv']:
-            print(f"📊 Processing spreadsheet file: {file_name}")
-            file_content = read_file_tool.invoke(local_file_path)
-            enhanced_question = f"{question_text}\n\n[Spreadsheet Content: {file_content}]"
-        elif file_extension == 'py':
-            print(f"🐍 Processing Python file: {file_name}")
-            code_content = read_file_tool.invoke(local_file_path)
-            enhanced_question = f"{question_text}\n\n[Python Code: {code_content}]"
+
+        # Route based on file type
+        ext = file_name.lower().split('.')[-1]
+        if ext in ['mp3', 'wav', 'm4a', 'flac', 'ogg']:
+            result = audio_processing_tool.invoke(local_file_path)
+            tag = "Audio Transcription"
+        elif ext in ['png', 'jpg', 'jpeg', 'gif', 'bmp']:
+            result = image_recognition_tool.invoke(local_file_path)
+            tag = "Image Analysis"
+        elif ext in ['csv', 'xls', 'xlsx']:
+            result = read_file_tool.invoke(local_file_path)
+            tag = "Spreadsheet Content"
+        elif ext in ['txt', 'md', 'py', 'json']:
+            result = read_file_tool.invoke(local_file_path)
+            tag = "File Content"
         else:
-            print(f"📄 Processing general file: {file_name}")
-            file_content = read_file_tool.invoke(local_file_path)
-            enhanced_question = f"{question_text}\n\n[File Content: {file_content}]"
+            result = read_file_tool.invoke(local_file_path)
+            tag = "File Content"
+
+        # Clean up
         try:
             os.remove(local_file_path)
-            print(f"🧹 Cleaned up temporary file: {local_file_path}")
-        except:
+        except Exception:
             pass
-        return enhanced_question
+
+        return f"{question_text}\n\n[{tag}: {result}]"
+
     except Exception as e:
-        print(f"❌ Error processing file {file_name}: {e}")
         return f"{question_text}\n\n[Note: Error processing attached file {file_name}: {str(e)}]"
 
 
@@ -173,27 +167,29 @@ DECISION MAKING:
 1. First, try to answer from your knowledge if it's a general fact.
 2. If you need specific, current, or detailed information, use serp_search_tool ONCE.
 3. If the question looks reversed (starts with a period), use reverse_text_tool ONCE first.
-4. For file-based questions, use the appropriate file tool.
-5. After using a tool, analyze the result and provide your final answer.
-6. Do NOT cycle between tools unnecessarily.
+4. For factual or historical questions, use wiki_search_tool ONCE.
+5. For file-based questions, use the appropriate file tool.
+6. After using a tool, analyze the result and provide your final answer.
+7. Do NOT cycle between tools unnecessarily.
 
 Tool Use Guidelines:
 1. Do **not** use any tools outside of the provided tools list.
 2. Always use **only one tool at a time** in each step of your execution.
 3. You have a MAXIMUM of 3 tool uses per question.
 4. For web searches and current information, use **serp_search_tool** (15s timeout).
-5. If the question looks reversed (starts with a period or reads backward), first use **reverse_text_tool** to reverse it, then process the question.
-6. For image analysis and description, use **image_recognition_tool** (requires OpenAI API key).
-7. For Python code execution, use **python_execution_tool**.
-8. For video analysis, use **video_analysis_tool**.
-9. For audio processing, use **audio_processing_tool**.
-10. For file type detection, use **file_type_detection_tool**.
-11. For reading file contents, use **read_file_tool**.
-12. File downloading is handled automatically - you don't need to download files manually.
-13. Keep responses concise and efficient.
-14. If you can't find the answer after using 2-3 tools, provide your best estimate based on available information.
-15. NEVER use more than 3 tools for a single question.
-16. After using a tool, provide your final answer immediately.
+5. For factual or historical questions, use **wiki_search_tool**.
+6. If the question looks reversed (starts with a period or reads backward), first use **reverse_text_tool** to reverse it, then process the question.
+7. For image analysis and description, use **image_recognition_tool** (requires OpenAI API key).
+8. For Python code execution, use **python_execution_tool**.
+9. For video analysis, use **video_analysis_tool**.
+10. For audio processing, use **audio_processing_tool**.
+11. For file type detection, use **file_type_detection_tool**.
+12. For reading file contents, use **read_file_tool**.
+13. File downloading is handled automatically - you don't need to download files manually.
+14. Keep responses concise and efficient.
+15. If you can't find the answer after using 2-3 tools, provide your best estimate based on available information.
+16. NEVER use more than 3 tools for a single question.
+17. After using a tool, provide your final answer immediately.
 
 FILE PROCESSING:
 - Questions may come with attached files (mp3, excel, images, etc.)

tools.py

@@ -20,6 +20,7 @@ import openai
 from pydub import AudioSegment
 import pandas as pd
 from PIL import Image
+from langchain_community.document_loaders import WikipediaLoader
 
 # Load environment variables
 print("Current working directory:", os.getcwd())
@@ -526,3 +527,22 @@ math_calculation_tool = Tool(
     func=calculate_simple_math,
     description="Safely evaluates simple mathematical expressions. Use this when you need to perform basic math calculations."
 )
+
+
+def wiki_search(query: str) -> str:
+    """Search Wikipedia for a query and return maximum 2 results."""
+    search_docs = WikipediaLoader(query=query, load_max_docs=2).load()
+    formatted_search_docs = "\n\n---\n\n".join(
+        [
+            f'<Document source="{doc.metadata["source"]}" page="{doc.metadata.get("page", "")}"/>'
+            f"\n{doc.page_content}\n</Document>"
+            for doc in search_docs
+        ])
+    return formatted_search_docs
+
+
+wiki_search_tool = Tool(
+    name="wiki_search_tool",
+    func=wiki_search,
+    description="Search Wikipedia for a query and return up to 2 results. Use this for factual or historical questions."
+)