wikipedia_search_tool

helper.py

@@ -266,4 +266,60 @@ audio_transcriber_tool = Tool(
         "Returns the transcribed text or an error message if transcription fails."
     ),
     func=transcribe_audio_from_path_or_url,
-)
+)
+
+from langchain_community.tools import WikipediaQueryRun
+from langchain_community.utilities import WikipediaAPIWrapper
+from langchain.tools import Tool # Ensure Tool is imported
+
+def get_wikipedia_tool() -> Tool:
+    """
+    Creates and returns a LangChain Tool for querying Wikipedia.
+    """
+    # Initialize the Wikipedia API Wrapper
+    # Configure the wrapper to control how much content is returned.
+    # 'top_k_results': Number of search results to return.
+    # 'doc_content_chars_max': Maximum number of characters from the page content.
+    #                        Setting this to a reasonable limit (e.g., 2000-4000)
+    #                        is crucial for token management, as Wikipedia articles can be very long.
+    #                        Adjust based on your LLM's context window and typical use cases.
+    wikipedia_api_wrapper = WikipediaAPIWrapper(
+        top_k_results=3,             # Get up to 3 most relevant articles
+        doc_content_chars_max=2000   # Limit content to 2000 characters per article
+    )
+
+    # Initialize the Wikipedia tool with the wrapper
+    wikipedia_query_tool = WikipediaQueryRun(api_wrapper=wikipedia_api_wrapper)
+
+    return Tool(
+        name="wikipedia_search_tool",
+        description="""
+        A specialized search tool for retrieving information from Wikipedia.
+        Use this tool when you need:
+        - Authoritative and well-established facts.
+        - Historical information, biographies, or scientific explanations.
+        - General knowledge about specific concepts, people, places, or events.
+        - Background information that is unlikely to be very recent.
+
+        **Input Format (CRITICAL):**
+        The input MUST be a concise and clear query string representing the topic you want to search on Wikipedia.
+        Think of a noun phrase or a short question that directly names the subject.
+        Example: "Albert Einstein"
+        Example: "Battle of Gettysburg"
+        Example: "photosynthesis process"
+
+        **DO NOT:**
+        - Ask natural language questions that are not search queries.
+        - Seek real-time information (e.g., current news, today's weather). For real-time data, use 'tavily_search'.
+        - Provide incomplete sentences or ambiguous terms.
+        - Expect this tool to perform calculations or access external websites beyond Wikipedia.
+
+        **Output:**
+        The tool returns a string containing snippets of relevant Wikipedia articles.
+        The output is limited in length to save tokens. If the answer is not found in the snippet,
+        you might need to refine your query or consider using a broader search tool like 'tavily_search'.
+        """,
+        func=wikipedia_query_tool.run,
+    )
+
+wikipedia_search_tool = get_wikipedia_tool()