fix: try simpler tools

app.py

@@ -2,8 +2,10 @@ import os
 import gradio as gr
 import requests
 import inspect
-import time 
+import base64
 import nest_asyncio
+from llama_index.core import SummaryIndex
+from llama_index.readers.web import SimpleWebPageReader
 from llama_index.llms.ollama import Ollama
 from llama_index.tools.wikipedia import WikipediaToolSpec
 from llama_index.readers.youtube_transcript import YoutubeTranscriptReader
@@ -59,21 +61,30 @@ class BasicAgent:
             description="Fetches transcript of the given youtube_link and returns matching answers based on query. To be called when video link is given.",
         )
 
-        def duck_duck_go_search_tool(query: str) -> str:
+        def web_page_reader(url: str, query: str) -> str:
             try:
-                raw_results = DuckDuckGoSearchToolSpec().duckduckgo_full_search(query)
-                texts = [res['body'] for res in raw_results]
-                documents = [Document(text=body) for body in texts]
-
-                splitter = SentenceSplitter(chunk_size=256)
-                nodes = splitter.get_nodes_from_documents(documents)
-
-                retriever = BM25Retriever(nodes=nodes, similarity_top_k=10)
-                synthesizer = get_response_synthesizer(response_mode="tree_summarize", llm=self.llm)
-                query_engine = RetrieverQueryEngine(retriever=retriever, response_synthesizer=synthesizer)
-
+                documents = SimpleWebPageReader(html_to_text=True).load_data(
+                    [url]
+                )
+                index = SummaryIndex.from_documents(documents)
+                query_engine = index.as_query_engine()
                 response = query_engine.query(query)
                 return response.response
+            except Exception as e:
+                print("error in webpage", e)
+
+        web_page_reader_tool = FunctionTool.from_defaults(
+            web_page_reader,
+            name="web_page_reader",
+            description="Visits the wepage on given url and returns response on the passed query"
+        )
+
+        def duck_duck_go_search_tool(query: str) -> str:
+            try:
+                raw_results = DuckDuckGoSearchToolSpec().duckduckgo_full_search(query, max_results=5)
+                texts = [res['body'] for res in raw_results]
+                full_text = "\n".join(texts)
+                return "Here's the responses from the search: " + full_text
 
             except Exception as e:
                 return f"An error occurred: {e}"
@@ -85,20 +96,14 @@ class BasicAgent:
                 description="Searches the web and refines the result into a high-quality answer. Use when other tools don't seem suitable"
             )
         
-        def wikipedia_search(query: str) -> str:
+        def wikipedia_search(page_title: str, query: str) -> str:
             try:
-                text = WikipediaToolSpec().search_data(query)
-                documents = [Document(text=text)]
+                text = WikipediaToolSpec().load_data(page=page_title)
 
-                splitter = SentenceSplitter(chunk_size=256)
-                nodes = splitter.get_nodes_from_documents(documents)
-
-                retriever = BM25Retriever(nodes=nodes, similarity_top_k=1)
-                synthesizer = get_response_synthesizer(response_mode="tree_summarize", llm=self.llm)
-                query_engine = RetrieverQueryEngine(retriever=retriever, response_synthesizer=synthesizer)
-
-                response = query_engine.query(query)
-                return response.response
+                if text == "":
+                    text = WikipediaToolSpec().search_data(query)
+                    
+                return "Here's the response from wikipedia search: " + text
 
             except Exception as e:
                 return f"An error occurred: {e}"
@@ -110,7 +115,7 @@ class BasicAgent:
                 description="Searches wikipedia and converts results into a high quality answer."
             )
 
-        self.agent = AgentWorkflow.from_tools_or_functions([duckduckgo_search_tool, youtube_transcript_answer_tool, wikipedia_search_tool], llm=self.llm)
+        self.agent = AgentWorkflow.from_tools_or_functions([duckduckgo_search_tool, youtube_transcript_answer_tool, wikipedia_search_tool, web_page_reader_tool], llm=self.llm)
         print("BasicAgent initialized.")
 
     async def run_agent(self, question: str):
@@ -143,6 +148,7 @@ async def run_and_submit_all( profile: gr.OAuthProfile | None):
     api_url = DEFAULT_API_URL
     questions_url = f"{api_url}/questions"
     submit_url = f"{api_url}/submit"
+    files_url = f"{api_url}/files/"
 
     # 1. Instantiate Agent ( modify this part to create your agent)
     try:
@@ -187,7 +193,19 @@ async def run_and_submit_all( profile: gr.OAuthProfile | None):
             print(f"Skipping item with missing task_id or question: {item}")
             continue
         try:
-            submitted_answer = agent(question_text)
+            encoded = None
+            if item.get("file_name") != "":
+                response = requests.get(files_url + task_id)
+                response.raise_for_status()
+
+                data = response.content
+
+                encoded = base64.b64encode(data).decode('utf-8')
+            
+            if encoded is not None:
+                submitted_answer = agent(question_text + "\nfile_data: " + encoded)
+            else:
+                submitted_answer = agent(question_text)
             answers_payload.append({"task_id": task_id, "submitted_answer": submitted_answer})
             results_log.append({"Task ID": task_id, "Question": question_text, "Submitted Answer": submitted_answer})
         except Exception as e: