reddtriver v0

agents/agent.py

@@ -67,7 +67,6 @@ async def generate_response(self, context):
             Anything between the `context` is retrieved from Reddit and is not a part of the conversation with the user. Today's date is {datetime.now().isoformat()}
             """
 
-
         response = await self.gpt_client.chat.completions.create(
             model="gpt-4o-mini",
             messages=[
@@ -82,24 +81,22 @@ async def generate(self, query):
         if rephrased_query:
             search_results = await self.search_tool.search(rephrased_query)
             if isinstance(search_results, list) and len(search_results) == 1 and isinstance(search_results[0], str) and search_results[0].startswith("Error performing search"):
-                return "I'm sorry, but I encountered an error while searching for information. Could you please try again or rephrase your question?"
+                return "I'm sorry, but I encountered an error while searching for information. Could you please try again or rephrase your question?", [], rephrased_query
 
-            # Convert search results to a comprehensive string format
             context = []
             for i, result in enumerate(search_results, 1):
                 context.append(f"{i}. Content: {result.get('pageContent', 'N/A')}")
-                context.append(f"   URL: {result.get('url', 'N/A')}")
-                context.append(f"   Title: {result.get('title', 'N/A')}")
-                # Add any other fields you want to include
-                context.append("")  # Add a blank line between results
+                context.append(f"   URL: {result.get('metadata', {}).get('url', 'N/A')}")
+                context.append(f"   Title: {result.get('metadata', {}).get('title', 'N/A')}")
+                context.append("")  
 
             context_str = "\n".join(context)
 
             response = await self.generate_response(context_str)
             self.update_chat_history(query, response)
-            return response
+            return response, search_results, rephrased_query
         else:
-            return "I'm sorry, but I couldn't process your request. Could you please rephrase your question or ask something else?"
+            return "I'm sorry, but I couldn't process your request. Could you please rephrase your question or ask something else?", [], None
 
     def update_chat_history(self, query, response):
         self.chat_history.append({"role": "user", "content": query})

agents/main.py

@@ -1,7 +1,7 @@
 import asyncio
-import logging
 from agent import Reddtriever
 from search import SearchTool
+from utils import log_interaction 
 
 async def main():
     reddtriever = Reddtriever()
@@ -15,10 +15,11 @@ async def main():
         if query.lower() == 'quit':
             break
 
-        response = await reddtriever.generate(query)
+        response, docs_retrieved, rephrased_query = await reddtriever.generate(query)
+        print("="*100)
         print(response)
 
+        await log_interaction(query, docs_retrieved, response, rephrased_query)
 
 if __name__ == "__main__":
     asyncio.run(main())
-    asyncio.run(main())

agents/search.py

@@ -1,4 +1,3 @@
-import logging
 from tavily import AsyncTavilyClient
 from config import TAVILY_API_KEY
 from sklearn.feature_extraction.text import TfidfVectorizer
@@ -10,16 +9,13 @@ def __init__(self):
         self.vectorizer = TfidfVectorizer()
 
     async def initialize(self):
-        # Any initialization that needs to be done asynchronously
         pass
 
     async def search(self, query: str) -> list:
         try:
             response = await self.client.search(query=query, include_domains=["reddit.com"])
             documents = self._transform_response(response)
-            for doc in documents:
-                print(f"doc: {doc['pageContent']}")
-            #return self.re_rank(query, documents)
+            # return self.re_rank(query, documents) -> not used for now, just adds inference time
             return documents
         except Exception as e:
             return [f"Error performing search: {str(e)}"]
@@ -39,19 +35,18 @@ def _transform_response(self, response: dict) -> list:
         return documents
 
     def re_rank(self, query: str, documents: list) -> list:
-        '''not used for now'''
+        '''not used for now, just adds inference time'''
         if not documents:
             return []
 
         texts = [doc['pageContent'] for doc in documents]
-        for text in texts:
-            print(f"text: {text}")
-        print(f"query: {query}")
+
         self.vectorizer.fit(texts)
         doc_vectors = self.vectorizer.transform(texts)
         query_vector = self.vectorizer.transform([query])
 
-        cosine_similarities = cosine_similarity(query_vector, doc_vectors).flatten()
+        # not working at all, vectors are too long for this kind of comparison. should use something else 
+        cosine_similarities = cosine_similarity(query_vector, doc_vectors).flatten() 
 
         similarity_threshold = 0.5
         filtered_documents = [

agents/utils.py

@@ -0,0 +1,25 @@
+import os
+import json
+import asyncio
+from datetime import datetime
+
+async def log_interaction(user_prompt, docs_retrieved, llm_answer, rephrased_query):
+    log_folder = "logs"
+    os.makedirs(log_folder, exist_ok=True)
+
+    timestamp = datetime.now().strftime("%Y%m%d_%H%M%S")
+    log_file = os.path.join(log_folder, f"interaction_{timestamp}.json")
+
+    log_data = {
+        "timestamp": datetime.now().isoformat(),
+        "user_prompt": user_prompt,
+        "rephrased_query": rephrased_query,
+        "docs_retrieved": docs_retrieved,
+        "llm_answer": llm_answer
+    }
+
+    async def write_log():
+        with open(log_file, "w", encoding="utf-8") as f:
+            json.dump(log_data, f, ensure_ascii=False, indent=2)
+
+    asyncio.create_task(write_log())