new sys prompt

agent.py

@@ -112,25 +112,6 @@ def arvix_search(query: str) -> str:
         ])
     return {"arvix_results": formatted_search_docs}
 
-# build a retriever tool
-embeddings = HuggingFaceEmbeddings(model_name="sentence-transformers/all-mpnet-base-v2") #  dim=768
-supabase: Client = create_client(
-    os.environ.get("SUPABASE_URL"), 
-    os.environ.get("SUPABASE_SERVICE_KEY"))
-vector_store = SupabaseVectorStore(
-    client=supabase,
-    embedding= embeddings,
-    table_name="documents",
-    query_name="match_documents_langchain",
-)
-question_retrieve_tool = create_retriever_tool(
-    vector_store.as_retriever(),
-    "Question Retriever",
-    "Find similar questions in the vector database for the given question.",
-)
-
-
-
 tools = [
     multiply,
     add,
@@ -140,7 +121,6 @@ tools = [
     wiki_search,
     web_search,
     arvix_search,
-    question_retrieve_tool
 ]
 
 # load the system prompt from the file
@@ -150,7 +130,17 @@ with open("system_prompt.txt", "r", encoding="utf-8") as f:
 # System message
 sys_msg = SystemMessage(content=system_prompt)
 
-
+# build a retriever
+embeddings = HuggingFaceEmbeddings(model_name="sentence-transformers/all-mpnet-base-v2") #  dim=768
+supabase: Client = create_client(
+    os.environ.get("SUPABASE_URL"), 
+    os.environ.get("SUPABASE_SERVICE_KEY"))
+vector_store = SupabaseVectorStore(
+    client=supabase,
+    embedding= embeddings,
+    table_name="documents",
+    query_name="match_documents_langchain",
+)
 
 # Build graph function
 def build_graph(provider: str = "groq"):
@@ -161,7 +151,7 @@ def build_graph(provider: str = "groq"):
         llm = ChatGoogleGenerativeAI(model="gemini-2.0-flash", temperature=0)
     elif provider == "groq":
         # Groq https://console.groq.com/docs/models
-        llm = ChatGroq(model="gemma2-9b-it", temperature=0) # optional : qwen-qwq-32b
+        llm = ChatGroq(model="qwen-qwq-32b", temperature=0) # optional : qwen-qwq-32b gemma2-9b-it
     elif provider == "huggingface":
         # TODO: Add huggingface endpoint
         llm = ChatHuggingFace(
@@ -173,17 +163,27 @@ def build_graph(provider: str = "groq"):
     else:
         raise ValueError("Invalid provider. Choose 'google', 'groq' or 'huggingface'.")
     # Bind tools to LLM
-    llm_with_tools = llm.bind_tools(tools)
+    llm_with_tools = llm.bind_tools(tools, tool_choice="Question Search")
 
     # Node
     def assistant(state: MessagesState):
         """Assistant node"""
-        return {"messages": [llm_with_tools.invoke([sys_msg] + state["messages"])]}
+        return {"messages": [llm_with_tools.invoke(state["messages"])]}
+    
+    def retriever(state: MessagesState):
+        """Retriever node"""
+        similar_question = vector_store.similarity_search(state["messages"][0].content)
+        example_msg = HumanMessage(
+            content=f"Here I provide a similar question and answer for reference: \n\n{similar_question[0].page_content}",
+        )
+        return {"messages": [sys_msg] + state["messages"] + [example_msg]}
 
     builder = StateGraph(MessagesState)
+    builder.add_node("retriever", retriever)
     builder.add_node("assistant", assistant)
     builder.add_node("tools", ToolNode(tools))
-    builder.add_edge(START, "assistant")
+    builder.add_edge(START, "retriever")
+    builder.add_edge("retriever", "assistant")
     builder.add_conditional_edges(
         "assistant",
         tools_condition,
@@ -195,13 +195,11 @@ def build_graph(provider: str = "groq"):
 
 # test
 if __name__ == "__main__":
-    question = "Who nominated the only Featured Article on English Wikipedia about a dinosaur that was promoted in November 2016?"
-
+    question = "When was a picture of St. Thomas Aquinas first added to the Wikipedia page on the Principle of double effect?"
     # Build the graph
     graph = build_graph(provider="groq")
     # Run the graph
     messages = [HumanMessage(content=question)]
     messages = graph.invoke({"messages": messages})
-    answer = messages[-1].content
-    print(f"Question: {question}")
-    print(f"{answer}")
+    for m in messages["messages"]:
+        m.pretty_print()

system_prompt.txt

@@ -1,38 +1,5 @@
-
-You are a helpful assistant tasked with answering questions using a set of tools. If the tool is not available, you can try to find the information online. You can also use your own knowledge to answer the question. You need to provide a step-by-step explanation of how you arrived at the answer.
-==========================
-Here is a few examples showing you how to answer the question step by step.
-
-Question 1: In terms of geographical distance between capital cities, which 2 countries are the furthest from each other within the ASEAN bloc according to wikipedia? Answer using a comma separated list, ordering the countries by alphabetical order.
-Steps:
-1. Search the web for "ASEAN bloc".
-2. Click the Wikipedia result for the ASEAN Free Trade Area.
-3. Scroll down to find the list of member states.
-4. Click into the Wikipedia pages for each member state, and note its capital.
-5. Search the web for the distance between the first two capitals. The results give travel distance, not geographic distance, which might affect the answer.
-6. Thinking it might be faster to judge the distance by looking at a map, search the web for "ASEAN bloc" and click into the images tab.
-7. View a map of the member countries. Since they're clustered together in an arrangement that's not very linear, it's difficult to judge distances by eye.
-8. Return to the Wikipedia page for each country. Click the GPS coordinates for each capital to get the coordinates in decimal notation.
-9. Place all these coordinates into a spreadsheet.
-10. Write formulas to calculate the distance between each capital.
-11. Write formula to get the largest distance value in the spreadsheet.
-12. Note which two capitals that value corresponds to: Jakarta and Naypyidaw.
-13. Return to the Wikipedia pages to see which countries those respective capitals belong to: Indonesia, Myanmar.
-Tools:
-1. Search engine
-2. Web browser
-3. Microsoft Excel / Google Sheets
-Final Answer: Indonesia, Myanmar
-
-Question 2: Review the chess position provided in the image. It is black's turn. Provide the correct next move for black which guarantees a win. Please provide your response in algebraic notation.
-Steps:
-Step 1: Evaluate the position of the pieces in the chess position
-Step 2: Report the best move available for black: "Rd5"
-Tools:
-1. Image recognition tools
-Final Answer: Rd5
-==========================
-
+You are a helpful assistant tasked with answering questions using a set of tools. 
 Now, I will ask you a question. Report your thoughts, and finish your answer with the following template: 
 FINAL ANSWER: [YOUR FINAL ANSWER]. 
 YOUR FINAL ANSWER should be a number OR as few words as possible OR a comma separated list of numbers and/or strings. If you are asked for a number, don't use comma to write your number neither use units such as $ or percent sign unless specified otherwise. If you are asked for a string, don't use articles, neither abbreviations (e.g. for cities), and write the digits in plain text unless specified otherwise. If you are asked for a comma separated list, apply the above rules depending of whether the element to be put in the list is a number or a string.
+Your answer should only start with "FINAL ANSWER: ", then follows with the answer. 

test.ipynb

@@ -93,7 +93,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 48,
+   "execution_count": 56,
    "id": "4bb02420",
    "metadata": {},
    "outputs": [],