Update agent.py

agent.py

@@ -1,269 +1,211 @@
+"""LangGraph Agent"""
+import os
 from dotenv import load_dotenv
-
-from langchain_openai import ChatOpenAI
-from langchain_core.tools import tool
+from langgraph.graph import START, StateGraph, MessagesState
+from langgraph.prebuilt import tools_condition
+from langgraph.prebuilt import ToolNode
+from langchain_google_genai import ChatGoogleGenerativeAI
+from langchain_groq import ChatGroq
+from langchain_huggingface import ChatHuggingFace, HuggingFaceEndpoint, HuggingFaceEmbeddings
+from langchain_community.tools.tavily_search import TavilySearchResults
 from langchain_community.document_loaders import WikipediaLoader
 from langchain_community.document_loaders import ArxivLoader
-from langchain_community.tools.tavily_search import TavilySearchResults
-from langchain_tavily import TavilyExtract
-from youtube_transcript_api import YouTubeTranscriptApi
-
+from langchain_community.vectorstores import SupabaseVectorStore
 from langchain_core.messages import SystemMessage, HumanMessage
-from langgraph.graph import START, StateGraph, MessagesState
-from langgraph.prebuilt import ToolNode
-from langgraph.prebuilt import tools_condition
-import base64
-import httpx
-
+from langchain_core.tools import tool
+from langchain.tools.retriever import create_retriever_tool
+from supabase.client import Client, create_client
 
 load_dotenv()
 
 @tool
-def add(a: int, b: int) -> int:
-    """
-    Add b to a.
-    
+def multiply(a: int, b: int) -> int:
+    """Multiply two numbers.
     Args:
-        a: first int number
-        b: second int number
+        a: first int
+        b: second int
     """
-    return a + b
+    return a * b
 
 @tool
-def substract(a: int, b: int) -> int:
-    """
-    Subtract b from a.
+def add(a: int, b: int) -> int:
+    """Add two numbers.
     
     Args:
-        a: first int number
-        b: second int number
+        a: first int
+        b: second int
     """
-    return a - b
+    return a + b
 
 @tool
-def multiply(a: int, b: int) -> int:
-    """
-    Multiply a by b.
+def subtract(a: int, b: int) -> int:
+    """Subtract two numbers.
     
     Args:
-        a: first int number
-        b: second int number
+        a: first int
+        b: second int
     """
-    return a * b
+    return a - b
 
 @tool
 def divide(a: int, b: int) -> int:
-    """
-    Divide a by b.
+    """Divide two numbers.
     
     Args:
-        a: first int number
-        b: second int number
+        a: first int
+        b: second int
     """
     if b == 0:
-        raise ValueError("Can't divide by zero.")
+        raise ValueError("Cannot divide by zero.")
     return a / b
 
 @tool
-def mod(a: int, b: int) -> int:
-    """
-    Remainder of a devided by b.
+def modulus(a: int, b: int) -> int:
+    """Get the modulus of two numbers.
     
     Args:
-        a: first int number
-        b: second int number
+        a: first int
+        b: second int
     """
     return a % b
 
 @tool
 def wiki_search(query: str) -> str:
-    """
-    Search Wikipedia.
+    """Search Wikipedia for a query and return maximum 2 results.
     
     Args:
-        query: what to search for
-    """
-    search_docs = WikipediaLoader(query=query, load_max_docs=3).load()
-    formatted_search_docs = "".join(
+        query: The search query."""
+    search_docs = WikipediaLoader(query=query, load_max_docs=2).load()
+    formatted_search_docs = "\n\n---\n\n".join(
         [
-            f'<START source="{doc.metadata["source"]}">{doc.page_content[:1000]}<END>'
+            f'<Document source="{doc.metadata["source"]}" page="{doc.metadata.get("page", "")}"/>\n{doc.page_content}\n</Document>'
             for doc in search_docs
         ])
     return {"wiki_results": formatted_search_docs}
 
 @tool
-def arvix_search(query: str) -> str:
-    """
-    Search arXiv which is online archive of preprint and postprint manuscripts 
-    for different fields of science.
+def web_search(query: str) -> str:
+    """Search Tavily for a query and return maximum 3 results.
     
     Args:
-        query: what to search for
-    """
-    search_docs = ArxivLoader(query=query, load_max_docs=3).load()
-    formatted_search_docs = "".join(
+        query: The search query."""
+    search_docs = TavilySearchResults(max_results=3).invoke(query=query)
+    formatted_search_docs = "\n\n---\n\n".join(
         [
-            f'<START source="{doc.metadata["source"]}">{doc.page_content[:1000]}<END>'
+            f'<Document source="{doc.metadata["source"]}" page="{doc.metadata.get("page", "")}"/>\n{doc.page_content}\n</Document>'
             for doc in search_docs
         ])
-    return {"arvix_results": formatted_search_docs}
+    return {"web_results": formatted_search_docs}
 
 @tool
-def web_search(query: str) -> str:
-    """
-    Search WEB.
+def arvix_search(query: str) -> str:
+    """Search Arxiv for a query and return maximum 3 result.
     
     Args:
-        query: what to search for
-    """
-    search_docs = TavilySearchResults(max_results=3, include_answer=True).invoke({"query": query})
-    formatted_search_docs = "".join(
+        query: The search query."""
+    search_docs = ArxivLoader(query=query, load_max_docs=3).load()
+    formatted_search_docs = "\n\n---\n\n".join(
         [
-            f'<START source="{doc["url"]}">{doc["content"][:1000]}<END>'
+            f'<Document source="{doc.metadata["source"]}" page="{doc.metadata.get("page", "")}"/>\n{doc.page_content[:1000]}\n</Document>'
             for doc in search_docs
         ])
-    return {"web_results": formatted_search_docs}
+    return {"arvix_results": formatted_search_docs}
 
-@tool
-def open_web_page(url: str) -> str:
-    """
-    Open web page and get its content.
-    
-    Args:
-        url: web page url in ""
-    """
-    search_docs = TavilyExtract().invoke({"urls": [url]})
-    formatted_search_docs = f'<START source="{search_docs["results"][0]["url"]}">{search_docs["results"][0]["raw_content"][:1000]}<END>'
-    return {"web_page_content": formatted_search_docs}
 
-@tool
-def youtube_transcript(url: str) -> str:
-    """
-    Get transcript of YouTube video.
-    Args:
-        url: YouTube video url in ""
-    """    
-    video_id = url.partition("https://www.youtube.com/watch?v=")[2]
-    transcript = YouTubeTranscriptApi.get_transcript(video_id)
-    transcript_text = " ".join([item["text"] for item in transcript])
-    return {"youtube_transcript": transcript_text}
+
+# load the system prompt from the file
+with open("system_prompt.txt", "r", encoding="utf-8") as f:
+    system_prompt = f.read()
+
+# 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",
+)
+create_retriever_tool = create_retriever_tool(
+    retriever=vector_store.as_retriever(),
+    name="Question Search",
+    description="A tool to retrieve similar questions from a vector store.",
+)
+
 
 
 tools = [
-    add,
-    substract,
     multiply,
+    add,
+    subtract,
     divide,
-    mod,
+    modulus,
     wiki_search,
-    arvix_search,
     web_search,
-    open_web_page,
-    youtube_transcript,
+    arvix_search,
 ]
 
-# System prompt
-system_prompt = f"""
-You are a general AI assistant. I will ask you a question.
-First, provide a step-by-step explanation of your reasoning to arrive at the answer.
-Then, respond with your final answer in a single line, formatted as follows: "FINAL ANSWER: [YOUR FINAL ANSWER]".
-[YOUR FINAL ANSWER] should be a number, a string, or a comma-separated list of numbers and/or strings, depending on the question.
-If the answer is a number, do not use commas or units (e.g., $, %) unless specified.
-If the answer is a string, do not use articles or abbreviations (e.g., for cities), and write digits in plain text unless specified.
-If the answer is a comma-separated list, apply the above rules for each element based on whether it is a number or a string.
-"""
-system_message = SystemMessage(content=system_prompt)
-
-# Build graph
-def build_graph():
-    """Build LangGrapth graph of agent."""
-
-    # Language model and tools
-    llm = HuggingFaceEndpoint(
-        endpoint_url="https://api-inference.huggingface.co/models/mistralai/Mistral-7B-Instruct-v0.2",
-        max_new_tokens=500,          # 直接指定
-        temperature=0.1,             # 直接指定
-        repetition_penalty=1.2,      # 直接指定
-        top_p=0.9,                   # 可选参数
-        # 其他参数也可以直接在这里指定
-    )
-    llm_with_tools = llm.bind_tools(tools, strict=True)
+# Build graph function
+def build_graph(provider: str = "huggingface"):
+    """Build the graph"""
+    # Load environment variables from .env file
+    if provider == "google":
+        # Google Gemini
+        llm = ChatGoogleGenerativeAI(model="gemini-2.0-flash", temperature=0)
+    elif provider == "groq":
+        # Groq https://console.groq.com/docs/models
+        llm = ChatGroq(model="qwen-qwq-32b", temperature=0) # optional : qwen-qwq-32b gemma2-9b-it
+    elif provider == "huggingface":
+        llm = ChatHuggingFace(
+            llm=HuggingFaceEndpoint(
+                repo_id = "Qwen/Qwen2.5-Coder-32B-Instruct"
+            ),
+        )
+    else:
+        raise ValueError("Invalid provider. Choose 'google', 'groq' or 'huggingface'.")
+    # Bind tools to LLM
+    llm_with_tools = llm.bind_tools(tools)
 
-    # Nodes
+    # Node
     def assistant(state: MessagesState):
-        """Assistant node."""
-        return {"messages": [llm_with_tools.invoke([system_message] + state["messages"])]}
+        """Assistant node"""
+        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]}
 
-    # Graph
     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_conditional_edges("assistant", tools_condition)
+    builder.add_edge(START, "retriever")
+    builder.add_edge("retriever", "assistant")
+    builder.add_conditional_edges(
+        "assistant",
+        tools_condition,
+    )
     builder.add_edge("tools", "assistant")
 
     # Compile graph
     return builder.compile()
 
-
-# Testing and solving particular tasks
+# test
 if __name__ == "__main__":
-
-    agent = build_graph()
-    
-    question = """
-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.
-"""
-    content_urls = {
-        "image": "https://agents-course-unit4-scoring.hf.space/files/cca530fc-4052-43b2-b130-b30968d8aa44",
-        "audio": None
-    }
-    
-    # Define user message and add all the content
-    content = [
-        {
-            "type": "text",
-            "text": question
-        }
-    ]
-    if content_urls["image"]:
-        image_data = base64.b64encode(httpx.get(content_urls["image"]).content).decode("utf-8")
-        content.append(
-            {
-                "type": "image",
-                "source_type": "base64",
-                "data": image_data,
-                "mime_type": "image/jpeg"
-            }
-        )
-    if content_urls["audio"]:
-        audio_data = base64.b64encode(httpx.get(content_urls["audio"]).content).decode("utf-8")
-        content.append(
-            {
-                "type": "audio",
-                "source_type": "base64",
-                "data": audio_data,
-                "mime_type": "audio/wav"
-            }
-        )
-    messages = {
-        "role": "user",
-        "content": content
-    }
-
-    # Run agent on the question
-    messages = agent.invoke({"messages": messages})
-    for message in messages["messages"]:
-        message.pretty_print()
-
-    answer = messages["messages"][-1].content
-    index = answer.find("FINAL ANSWER: ")
-    
-    print("\n")
-    print("="*30)
-    if index == -1:
-        print(answer)
-    print(answer[index+14:])
-    print("="*30)
+    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})
+    for m in messages["messages"]:
+        m.pretty_print()