agent.py

import os
from dotenv import load_dotenv
from langgraph.prebuilt import ToolNode, tools_condition
from langchain_core.messages import SystemMessage, HumanMessage
from langchain_community.document_loaders import WikipediaLoader, ArxivLoader
from langchain_community.tools.tavily_search import TavilySearchResults
from langchain.tools.retriever import create_retriever_tool
from langchain_core.tools import tool
from supabase.client import Client, create_client
from langchain_community.vectorstores import SupabaseVectorStore
from langchain_huggingface import (
    ChatHuggingFace,
    HuggingFaceEndpoint,
    HuggingFaceEmbeddings,
)
from langgraph.graph import START, StateGraph, MessagesState

load_dotenv()


@tool
def wikipedia_search(query: str) -> str:
    """Search Wikipedia for a query and return maximum 2 results
    Args:
        query: The search string
    """
    docs = WikipediaLoader(query=query, load_max_docs=2).load()
    all_search_docs = "\n\n---\n\n".join(
        [
            f'<Document source="{doc.metadata["source"]}" page="{doc.metadata.get("page", "")}"/>\n{doc.page_content}\n</Document>'
            for doc in docs
        ]
    )
    return {"wikipedia_results": all_search_docs}


@tool
def web_search(query: str) -> str:
    """Search Tavily for a query and return maximum 3 results.

    Args:
        query: The search query."""
    docs = TavilySearchResults(max_results=3).invoke(query=query)
    all_search_docs = "\n\n---\n\n".join(
        [
            f'<Document source="{doc.metadata["source"]}" page="{doc.metadata.get("page", "")}"/>\n{doc.page_content}\n</Document>'
            for doc in docs
        ]
    )
    return {"web_results": all_search_docs}


@tool
def arvix_search(query: str) -> str:
    """Search Arxiv for a query and return maximum 3 result.

    Args:
        query: The search query."""
    search_docs = ArxivLoader(query=query, load_max_docs=3).load()
    formatted_search_docs = "\n\n---\n\n".join(
        [
            f'<Document source="{doc.metadata["source"]}" page="{doc.metadata.get("page", "")}"/>\n{doc.page_content[:1000]}\n</Document>'
            for doc in search_docs
        ]
    )
    return {"arvix_results": formatted_search_docs}


with open("system_prompt.txt", "r", encoding="utf-8") as f:
    system_prompt = f.read()

sys_msg = SystemMessage(system_prompt)
supabase: Client = create_client(
    os.environ.get("SUPABASE_URL"), os.environ.get("SUPABASE_SERVICE_KEY")
)

supabase_store = SupabaseVectorStore(
    client=supabase,
    embedding=HuggingFaceEmbeddings(
        model_name="sentence-transformers/all-mpnet-base-v2"
    ),
    table_name="search_documents",
    query_name="langchain_match_documents",
)

retriever_tool = create_retriever_tool(
    retriever=supabase_store.as_retriever(
        search_type="similarity", search_kwargs={"k": 5}
    ),
    name="question_search",
    description="A tool to retrieve similar questions from a vector store.",
)

tools = [
    wikipedia_search,
    web_search,
    arvix_search,
    retriever_tool,
]


def build_graph():
    llm = ChatHuggingFace(
        llm=HuggingFaceEndpoint(repo_id="Qwen/Qwen2.5-Coder-32B-Instruct"),
    )

    llm_with_tools = llm.bind_tools(tools)

    def assistant(state: MessagesState):
        """Assistant node"""
        return {"messages": [llm_with_tools.invoke(state["messages"])]}

    def retriever(state: MessagesState):
        """Retriever node"""
        similar_question = supabase_store.similarity_search(
            state["messages"][0].content
        )
        print("Similar questions:")
        print(similar_question)
        if len(similar_question) > 0:
            example_msg = HumanMessage(
                content=f"Here I provide a similar question and answer for reference: \n\n{similar_question[0].page_content}",
            )
            # return {"messages": [{"role": "system", "content": similar_question[0].page_content}]}
            return {"messages": [sys_msg] + state["messages"] + [example_msg]}
        return {"messages": [sys_msg] + state["messages"]}

    builder = StateGraph(MessagesState)
    builder.add_node("retriever", retriever)
    builder.add_node("assistant", assistant)
    builder.add_node("tools", ToolNode(tools))
    builder.add_edge(START, "retriever")
    builder.add_edge("retriever", "assistant")
    builder.add_conditional_edges(
        "assistant",
        tools_condition,
    )
    builder.add_edge("tools", "assistant")

    return builder.compile()