workflow.py

# https://langchain-ai.github.io/langgraph/how-tos/memory/manage-conversation-history/#build-the-agent
# https://docs.tavily.com/docs/rest-api/api-reference

import os
from langchain_core.tools import tool
from langgraph.checkpoint.memory import MemorySaver
from langgraph.graph import MessagesState, StateGraph, START, END
from langgraph.prebuilt import ToolNode
from langchain_community.tools.tavily_search import TavilySearchResults
from langchain_core.messages import SystemMessage, HumanMessage, AIMessage


import logging
logger = logging.getLogger(__name__)
logger.setLevel(logging.INFO)

memory = MemorySaver()


dsa_search_domains = [
    # DSA
    'algs4.cs.princeton.edu',
    'chalmersgu-data-structure-courses.github.io',
    'pressbooks.palni.org/anopenguidetodatastructuresandalgorithms',
    'en.wikibooks.org/wiki/Algorithms',
    'people.mpi-inf.mpg.de',
    'jeffe.cs.illinois.edu/teaching/algorithms',
    'opendatastructures.org',
    'github.com/aibooks14',
    'open.umn.edu/opentextbooks',
    'opendsa-server.cs.vt.edu/OpenDSA/Books',
    'www.programiz.com'
    # Discrete Math
    'discrete.openmathbooks.org',
    'stephendavies.org',
    'www.fecundity.com',
    'ocw.mit.edu',
    'discretemath.org',
    'www.khanacademy.org',
    # More general
    'www.w3schools.com',
    'www.geeksforgeeks.org',
    'leetcode.com',
    'www.hackerrank.com',
    'www.freecodecamp.org',
    'www.codechef.com',
    'www.w3resource.com',
    'www.hackerearth.com',
    'openstax.org'
]

tavily_search = TavilySearchResults(max_results=8, verbose=False, include_domains=dsa_search_domains)
tools = [tavily_search]
tool_node = ToolNode(tools)

# from langchain_openai import ChatOpenAI
# logger.debug(f"LLM model: {os.getenv('OPENAI_MODEL_NAME')}")
# llm = ChatOpenAI(model=os.getenv("OPENAI_MODEL_NAME"))

# from huggingface_hub import InferenceClient
# llm = InferenceClient("HuggingFaceH4/zephyr-7b-beta")
# DOESN'T support bind_tools
# from langchain_huggingface import HuggingFaceEndpoint
# llm = HuggingFaceEndpoint(
#     repo_id="HuggingFaceH4/zephyr-7b-beta",
#     max_length=128,
#     temperature=0.5,
#     huggingfacehub_api_token=os.getenv('HUGGINGFACEHUB_API_TOKEN'),
# )

from langchain_groq import ChatGroq
logger.debug(f"LLM model: {os.getenv('GROQ_MODEL')}")
llm = ChatGroq(model_name=os.getenv('GROQ_MODEL'), temperature=0.1)

# from langchain_mistralai import ChatMistralAI
# logger.debug(f"LLM model: {os.getenv('MISTRAL_MODEL_NAME')}")
# llm = ChatMistralAI(
#     model=os.getenv('MISTRAL_MODEL_NAME'),
#     temperature=0,
#     max_retries=2,
# )
bound_model = llm.bind_tools(tools)


def should_continue(state: MessagesState):
    """Return the next node to execute."""
    last_message = state["messages"][-1]
    logger.debug(f'***should_continue*** : last_message = {last_message}')
    if not last_message.tool_calls:
        return END
    return "action"


def search_agent(state: MessagesState):
    response = bound_model.invoke(state["messages"])
    return {"messages": response}


# Define a new graph
workflow = StateGraph(MessagesState)

# Define the two nodes we will cycle between
workflow.add_node("agent", search_agent)
workflow.add_node("action", tool_node)

# Set the entrypoint as `agent`
# This means that this node is the first one called
workflow.add_edge(START, "agent")

# We now add a conditional edge
workflow.add_conditional_edges(
    "agent",
    should_continue,
    ["action", END],
)

workflow.add_edge("action", "agent")

app = workflow.compile(checkpointer=memory)