AlanRocha's picture
Update agent.py
32ed98c verified
Raw
History Blame Contribute Delete
2.05 kB
"""LangGraph Agent"""
import os
from dotenv import load_dotenv
from langchain_google_genai import ChatGoogleGenerativeAI
from langchain_groq import ChatGroq
from langchain_huggingface import ChatHuggingFace, HuggingFaceEndpoint
from langchain_core.messages import SystemMessage
from langgraph.prebuilt import create_react_agent # Importação corrigida para o LangGraph
from tools import multiply, wiki_search, web_search, arvix_search, execute_python_code, YouTubeVideoAnalysisTool, read_excel_format, transcribe_mp3
load_dotenv()
_AGENT_DIR = os.path.dirname(os.path.abspath(__file__))
# load the system prompt from the file
with open(os.path.join(_AGENT_DIR, "system_prompt.txt"), "r", encoding="utf-8") as f:
system_prompt = f.read()
# System message
sys_msg = SystemMessage(content=system_prompt)
tools = [
multiply,
wiki_search,
web_search,
arvix_search,
execute_python_code,
YouTubeVideoAnalysisTool,
read_excel_format,
transcribe_mp3,
]
# Build graph function
def build_graph(provider: str | None = None):
if provider is None:
provider = os.getenv("LLM_PROVIDER", "groq").strip().lower()
if provider == "google":
llm = ChatGoogleGenerativeAI(model="gemini-2.5-flash", temperature=0)
elif provider == "groq":
model = os.getenv("GROQ_MODEL")
seed = int(os.getenv("GROQ_SEED", "42"))
llm = ChatGroq(model=model, temperature=0, model_kwargs={"seed": seed})
elif provider == "huggingface":
# TODO: Add huggingface endpoint. crédits tres limités...
llm = ChatHuggingFace(
llm=HuggingFaceEndpoint(
url="https://api-inference.huggingface.co/models/Meta-DeepLearning/llama-2-7b-chat-hf",
temperature=0,
),
)
else:
raise ValueError("Invalid provider. Choose 'google', 'groq' or 'huggingface'.")
# Formato correto e atualizado para as novas versões do LangGraph:
return create_react_agent(llm, tools, prompt=system_prompt)