58 lines
2.1 KiB
Python
58 lines
2.1 KiB
Python
from langchain_openai import ChatOpenAI
|
||
from pydantic import SecretStr
|
||
from langchain.tools import tool
|
||
from langchain.agents import create_agent
|
||
import json
|
||
|
||
# 1. LLM connection
|
||
llm = ChatOpenAI(
|
||
model="gpt-4o-mini", # replace with your LM Studio model name
|
||
base_url="http://localhost:1234/v1",
|
||
api_key=SecretStr("fake"),
|
||
temperature=0.7,
|
||
)
|
||
|
||
# 2. Sub‑agent that generates a price table
|
||
@tool
|
||
def get_price(product: str, city: str) -> str:
|
||
"""Return a realistic price for the product in the given city.
|
||
The function internally creates a small agent that asks the LLM to produce a markdown table.
|
||
"""
|
||
# Sub‑agent prompt – keep it short and deterministic
|
||
sub_prompt = (
|
||
f"You are a market analyst. Provide a realistic price for {product} in {city}. "
|
||
"Return a markdown table with columns: Продукт, Цена (руб.), Магазин."
|
||
)
|
||
# Create the sub‑agent
|
||
sub_agent = create_agent(
|
||
model=llm,
|
||
tools=[], # no external tools needed for this simple query
|
||
system_prompt=sub_prompt,
|
||
)
|
||
# Ask the sub‑agent and get its response
|
||
result = sub_agent.invoke({"messages": [{"role": "human", "content": "Generate table"}]})
|
||
return result["messages"][-1]["content"]
|
||
|
||
# 3. Main agent with get_price tool
|
||
main_agent = create_agent(
|
||
model=llm,
|
||
tools=[get_price],
|
||
system_prompt="Ты помощник по планированию покупок.",
|
||
)
|
||
|
||
# 4. Run the main agent on a sample query
|
||
query = "Помоги составить список покупок: молоко, хлеб, яблоки. Я нахожусь в Казани."
|
||
response = main_agent.invoke({"messages": [{"role": "human", "content": query}]})
|
||
# Pretty‑print all messages (including tool calls)
|
||
for msg in response["messages"]:
|
||
if msg.get("content"):
|
||
print(msg["content"])
|
||
elif msg.get("tool_calls"):
|
||
for call in msg["tool_calls"]:
|
||
name = call["name"]
|
||
args = json.dumps(call["args"], ensure_ascii=False)
|
||
print(f"{name}({args})")
|
||
|
||
# Final answer (last message content)
|
||
print("\n---\nAnswer:\n", response["messages"][-1]["content"])
|