106 lines
4.2 KiB
Python
106 lines
4.2 KiB
Python
import os
|
|
import asyncio
|
|
from typing import List
|
|
|
|
from pydantic import BaseModel, Field
|
|
from langchain_openai import ChatOpenAI
|
|
from langchain_core.prompts import PromptTemplate
|
|
from langchain_core.output_parsers import PydanticOutputParser
|
|
from langchain_core.messages import HumanMessage
|
|
from langchain.tools import tool
|
|
|
|
from deepagents import create_deep_agent
|
|
from deepagents.backends import FilesystemBackend, LocalShellBackend, CompositeBackend
|
|
|
|
# -------------------- Pydantic model --------------------
|
|
class AssignmentCard(BaseModel):
|
|
title: str = Field(description="Краткое название задания")
|
|
subject: str = Field(description="Тема или предмет задания")
|
|
deadline_hint: str = Field(description="Подсказка о сроке сдачи (например, к пятнице, 12.09)")
|
|
deliverable_type: str = Field(description="Что нужно сдать: отчёт, код, презентация и т.п.")
|
|
grading_hints: List[str] = Field(description="Список упомянутых критериев оценки")
|
|
|
|
# -------------------- Structured output parser --------------------
|
|
parser = PydanticOutputParser(pydantic_object=AssignmentCard)
|
|
|
|
# -------------------- Prompt template --------------------
|
|
prompt_template = PromptTemplate(
|
|
template=(
|
|
"Ты получаешь неформальное описание учебного задания и должен вернуть "
|
|
"структурированные данные в формате JSON, соответствующем схеме ниже.\n"
|
|
"Верни только JSON, без пояснений.\n"
|
|
"Описание задания:\n{task_description}\n\n"
|
|
"{format_instructions}"
|
|
),
|
|
input_variables=["task_description"],
|
|
partial_variables={"format_instructions": parser.get_format_instructions()},
|
|
)
|
|
|
|
# -------------------- LLM (OpenRouter) --------------------
|
|
llm = ChatOpenAI(
|
|
model="openai/gpt-oss-20b:free",
|
|
base_url="https://openrouter.ai/api/v1",
|
|
api_key=os.getenv("OPENAI_API_KEY"),
|
|
temperature=0.0,
|
|
)
|
|
|
|
# -------------------- Tool that runs the chain --------------------
|
|
@tool
|
|
def parse_assignment(task_description: str) -> str:
|
|
"""
|
|
Parse a free-form assignment description into a structured AssignmentCard.
|
|
Returns a JSON string that can be parsed by the Pydantic model.
|
|
"""
|
|
chain = prompt_template | llm | parser
|
|
result = chain.invoke({"task_description": task_description})
|
|
# result is an AssignmentCard instance
|
|
return result.model_dump_json(indent=2)
|
|
|
|
# -------------------- DeepAgent setup --------------------
|
|
backend = CompositeBackend(
|
|
[
|
|
LocalShellBackend(workspace_dir="./workspace"),
|
|
FilesystemBackend(),
|
|
]
|
|
)
|
|
|
|
agent = create_deep_agent(
|
|
model=llm,
|
|
tools=[parse_assignment],
|
|
backend=backend,
|
|
system_prompt="You are a helpful assistant that extracts assignment data.",
|
|
)
|
|
|
|
# -------------------- Main execution --------------------
|
|
async def main():
|
|
# Пример входного текста
|
|
user_input = (
|
|
"Сдайте к пятнице мини-отчёт по LangChain: 2 страницы, упор на агентов. "
|
|
"Оценка: за полноту и за пример кода."
|
|
)
|
|
result = await agent.ainvoke(
|
|
{"messages": [HumanMessage(content=user_input)]},
|
|
{"configurable": {"thread_id": "session-1"}},
|
|
)
|
|
# Последнее сообщение агента содержит JSON строку
|
|
json_output = result["messages"][-1].content
|
|
print("=== Structured JSON ===")
|
|
print(json_output)
|
|
|
|
# Для наглядности выводим человекочитаемую сводку
|
|
try:
|
|
import json
|
|
|
|
data = json.loads(json_output)
|
|
print("\n=== Human-readable summary ===")
|
|
print(f"Title: {data.get('title')}")
|
|
print(f"Subject: {data.get('subject')}")
|
|
print(f"Deadline hint: {data.get('deadline_hint')}")
|
|
print(f"Deliverable type: {data.get('deliverable_type')}")
|
|
print(f"Grading hints: {', '.join(data.get('grading_hints', []))}")
|
|
except Exception as e:
|
|
print("Failed to parse JSON:", e)
|
|
|
|
|
|
if __name__ == "__main__":
|
|
asyncio.run(main()) |