112 lines
4.9 KiB
Python
112 lines
4.9 KiB
Python
"""
|
||
Main entry point for the "сырой текст задания → плоская карточка" task.
|
||
|
||
The script demonstrates how to convert a free‑form description of an assignment into a structured data object using LangChain and Pydantic.
|
||
|
||
Usage examples are provided in the ``__main__`` section – three different raw texts are parsed and printed.
|
||
"""
|
||
|
||
from __future__ import annotations
|
||
|
||
import os
|
||
from typing import List
|
||
|
||
from langchain_openai import ChatOpenAI
|
||
from langchain_core.prompts import PromptTemplate
|
||
from langchain_core.output_parsers import PydanticOutputParser
|
||
from pydantic import BaseModel, Field
|
||
|
||
# ---------------------------------------------------------------------------
|
||
# 1. Define the data model that represents a task card.
|
||
# ---------------------------------------------------------------------------
|
||
class TaskCard(BaseModel):
|
||
"""Structured representation of an assignment description.
|
||
|
||
The fields are intentionally generic – they capture the most common pieces of information
|
||
that appear in the course tasks:
|
||
|
||
* ``title`` – short name of the task.
|
||
* ``subject`` – subject or topic area.
|
||
* ``deadline_hint`` – free‑form hint about when the task should be finished.
|
||
* ``deliverable_type`` – what is expected to be submitted (report, code, presentation…).
|
||
* ``grading_hints`` – list of items that influence grading.
|
||
"""
|
||
|
||
title: str = Field(..., description="Short name of the task")
|
||
subject: str | None = Field(None, description="Subject or topic area")
|
||
deadline_hint: str | None = Field(
|
||
None,
|
||
description="Free‑form hint about when the task should be finished",
|
||
)
|
||
deliverable_type: str | None = Field(
|
||
None,
|
||
description="What is expected to be submitted (report, code, presentation…)",
|
||
)
|
||
grading_hints: List[str] = Field(
|
||
default_factory=list,
|
||
description="List of items that influence grading",
|
||
)
|
||
|
||
# ---------------------------------------------------------------------------
|
||
# 2. Build the prompt and parser.
|
||
# ---------------------------------------------------------------------------
|
||
parser = PydanticOutputParser(pydantic_object=TaskCard)
|
||
|
||
prompt_template = PromptTemplate(
|
||
template=(
|
||
"You are an assistant that extracts structured information from a free‑form task description."
|
||
" Return only the data in JSON format that matches the following schema:\n{format_instructions}\n"
|
||
"Input: {text}"
|
||
),
|
||
input_variables=["text"],
|
||
partial_variables={"format_instructions": parser.get_format_instructions()},
|
||
)
|
||
|
||
# ---------------------------------------------------------------------------
|
||
# 3. Create the LLM instance.
|
||
# ---------------------------------------------------------------------------
|
||
llm = ChatOpenAI(
|
||
model="openai/gpt-oss-20b:free",
|
||
base_url="https://platform.brojs.ru/jrnl-bh/api/inference/v1",
|
||
api_key=os.getenv("JOURNAL_MCP_PAT"),
|
||
temperature=0.0,
|
||
)
|
||
|
||
# ---------------------------------------------------------------------------
|
||
# 4. Helper that runs the chain and returns a TaskCard.
|
||
# ---------------------------------------------------------------------------
|
||
async def parse_task(text: str) -> TaskCard:
|
||
"""Parse *text* into a :class:`TaskCard` using LangChain.
|
||
|
||
The function is asynchronous because ``ChatOpenAI`` uses an async API. It can be called from
|
||
synchronous code via ``asyncio.run``.
|
||
"""
|
||
|
||
chain = prompt_template | llm | parser
|
||
result = await chain.ainvoke({"text": text})
|
||
return result
|
||
|
||
# ---------------------------------------------------------------------------
|
||
# 5. Demo – three example texts.
|
||
# ---------------------------------------------------------------------------
|
||
if __name__ == "__main__":
|
||
import asyncio
|
||
|
||
examples = [
|
||
"Сдайте к пятнице мини‑отчёт по LangChain: 2 страницы, упор на агентов. Оценка: за полноту и за пример кода.",
|
||
"На следующей неделе подготовьте презентацию о Qdrant. Должно быть 10 слайдов, включать примеры кода. Оценка по содержанию и дизайну.",
|
||
"Разработайте скрипт на Python, который генерирует случайный пароль длиной 12 символов. Сдача – код в репозитории. Оценка: корректность и безопасность.",
|
||
]
|
||
|
||
async def demo():
|
||
for i, txt in enumerate(examples, start=1):
|
||
card = await parse_task(txt)
|
||
print(f"\nExample {i}:")
|
||
print("Raw text:")
|
||
print(txt)
|
||
print("\nParsed card: ")
|
||
# Pretty‑print the model using ``model_dump`` – it returns a dict.
|
||
print(card.model_dump(indent=2, sort_keys=False))
|
||
|
||
asyncio.run(demo())
|