From 3a76317388a52549e15826a8d77fbeefd0ef7bde Mon Sep 17 00:00:00 2001
From: Danil Parunin 5f1b81b8-4f5d-11e8-9c2d-fa7ae01bbebc <dapa46@mail.ru>
Date: Mon, 15 Jun 2026 12:24:41 +0000
Subject: [PATCH] add: main.py

---
 main.py | 93 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 93 insertions(+)
 create mode 100644 main.py

diff --git a/main.py b/main.py
new file mode 100644
index 0000000..4ba12fd
--- /dev/null
+++ b/main.py
@@ -0,0 +1,93 @@
+import os
+import asyncio
+from dotenv import load_dotenv
+from langchain_openai import ChatOpenAI
+from langchain_core.messages import HumanMessage
+from langchain_core.prompts import PromptTemplate
+from langchain_core.output_parsers import PydanticOutputParser
+from pydantic import BaseModel, Field
+from langchain.tools import tool
+from deepagents import create_deep_agent
+from deepagents.backends import CompositeBackend, LocalShellBackend, FilesystemBackend
+
+# Load environment variables (OPENAI_API_KEY)
+load_dotenv()
+
+# 1. Define the Pydantic model for the task card
+class TaskCard(BaseModel):
+    title: str = Field(description="Краткое название задания")
+    subject: str = Field(description="Предмет или тема задания")
+    deadline_hint: str = Field(description="Краткая подсказка о сроке выполнения, например 'к пятнице'")
+    deliverable_type: str = Field(description="Тип сдачи: отчёт, код, презентация и т.п.")
+    grading_hints: list[str] = Field(description="Список критериев оценки, упомянутых в тексте")
+
+# 2. Create the parser that will enforce the structure
+parser = PydanticOutputParser(pydantic_object=TaskCard)
+
+# 3. Prompt template that asks the model to output the data in the required format
+prompt = PromptTemplate(
+    template="""
+Ниже приведено описание задания. Ваша задача — извлечь из него следующую информацию и вернуть в формате JSON, соответствующем модели TaskCard:
+
+{input_text}
+
+{format_instructions}
+""",
+    input_variables=["input_text"],
+    partial_variables={"format_instructions": parser.get_format_instructions()},
+)
+
+# 4. LLM configuration (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,
+)
+
+# 5. Define a tool that performs the parsing chain
+@tool
+def parse_task(input_text: str) -> str:
+    """Parse a raw task description into a structured JSON string."""
+    chain = prompt | llm | parser
+    result = chain.invoke({"input_text": input_text})
+    # parser returns a dict; convert to JSON string for the agent to return
+    return result.model_dump_json()
+
+# 6. Backend for the agent (filesystem + local shell, though not used here)
+backend = CompositeBackend([
+    LocalShellBackend(workspace_dir="./workspace"),
+    FilesystemBackend(),
+])
+
+# 7. Create the deep agent with the parsing tool
+agent = create_deep_agent(
+    model=llm,
+    tools=[parse_task],
+    backend=backend,
+    system_prompt="You are a task‑card extractor. Use the provided tool to parse the input and return the JSON string.",
+)
+
+# 8. Main entry point
+async def main():
+    # Example input – replace with any user text
+    user_text = "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_text)]},
+        {"configurable": {"thread_id": "session-1"}},
+    )
+    # The agent returns the tool output as the last message
+    output = result["messages"][-1].content
+    print("Parsed JSON:")
+    print(output)
+    # Pretty‑print the parsed object
+    card = TaskCard.parse_raw(output)
+    print("\nHuman‑readable summary:")
+    print(f"Title: {card.title}")
+    print(f"Subject: {card.subject}")
+    print(f"Deadline hint: {card.deadline_hint}")
+    print(f"Deliverable type: {card.deliverable_type}")
+    print(f"Grading hints: {', '.join(card.grading_hints)}")
+
+if __name__ == "__main__":
+    asyncio.run(main())