# app/agents.py
import asyncio
import random
from .schemas import ReceiptInfo, InvoiceInfo, ReceiptItem

# --- Agent核心功能 (占位符/模拟实现) ---
# 在实际应用中，这些函数将被替换为调用LangChain和LLM的真实逻辑。

async def agent_classify_document(text: str) -> str:
    """Agent 1: 文件分类 (模拟)"""
    print("--- [Agent 1] 正在进行文档分类...")
    await asyncio.sleep(0.5) # 模拟网络延迟
    doc_types = ["信件", "收据", "发票", "合约"]
    if "发票" in text: return "发票"
    if "收据" in text or "小票" in text: return "收据"
    if "合同" in text or "协议" in text: return "合约"
    return random.choice(doc_types)

async def agent_extract_receipt_info(text: str) -> ReceiptInfo:
    """Agent 2: 收据信息提取 (模拟)"""
    print("--- [Agent 2] 正在提取收据信息...")
    await asyncio.sleep(1) # 模拟LLM处理时间
    return ReceiptInfo(
        merchant_name="模拟超市",
        transaction_date="2025-08-10",
        total_amount=198.50,
        items=[ReceiptItem(name="牛奶", quantity=2, price=11.5)]
    )

async def agent_extract_invoice_info(text: str) -> InvoiceInfo:
    """Agent 3: 发票信息提取 (模拟)"""
    print("--- [Agent 3] 正在提取发票信息...")
    await asyncio.sleep(1) # 模拟LLM处理时间
    return InvoiceInfo(
        invoice_number="INV123456789",
        issue_date="2025-08-09",
        seller_name="模拟科技有限公司",
        total_amount_in_figures=12000.00
    )

async def agent_vectorize_and_store(doc_id: str, text: str, category: str, vector_db: dict):
    """Agent 4: 向量化并存储 (模拟)"""
    print(f"--- [Agent 4] 正在向量化文档 (ID: {doc_id})...")
    await asyncio.sleep(0.5)
    chunks = [text[i:i+200] for i in range(0, len(text), 200)]
    vector_db[doc_id] = {
        "metadata": {"category": category, "chunk_count": len(chunks)},
        "content_chunks": chunks,
        "vectors": [random.random() for _ in range(len(chunks) * 128)]
    }
    print(f"--- [Agent 4] 文档 {doc_id} 已存入向量数据库。")