Enterprise Intelligent Document Processing (IDP): Automated Contract Extraction with RAG and LLMs

from pydantic import BaseModel, Field, validator
from typing import List, Optional
from langchain_openai import ChatOpenAI

class ContractClause(BaseModel):
    clause_type: str = Field(description="Type of clause: INDEMNITY, TERMINATION, GOVERNING_LAW")
    summary: str = Field(description="Brief summary of clause terms")
    risk_level: str = Field(description="HIGH, MEDIUM, LOW risk assessment")

class EnterpriseContractExtraction(BaseModel):
    contract_title: str
    effective_date: Optional[str]
    governing_jurisdiction: str
    liability_cap_usd: float = Field(description="Maximum financial liability cap in USD")
    clauses: List[ContractClause]

    @validator('liability_cap_usd')
    def validate_positive_liability(cls, v):
        if v < 0:
            raise ValueError("Liability cap cannot be negative")
        return v

llm = ChatOpenAI(model="gpt-4o", temperature=0.0)
structured_extractor = llm.with_structured_output(EnterpriseContractExtraction)

async def extract_contract_metadata(contract_text: str) -> EnterpriseContractExtraction:
    prompt = f"Analyze the following legal document chunk and extract key metadata:\n\n{contract_text}"
    result = await structured_extractor.ainvoke(prompt)
    return result