Workflow Automation

from langchain.agents import AgentExecutor, create_tool_calling_agent
from langchain_openai import ChatOpenAI
from langchain.tools import tool
from langchain.prompts import ChatPromptTemplate

# Define workflow tools
@tool
def extract_invoice_data(text: str) -> dict:
    """Extract key fields from invoice text: vendor, amount, date, PO number."""
    # In production, use a fine-tuned model or structured extraction
    return {"vendor": "...", "amount": "...", "date": "...", "po": "..."}

@tool
def validate_against_po(po_number: str) -> str:
    """Check purchase order exists and matches the invoice amount."""
    # Query ERP system
    return f"PO {po_number} validated: amount matches, approved for payment."

@tool
def create_accounting_entry(invoice_data: str) -> str:
    """Create an entry in the accounting system for the validated invoice."""
    return "Accounting entry created. Journal ID: JE-2025-00142"

@tool
def send_notification(recipient: str, message: str) -> str:
    """Send a notification to the relevant stakeholder."""
    return f"Notification sent to {recipient}"

# Create the workflow agent
llm = ChatOpenAI(model="gpt-4o", temperature=0)
tools = [extract_invoice_data, validate_against_po, 
         create_accounting_entry, send_notification]

prompt = ChatPromptTemplate.from_messages([
    ("system", """You are an Invoice Processing Agent. 
    Follow this workflow:
    1. Extract invoice data from the document
    2. Validate against the purchase order
    3. Create the accounting entry
    4. Notify the approver
    Handle errors gracefully and report any issues."""),
    ("human", "{input}"),
    ("placeholder", "{agent_scratchpad}")
])

agent = create_tool_calling_agent(llm, tools, prompt)
executor = AgentExecutor(agent=agent, tools=tools, verbose=True)

# Run the workflow
result = executor.invoke({
    "input": "Process this invoice: Vendor=Acme Corp, Amount=$5,240, "
             "Date=2025-01-15, PO=PO-2025-0089"
})
print(result["output"])