LG-RackandStack - Loan Data Processing Workflow

Overview

LG-RackandStack is a comprehensive LangGraph workflow designed for automated loan data processing and document management. This workflow integrates with ESFuse APIs, Encompass loan management systems, and document repositories to create a complete loan processing pipeline.

What's Included

Complete Workflow Pipeline

The workflow includes six main processing nodes:

• extract_input: Parse and validate user input
• pull_data: Retrieve loan data from ESFuse API and auto-parse borrower information
• pull_doc: Download document data for all specified document IDs
• push_data: Update Encompass loan fields with parsed borrower data
• push_doc: Create loan submissions and associate documents
• summary: Generate comprehensive workflow reporting

Advanced State Management

• State class with comprehensive fields:
  • user_input: JSON input with loan/document data
  • loan_id, task_id, client_id: Core identifiers
  • document_ids: JSON array of document IDs to process
  • field_updates: Auto-generated Encompass field mappings
  • Workflow tracking: pull_data_result, push_data_result, etc.

API Integrations

• ESFuse APIs: Loan and document data retrieval
• Encompass Integration: Automatic field updates with parsed borrower data
• TaskDoc/DocRepo: Document submission and task management
• Async Processing: Non-blocking API calls for performance

Getting Started

Prerequisites

• Python 3.9+
• LangGraph
• cuteagent (Document and loan processing library)
• aiohttp (for async API calls)

Required Environment Variables

Create a .env file with the following variables:

# ESFuse API Configuration
ESFUSE_TOKEN=your_esfuse_token
LOAN_API_BASE_URL=https://your-loan-api.com/prod
DOC_API_BASE_URL=https://your-doc-api.com/prod

# Encompass API Configuration  
ENCOMPASS_BASE_URL=https://your-encompass-api.com
ENCOMPASS_ACCESS_TOKEN=your_encompass_token

# TaskDoc API Configuration
TASKDOC_API_TOKEN=your_taskdoc_api_token
TASKDOC_AUTH_TOKEN=your_taskdoc_auth_token

# Submission Configuration (optional)
SUBMISSION_TYPE=Initial Submission
AUTO_LOCK=false

Installation

# Navigate to project directory
cd LG-RackandStack-main

# Install dependencies
pip install -e .

# Run the workflow
langgraph dev

Workflow Process

Input Format

The workflow accepts JSON input with the following structure:

{
  "status": "completed",
  "loan_id": "25",
  "task_id": "347360", 
  "client_id": "loan_25",
  "document_ids": "[\"953\", \"954\", \"955\", \"956\", \"957\", \"958\", \"959\", \"960\", \"962\", \"963\", \"964\", \"965\", \"966\", \"967\", \"968\", \"969\"]",
  "documents_stored": "16",
  "documents_processed": "17"
}

Processing Flow

START → extract_input → pull_data → pull_doc → push_data → push_doc → summary → END

1. extract_input

• Validates environment variables
• Parses input JSON and extracts all required fields
• Sets up state for subsequent nodes

2. pull_data

• Calls ESFuse loan API: GET /loan?clientId={client_id}&loanId={loan_id}
• Automatically parses borrower data and creates Encompass field mappings:
  • 4000: first_name
  • 4001: middle_name
  • 4002: last_name
  • 1204: email
  • 65: SSN
  • 1402: date_of_birth
  • FR0106: city
  • FR0126: address1
• Stores parsed field updates in state

3. pull_doc

• Processes all document IDs from the input array
• Calls ESFuse doc API: GET /doc?clientId={client_id}&docId={doc_id} for each document
• Handles both JSON and PDF responses
• Creates summary with success/failure counts

4. push_data

• Uses the auto-generated field updates from pull_data
• Retrieves Encompass loan GUID from the loan API
• Updates Encompass: POST /api/v1/write_loan_data
• Pushes all parsed borrower data to corresponding Encompass fields

5. push_doc

• Creates loan submission using the CuteAgent library
• Library internally handles task processing using the task_id
• Associates all documents with the submission
• Supports submission types and auto-lock functionality

6. summary

• Generates comprehensive workflow report
• Shows success/failure status for each node
• Provides document processing statistics
• Creates final workflow summary

Key Features

Automatic Field Mapping

The workflow automatically extracts borrower information from loan data and maps it to Encompass field IDs:

Data Field	Encompass Field ID	Source
First Name	4000	loaninfo.borrowers_attributes[0].first_name
Middle Name	4001	loaninfo.borrowers_attributes[0].middle_name
Last Name	4002	loaninfo.borrowers_attributes[0].last_name
Email	1204	loaninfo.borrowers_attributes[0].email
SSN	65	loaninfo.borrowers_attributes[0].ssn
Date of Birth	1402	loaninfo.borrowers_attributes[0].date_of_birth
City	FR0106	loaninfo.city
Address 1	FR0126	loaninfo.address1

Error Handling & Logging

• Comprehensive error handling at each node
• Detailed logging for debugging and monitoring
• Graceful failure handling with error reporting
• State tracking throughout the workflow

Performance Optimizations

• Async API calls using aiohttp for non-blocking operations
• Efficient JSON parsing and data extraction
• Minimal memory footprint (no file downloads)
• Parallel document processing capabilities

Development

File Structure

LG-RackandStack-main/
├── src/agent/
│   ├── __init__.py
│   └── graph.py          # Main workflow definition
├── lgBlank.egg-info/     # Package metadata
├── pyproject.toml        # Project dependencies and configuration
├── langgraph.json        # LangGraph configuration
├── workflow_config.json  # Workflow configuration template
└── README.md            # This file

Testing & Development

Use LangGraph Studio to test and develop:

1. Start the development server:

   langgraph dev

2. Access the interfaces:

   • 🚀 API: http://127.0.0.1:2024
   • 🎨 Studio UI: https://smith.langchain.com/studio/?baseUrl=http://127.0.0.1:2024
   • 📚 API Docs: http://127.0.0.1:2024/docs

3. Test with sample input:

   {
     "loan_id": "25",
     "task_id": "347360",
     "client_id": "loan_25", 
     "document_ids": "[\"953\", \"954\"]"
   }

Workflow Configuration

The workflow_config.json file can be used to customize:

• API endpoints and tokens
• Field mapping configurations
• Submission settings
• Error handling preferences

API Reference

Input Parameters

• loan_id (required): Unique loan identifier
• task_id (required): TaskDoc task identifier for document processing
• client_id (required): ESFuse client identifier
• document_ids (required): JSON array of document IDs to process
• documents_stored (optional): Count of stored documents
• documents_processed (optional): Count of processed documents

Output Structure

{
  "workflow_status": "Success",
  "total_nodes_completed": 6,
  "pull_data_success": true,
  "pull_doc_success": true, 
  "push_data_success": true,
  "push_doc_success": true,
  "documents_processed": 16,
  "successful_documents": 15,
  "failed_documents": 1,
  "timestamp": "2025-01-04 23:45:30"
}

License

This template follows the same license as the parent project.