Process Library & AI Evaluation - Configuration Guide

---

Overview

The Process Library is a collection of defined "Processes" that enable AI-powered document evaluation and automated determination. Each process configures context, criteria, and knowledge that allows the AI system to evaluate submitted applications, documents, or forms.

Process Library Capabilities

---

Prerequisites

Before creating processes, ensure:

• User has appropriate permissions for Document Processing module
• Forms/entities exist if using existing form option
• Template documents are prepared (if using template upload)

---

Section 1: Process Library Overview

1.1 Process Types

1.2 Process Statuses

1.3 Process Library Navigation

Navigation: Document Processing -> Process Library

The Process Library table displays:

• Process name
• Process type
• Intake type
• Description
• Last modified date

Available Actions:

• Edit process
• View process
• Delete process

---

Section 2: Creating a New Process

2.1 Process Creation Steps

Creating a process follows a 5-step wizard:

Step 1: Basic Info
    |-- Name, Type, Description

Step 2: Template
    |-- Upload template OR Select existing form OR Start from scratch

Step 3: Entity
    |-- Configure fields, Create AI fields, Add custom fields

Step 4: Form
    |-- Configure form layout, Add fields, Set validation

Step 5: Intake
    |-- Configure how submissions are received

2.2 Step 1: Basic Info

Navigation: Process Library -> + Add New Process -> Basic Info

2.3 Step 2: Template

Navigation: Process Creation -> Step 2: Template

Three options for template setup:

Template Fields:

2.4 Step 3: Entity (Field Configuration)

Navigation: Process Creation -> Step 3: Entity

In this step, you configure the entity fields for the process:

1. Automatic Field Detection: System detects fields from uploaded template
2. Add Missing Fields: Manually add fields not detected
3. Configure AI Fields: Enable AI autofill for fields
4. Create Custom Fields: Add dropdown, text, number fields

Field Types Available

The process supports all standard field types plus AI-enhanced fields:

• Text, Text Area
• Dropdown, Radio, Checkbox
• Date, Number
• File, Signature
• And all other entity field types

AI Autofill Configuration

Navigation: Field -> Three-dot menu -> Autofill with AI

This opens the AI sidebar for configuring AI-powered field filling.

---

Section 3: AI Autofill Configuration

3.1 AI Autofill Overview

AI Autofill allows the system to automatically extract information from documents and fill form fields using AI analysis.

3.2 AI Autofill Settings

3.3 Configuring AI Instructions

Navigation: AI Sidebar -> Add Instruction

When configuring AI autofill:

1. Select Field to Analyze: Choose which field the AI should read/analyze
2. Write AI Instructions: Provide clear instructions for what to extract
3. Select Field to Assign: Choose which field receives the extracted value
4. Add Instruction: Click to add the instruction to the list
5. Test AI Instruction: Test the instruction with sample data

Instruction List Features:

• View all configured instructions
• Edit existing instructions
• Delete instructions
• Reorder instructions

3.4 AI Instruction Best Practices

When writing AI instructions:

3.5 Evaluation Data Structure

When AI evaluates instructions, the system stores detailed evaluation data:

Evaluation Instruction Data:

Watched Field Instruction:

3.6 AI Instruction Data Model

Each AI instruction contains:

---

Section 4: Knowledge Base Integration

4.1 Knowledge Base Overview

The AI system can be enhanced with external knowledge sources to improve accuracy and context-awareness. Knowledge sources provide reference data that the AI uses during field extraction and evaluation.

Navigation: AI Sidebar -> Knowledge Tab -> Add Knowledge Dropdown

4.2 Knowledge Source Types

4.3 Adding API Knowledge

Navigation: Add Knowledge -> API

Configure external API endpoints to provide reference data:

HTTP Configuration:

Header Configuration:

Query Parameters:

Body Parameters (for POST/PUT):

Property Schema:

API Knowledge Data Structure:

type ApiKnowledge = {
  type: 'api';
  data: {
    api_url: string;
    request_method: string;
    request_headers: Record<string, string>;
    query_params: Record<string, string>;
    request_body: Record<string, any>;
  };
};

4.4 Adding Document Knowledge

Navigation: Add Knowledge -> Document

Upload reference documents for the AI to use:

File Requirements:

Supported Formats:

• PDF documents
• Word documents (.doc, .docx)
• Text files (.txt)
• Image files (for OCR processing)

Document Knowledge Data Structure:

type DocumentKnowledge = {
  type: 'document';
  data: {
    file_details: {
      file_id: string;
      file_name?: string;
      size?: number;
    }[];
  };
};

4.5 Adding Website Knowledge

Navigation: Add Knowledge -> Website

Add website URLs as reference sources:

URL Requirements:

Website Knowledge Data Structure:

type WebsiteKnowledge = {
  type: 'website';
  data: {
    web_url: string;  // Field name: 'web_url'
  };
};

4.6 Adding SharePoint Knowledge

Navigation: Add Knowledge -> SharePoint

Connect to SharePoint document libraries:

Configuration:

SharePoint Knowledge Data Structure:

type SharepointKnowledge = {
  type: 'sharepoint';
  data: {
    siteUrl: string;
    listId: string;
  };
};

4.7 Knowledge Item Structure

All knowledge sources share a common base structure:

type KnowledgeItem = {
  _id: string;           // Unique identifier
  name: string;          // Display name
  description: string;   // Description of the knowledge
} & KnowledgeData;       // Type-specific data

---

Section 5: Form Configuration

5.1 Step 4: Form

Navigation: Process Creation -> Step 4: Form

In this step, configure how the form appears to users:

Form Settings

Adding Fields to Form

Navigation: Form Builder -> Add Fields

1. Click Add Fields button
2. Select fields from the available entity fields
3. Configure field settings (label, placeholder, description)
4. Drag and drop to reorder fields

Field Configuration Options

For each field in the form:

---

Section 6: Intake Configuration

6.1 Step 5: Intake

Navigation: Process Creation -> Step 5: Intake

Configure how submissions are received and processed.

6.2 Intake Types

6.3 Intake Configuration Fields

6.4 Instructions Configuration

Additional instructions for the process:

---

Section 7: Permitting Solution Patterns

7.1 Permit Application Process Pattern

Process Name: Permit Application Evaluation

Configuration:

AI Autofill Examples:

7.2 Document Evaluation Process Pattern

Process Name: Contract Document Review

Configuration:

7.3 Zoning Compliance Check Pattern

AI Autofill for Zoning:

---

Section 8: Evaluation Results & Metrics

8.1 Evaluation Result Structure

When documents/forms are processed, the system generates evaluation results:

8.2 Process History Fields

Track the history of evaluations with these fields:

8.3 Dashboard Metrics

The system tracks the following metrics:

8.4 Final Evaluation Output

Each evaluation produces a final output:

type FinalOutput = {
  value: string[];    // Array of extracted/determined values
  reasoning: string;  // AI's reasoning for the decision
};

---

Section 9: Testing and Publishing

9.1 Testing AI Instructions

Before publishing, test all AI instructions:

1. Navigate to AI Sidebar
2. Click Test AI Instruction
3. Upload sample document
4. Verify extracted values
5. Adjust instructions if needed

9.2 Publishing the Process

After completing all steps:

1. Review all configurations
2. Click Submit to save to Process Library
3. Process status changes to "Published"
4. Process becomes available for use

---

Section 10: Field Configuration in Process

10.1 Entity Field Configuration

Fields in a process can be configured with various settings:

Field Properties:

10.2 Field Validation Schema

10.3 Field Configuration UI Constants

---

Troubleshooting

Common Issues

AI Autofill Not Extracting Data

Check:

1. Verify the source field contains readable text
2. Ensure AI instructions are clear and specific
3. Check that document quality is sufficient for OCR
4. Test with sample documents

Field Detection Not Working

Check:

1. Ensure template PDF has proper form fields
2. Try different template formats
3. Manually add missing fields

Intake Not Receiving Documents

Check:

1. Verify SharePoint connection credentials
2. Check folder permissions
3. Confirm API key is valid
4. Review intake URL configuration

Knowledge Source Not Updating

Check:

1. Verify API endpoint is accessible
2. Check authentication headers
3. Ensure website URL is valid HTTP/HTTPS
4. Confirm SharePoint permissions

---

Code Reference

Source Files - Process Library

Source Files - AI Autofill

Source Files - Knowledge Base

Source Files - Evaluation & Dashboard

Source Files - Entity Configuration

---

This guide is aligned with the document-processing-app codebase as of January 2026.