Stage 2 — Document Collection

Goal

Collect every available public record and supporting document required for title examination, producing a complete Evidence Package for downstream processing.

Scope

This stage is responsible only for collecting documents and metadata.

It does not:

• Extract document contents
• Build chain of title
• Detect liens
• Perform legal analysis

Output: Evidence Package.

Input

• Normalized Order
• Property Address
• APN (if available)
• County
• State
• Product Type

Responsibilities

• Execute collection workflow
• Query all configured sources
• Download available documents
• Collect source metadata
• Store raw files
• Record failures and retry status
• Produce immutable evidence package

Collection Sources

Public Sources

• County Recorder
• County Assessor
• Tax Office
• Court Records
• GIS
• Bankruptcy
• Secretary of State (UCC)

Future Commercial Providers

This platform will initially use public sources.

Commercial integrations (DataTrace, First American, CoreLogic, Fidelity, ATTOM, etc.) will be added later through dedicated provider adapters.

Third-Party Integrations

Browser Automation

• Playwright (primary)

Browser Infrastructure

• Browser worker pool
• Session management
• Cookie persistence

Automation

• Login automation
• Navigation
• Search automation
• File downloads
• Pagination

Infrastructure

• Object Storage (MinIO / S3)
• Queue (RabbitMQ)
• Retry engine
• Proxy support (optional)
• CAPTCHA solving (optional)

Connector Architecture

Collection Orchestrator
        │
        ├── Recorder Connector
        ├── Assessor Connector
        ├── Tax Connector
        ├── Court Connector - not needed
        ├── GIS Connector
        ├── Bankruptcy Connector - not needed
        ├── UCC Connector - Recorder
        └── Future Provider Connectors

Every connector implements the same lifecycle.

Authenticate
      ↓
Search
      ↓
Collect
      ↓
Download
      ↓
Normalize Metadata
      ↓
Return Collection Result

Evidence Package

Each collected document stores:

• Source
• Search Parameters
• Download Timestamp
• Raw File
• Metadata
• Checksum
• Download Status
• Retry Count
• Collection Logs

Edge Cases

• Website unavailable
• Login expired
• Session timeout
• CAPTCHA
• Rate limiting
• Corrupted downloads
• Duplicate documents
• Partial downloads
• Large result sets
• County website redesign

Retry Strategy

Failures should be isolated per connector.

If one connector fails:

• Retry
• Record failure reason
• Continue remaining connectors
• Never fail the entire collection pipeline because of a single source.

Parallel Execution

All connectors execute independently wherever possible to minimize collection time.

Human Involvement

Minimal.

Manual intervention only when:

• Websites become incompatible
• Authentication cannot be completed
• Documents are unavailable online

Expected automation: 90–100% depending on jurisdiction.

Definition of Done

• All configured connectors executed.
• Successful documents downloaded.
• Failures recorded.
• Raw evidence stored.
• Metadata persisted.
• Evidence Package created.
• Ready for Stage 3 (OCR & Data Extraction).