Modernizing legal document ingestion and archival with an event-driven, serverless architecture

LiteBreeze helped Actapublica modernize its outdated, non-scalable document processing system. Robotea, the new solution, built on AWS, improved throughput, reduced infrastructure costs, and gave editorial teams faster access to archived content.

Client Overview

The client is a Swedish news agency operating in the journalism and legal document archiving sector.

1. Acts as a national source of legal news and documents
2. Maintains a real-time searchable archive of legal records
3. Supports both media professionals and public users through timely reporting and easy access to legal documentation

Business Challenge

The agency’s workflow involved:

• Collecting and archiving legal documents from various Swedish government agencies
• Processing scanned PDFs through OCR pipelines to extract text
• Indexing documents into Elasticsearch for search and archival
• Enriching metadata (region, type, authority) — partly via Amazon Mechanical Turk (MTurk)
• Making the enriched archive accessible to editorial teams and the public

However, their legacy PHP-based system created several issues:

• Loosely connected scripts worked sequentially, creating bottlenecks
• Single-threaded design caused slow processing and frequent backlogs
• Could not keep up with the growing volume of incoming documents
• Editorial teams faced delays in reporting due to late document availability

The solution: Robotea

LiteBreeze re-architected the pipeline into a modular, serverless, event-driven system on AWS:

• AWS Lambda (Python) replaced sequential scripts, enabling parallel execution
• Workflow steps (ingestion, OCR, metadata extraction, indexing) were decoupled and orchestrated via AWS Step Functions
• Built-in error handling, retries, and notifications improved reliability
• Serverless design removed infrastructure management, cutting operational overhead

This allowed the client to achieve immediate scalability and handle high volumes efficiently.

Implementation Process

The migration was phased to minimize risk:

• Pilot Phase → Rebuilt only the ingestion module in AWS Lambda to test feasibility
• Expansion Phase → Gradually integrated OCR, metadata, and indexing into separate Step Functions
• Automation → Infrastructure provisioned via AWS SAM (Serverless Application Model) and CloudFormation for version-controlled deployments
• Cutover → Full production switch after successful staging validation, with minimal disruption

Results and Benefits

The new system delivered measurable improvements:

• Scalability: From a few hundred → ~10,000 documents per day
• Flexibility: Handled both steady ingestion and sudden spikes in volume
• Cost Efficiency: Serverless design scaled down during off-peak hours, reducing costs
• Observability: AWS CloudWatch + Slack alerts improved visibility and incident response

Editorial workflows became faster, more reliable, and cost-efficient.

Conclusion

By moving to a serverless, event-driven architecture on AWS, ActaPublica:

• Eliminated backlogs from sequential processing
• Achieved high-concurrency document handling
• Reduced operational costs with on-demand scaling
• Gained better monitoring and real-time alerting

This project established a scalable, resilient, and cost-effective foundation for ongoing document ingestion and archival.

Technologies Used

• Programming Language: Python
• Serverless Compute and Workflow Orchestration: AWS Lambda, AWS Step Functions
• Document Processing and OCR: AWS Textract
• Search and Indexing: Elasticsearch
• Relational Data Storage: AWS RDS
• Crowdsourced Metadata Extraction: AWS Mechanical Turk (MTurk)
• Monitoring and Alerts: AWS CloudWatch, Slack integration
• Infrastructure as Code: AWS Serverless Application Model (SAM)