pdf-processing

Pdf Processing

A compliance department receives regulatory filings as PDF documents. Each PDF needs text extraction, metadata parsing (author, creation date, page count), content classification by section type, and indexing for full-text search. Scanned-image PDFs need a different extraction path than native-text PDFs.

Pipeline

[pd_extract_text]
     |
     v
[pd_parse_sections]
     |
     v
[pd_analyze_content]
     |
     v
[pd_generate_summary]

Workflow inputs: pdfBase64, pdfPath, options

Workers

AnalyzeContentWorker (task: pd_analyze_content)

Analyzes content of parsed sections — counts words and finds keywords.

• Lowercases strings, rounds with math.round(), formats output strings
• Reads sections. Writes wordCount, keywords, analysis

ExtractTextWorker (task: pd_extract_text)

Extracts text from a real PDF document using Apache PDFBox.

• Base64 encodes data
• Reads pdfPath, pdfBase64, pdfUrl. Writes error, text, pageCount, charCount

GenerateSummaryWorker (task: pd_generate_summary)

Generates a summary based on sections and analysis.

• Lowercases strings, uses java streams
• Reads sections, analysis. Writes summary

ParseSectionsWorker (task: pd_parse_sections)

Parses raw text into sections based on chapter headings.

• Trims whitespace, applies compiled regex
• Reads rawText. Writes sections, sectionCount

---

34 tests | Workflow: pdf_processing | Timeout: 60s

See RUNNING.md for setup and usage.