Skip to main content

Document Ingestion & Processing

Aragora ingests documents as evidence for debates, audits, and workflows. The DocumentConnector uses the unified DocumentParser to extract text, tables, and metadata into structured evidence objects.

Supported Formats​

The parser supports common business and technical formats, including:

  • Office: PDF, DOC/DOCX, XLS/XLSX, PPT/PPTX
  • Text: TXT, Markdown, RST, HTML, RTF
  • Structured data: JSON, YAML, XML, CSV
  • Notebooks: IPYNB
  • E-books: EPUB, MOBI
  • Archives: ZIP, TAR, GZIP (expanded before parsing)

Source: aragora/connectors/documents/parser.py

Connector Usage​

from aragora.connectors.documents import DocumentConnector

connector = DocumentConnector(max_pages=100, extract_tables=True)

# Parse a file from disk
results = await connector.search_file("/path/to/report.pdf")

# Parse raw bytes
with open("policy.docx", "rb") as f:
    results = await connector.search_bytes(f.read(), filename="policy.docx")

# Search within parsed content
matches = await connector.search("access control", limit=5)

Source: aragora/connectors/documents/connector.py

Where Documents Flow​

Configuration Notes​

Document parsing limits and table extraction are configured in the connector constructor. For production workloads, tune page limits and max content size to match your document sizes and compute budget.