Document Processing (AKIOS V1.0)

Automate ingestion, extraction, and analysis with the cage.

Quick start

./akios init my-doc-project
./akios run templates/document_ingestion.yml

Supported inputs

• PDF (native + OCR fallback)
• DOCX
• TXT (encoding detection)
• Images (OCR)

Basic pipeline

name: "Document Analysis"
steps:
  - step: read_doc
    agent: filesystem
    action: read
    config: {allowed_paths: ["./data/input"]}
    parameters: {path: "./data/input/contract.pdf"}

  - step: analyze
    agent: llm
    action: complete
    parameters:
      prompt: |
        Extract parties, dates, financial terms, risks:
        {{read_doc.content}}

  - step: save
    agent: filesystem
    action: write
    config: {allowed_paths: ["./data/output"]}
    parameters:
      path: "./data/output/contract_analysis.txt"
      content: "{{analyze.text}}"

Patterns

• Batch: list files then process with tool_executor/parallel (limit procs).
• Routing: classify via LLM, then move to a category folder with tool_executor.
• PII: redaction is automatic; keep audit enabled for evidentiary trails.

Tips

• Keep inputs under data/input/; outputs under data/output/.
• Use size checks to switch to summary-only processing on very large files.
• Store OCR/parse failures separately for reprocessing.