Image and File Reader

Goal

Extract readable content from image and file inputs using local tools so the model can continue analyzing without native multimodal input.

Tools

• scripts/read_file.py — one command to inspect one or many files.

Core Workflow

1. Choose files and run the reader.
2. If the file is image format, use OCR mode (--ocr) when needed.
3. Return the extracted text (or clear error messages) for downstream reasoning.

Command Patterns

Read one file:

python3 scripts/read_file.py ./screenshot.png --ocr

Read multiple files and output JSON:

python3 scripts/read_file.py ./a.png ./b.pdf ./note.txt --ocr --json

Use JSON mode when you need consistent parsing for scripts/agents.

python3 scripts/read_file.py ./doc.pdf --json > doc_payload.json

Common flags

• --ocr : Enable OCR for image files.
• --ocr-lang : OCR language (default eng+chi_tra).
• --max-chars : Truncate long text output for a safer payload.
• --json : Output machine-readable JSON object.

Notes

• OCR needs one of:
  • tesseract CLI (preferred)
• PDF text extraction prefers pdftotext CLI.
• If no extractor is available, the tool will return a clear, actionable error and suggestions.

Output Rules

• Keep only extracted content and metadata needed for triage.
• Preserve raw text and mention truncation when it happens.
• Never invent content.

Next-Step Guidance

If OCR confidence is low or output is empty:

• Re-run with a higher-resolution image.
• Pre-process image (crop/contrast) before OCR.
• Provide a manual transcript if needed.