PDF Reader

Read PDFs with full text extraction and page rendering for diagrams/figures.

CLI

Prefer using the pre-built binary (available on all gLinux machines):

PDF_READER=/google/bin/releases/gemini-agents-pdf-reader/read_pdf_cli.par

Alternatively, install via apt (the read_pdf_cli.par binary will be available directly on PATH):

sudo glinux-add-repo -b gemini-agents-pdf-reader stable
sudo apt update && sudo apt install -y gemini-agents-pdf-reader

Use --version to check the build date. To verify the exact build CL: binfs ls /google/bin/releases/gemini-agents-pdf-reader.

If you modify the CLI source code, build from source:

blaze build //learning/gemini/agents/clis/pdf_reader:read_pdf_cli
PDF_READER=blaze-bin/learning/gemini/agents/clis/pdf_reader/read_pdf_cli

$PDF_READER \
  --url=URL \
  --output_dir=DIR \
  --output_file=FILE

The script outputs page-annotated text and renders each page as a PNG image. After running, use view_file on the rendered page images to inspect diagrams.

Library

from google3.learning.gemini.agents.clis.pdf_reader import read_pdf_lib

text = read_pdf_lib.extract_text('/path/to/doc.pdf')
pages = read_pdf_lib.render_pages('/path/to/doc.pdf', '/tmp/output')
output = read_pdf_lib.read_pdf('/path/to/doc.pdf', output_dir='/tmp/output')

Workflow

1. Run read_pdf_cli with --url or --file and --output_dir for page images
2. Read the text output for content understanding
3. Use view_file on specific page PNGs listed under RENDERED PAGES for diagrams, figures, or equations

Flags

Reporting Issues

Report bugs or improvements for this skill at Agent Skill: pdf-reader. See the skill_issue skill for instructions on filing and triaging skill bugs.