---

name: page-index description: "Routes PageIndex PDF, flash, Markdown, and workspace retrieval workflows." disable-model-invocation: true metadata: disco-role: operating license: MIT

PageIndex

PageIndex turns PDFs and Markdown into hierarchical trees and exposes retrieval helpers for workspace-backed document Q&A.

Use this skill when the request mentions:

• generating a tree, outline, TOC, or section hierarchy from a PDF
• the fast no-LLM Flash path or embedded bookmarks
• converting Markdown headings into a tree
• optimizing an existing tree for search cost
• PageIndexClient, workspace storage, or document retrieval by page range
• the agentic vectorless RAG demo or tool-call based document QA

Start here

• Install the runtime dependency set listed in references/configuration.md.
• Make pageindex importable in the current Python environment before using the bundled scripts.
• Quick import smoke: python scripts/check_env.py
• Main CLI help: python scripts/pageindex_cli.py --help

Route map

• sub-skills/pdf-indexing/ — classic PDF tree extraction with the LLM-assisted pipeline.
• sub-skills/flash-indexing/ — fast PDF extraction with PageIndex Flash, embedded bookmarks, and merge-only optimization.
• sub-skills/markdown-indexing/ — Markdown heading trees, thinning, and optional summaries.
• sub-skills/retrieval-client/ — PageIndexClient, workspace persistence, page retrieval, and agentic Q&A patterns.

Shared references

Read these when you need the full details behind a route:

• references/workflows.md — end-to-end workflows and the right route for each task family.
• references/api-reference.md — verified public APIs and signatures.
• references/cli-reference.md — bundled CLI wrapper and pageindex.tree_optimize flags.
• references/data-formats.md — tree JSON, workspace JSON, and Markdown semantics.
• references/configuration.md — model defaults, environment variables, and optional extras.
• references/troubleshooting.md — common failures and recovery steps.
• references/repo-provenance.md — staleness and source snapshot information.

Selection hints

• PDF with TOC recovery, page-number reconciliation, summaries, or document descriptions -> pdf-indexing.
• Fast PDF structure from layout statistics, embedded bookmarks, or merge-only optimization -> flash-indexing.
• Markdown files with headings -> markdown-indexing.
• Workspace-backed retrieval, get_document, get_document_structure, get_page_content, or the agentic RAG demo -> retrieval-client.
• If a task mentions --optimize, start from flash-indexing; its reference material covers the merge-only and expand paths.

Operating notes

• The package is source-first in this repository. Do not assume a pip-installable distribution exists; install the runtime dependency set from references/configuration.md and use the bundled scripts in a path-aware environment.
• Live LLM or agentic runs still need the user's API key and network access.
• Prefer the smallest safe workflow that matches the request.
• If the repository has changed, read references/repo-provenance.md before updating or refreshing the skill.