pageindex-grounded

Build and cache per-paper PageIndex trees for grounded question answering over local PDFs.

11|1|Updated Feb 23, 2026

Search Tags:#provenance #cross-paper #research-workflow #grounded-qa #local-papers #pagindex

Ready to run

npx skills add https://github.com/jimezsa/opencolab --skill pageindex-grounded

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: pageindex-grounded
Source: https://github.com/jimezsa/opencolab/tree/main/projects/SKILLS/pageindex-grounded
Command: npx skills add https://github.com/jimezsa/opencolab --skill pageindex-grounded

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Local researchers need precise, evidence-backed answers from a bounded set of locally stored PDFs. This skill builds and caches per-paper PageIndex trees to enable fast, verifiable retrieval and prevents overreach beyond the selected documents.

Core Features & Use Cases

Build and reuse per-paper PageIndex trees for fast, exact retrieval against local PDFs.
Maintain a bounded paper set per question, enabling focused cross-paper comparisons with provenance.
Return exact paper references and page ranges to ground answers and support verification.

Quick Start

Ask it to ground a question against a specific local paper and page range.

Frequently Asked Questions about pageindex-grounded

High-intent search queries and answers about installing and using this skill.

FAQPage Schema

How do I get grounded answers from local PDFs with exact page references?▼

Grounded QA over local PDFs is achieved by building per-paper PageIndex trees that return exact paper references and page ranges to support verification. The skill maintains a bounded paper set per question to prevent overreach beyond selected documents.

Can I verify exact claims across multiple local research papers?▼

Exact-claim verification and cross-paper comparisons are supported by caching per-paper PageIndex trees within a bounded set of PDFs. This enables focused retrieval with provenance, returning exact references and page ranges to ground answers.

What do I need to set up before doing grounded question answering on local PDFs?▼

Grounded QA requires a local PageIndex runner, PDFs stored under research/pdf, optional metadata under research/meta, and a local cache under research/pageindex with a manifest.json to enable reuse of the per-paper trees.

Does grounded question answering reuse cached PageIndex trees for faster retrieval?▼

Cached PageIndex trees are built and reused for fast, exact retrieval against local PDFs. The local cache under research/pageindex uses a manifest.json to manage and enable reuse of these per-paper structures across multiple queries.

What are the limitations of cross-paper comparisons with a bounded set of local PDFs?▼

Cross-paper comparisons are limited to the bounded set of local PDFs selected per question to prevent overreach beyond the scoped documents. Retrieval and provenance-enabled answers are constrained to this specific local paper set.

What is the best way to prevent overreach when querying a local set of research PDFs?▼

The best way to prevent overreach is maintaining a bounded paper set per question during grounded QA. This approach ensures retrieval and cross-paper comparisons are strictly limited to the selected local PDFs with exact provenance.