Ask Questions About Any Document

You have a 50-page document and need specific information from it. Instead of reading the entire thing, ask the AI to find what you need.

The Quick Recipe

“Read the file at ~/Documents/contract.pdf and tell me: what are the payment terms, cancellation policy, and auto-renewal clauses?”

Chapeta reads the file, processes it through your chosen AI model, and answers your specific questions with references to the relevant sections.

Document Types That Work

Chapeta’s file read tool handles:

• PDF files: Contracts, reports, research papers, manuals
• Text files: .txt, .md, .csv, .json, .xml
• Code files: Any programming language
• Log files: Server logs, application logs
• Config files: .yaml, .toml, .ini, .env

The key requirement: the file must contain text (not scanned images). For image-based PDFs, you would need OCR first.

Q&A Patterns

Specific Information Extraction

“Read ~/Documents/annual-report.pdf. What was the total revenue for Q3?”

Comparison

“Read ~/Documents/proposal-v1.pdf and ~/Documents/proposal-v2.pdf. What are the key differences between these two versions?”

Compliance Check

“Read ~/Documents/privacy-policy.pdf. Does this policy comply with GDPR requirements? List any gaps.”

Technical Understanding

“Read ~/projects/myapp/README.md. What steps are needed to set up the development environment?”

Working with Large Documents

For documents within the context window (most files under 100 pages):

“Read ~/Documents/handbook.pdf and answer: what is the vacation policy for employees in their first year?”

For very large documents, use a model with a larger context window:

Switch to Gemini (2M tokens) or Claude (200K tokens) for documents that exceed GPT’s 128K token limit. See the model comparison for context window sizes.

Progressive Reading

For enormous files, read in sections:

“Read the first 50 pages of ~/Documents/manual.pdf and summarize the setup instructions”

Then:

“Now read pages 51-100 and summarize the configuration options”

Multi-Document Research

Cross-Reference

“Read ~/Documents/report-2025.pdf and ~/Documents/report-2024.pdf. How did the metrics change year over year?”

Aggregate

“Read all .txt files in ~/Documents/interviews/ and identify the most common themes across all interviews”

This uses Glob to find files, then reads each one, giving you a synthesized analysis.

Fact-Check

“Read ~/Documents/draft-article.md. Search the web for the claims made in the article and flag any that appear inaccurate.”

Combining document reading with web search for verification.

Save Your Analysis

“Read ~/Documents/contract.pdf, analyze the key terms, and save a summary to ~/Documents/contract-summary.md”

The analysis is preserved as a file you can reference later or share with colleagues.

Building a Document Q&A Skill

Create a reusable skill for document analysis:

Name: Document Q&A
System Prompt: Read the provided file. Answer questions about its content precisely,
referencing specific sections when possible. If the answer is not in the document, say so.
Do not make up information that isn't in the source material.
Tools: File Read, Glob, File Write

This skill enforces grounded answers rather than allowing the AI to speculate beyond the document content.

Limitations

Text extraction from PDFs depends on the PDF being text-based (not scanned images). Very large files may exceed the context window of your chosen model. The AI cannot access password-protected files. Binary files (images, videos, compiled code) are not supported through the file read tool; for images, use the Screenshot tool. The AI analyzes the document content that the file read tool extracts, so any formatting or layout information that does not survive text extraction will be missed.