Document Attachments

​

Overview

​

Supported Formats

​

Text Files

• .txt — Plain text
• .md — Markdown
• .log — Log files

​

Code Files

• JavaScript/TypeScript: .js, .jsx, .ts, .tsx
• Python: .py
• Java/Kotlin: .java, .kt
• C/C++: .c, .cpp, .h
• Swift: .swift
• Go: .go
• Rust: .rs
• Ruby: .rb
• PHP: .php
• SQL: .sql
• Shell: .sh

​

Data Files

• .csv — Comma-separated values
• .json — JSON data
• .xml — XML documents
• .yaml / .yml — YAML configuration
• .toml — TOML configuration
• .ini / .cfg / .conf — Config files
• .html — HTML documents

​

PDF Documents

• .pdf — Native text extraction (see below)

​

How to Use

​

Attaching a Document

1. Open a conversation
2. Tap the paperclip icon in the chat input
3. Select a file from your device using the native file picker
4. The document appears as a badge in the input area
5. Type your question about the document
6. Send

​

Multiple Attachments

• Tap the paperclip again to add more files
• Each document appears as a separate badge
• All documents are included in the AI’s context

​

Viewing Attachments

• iOS: Opens in QuickLook (native preview)
• Android: Opens with Intent.ACTION_VIEW (your device’s default viewer)

​

Removing Attachments

• Tap the x on any document badge to remove it
• Removed documents are not included in the message

​

PDF Text Extraction

​

iOS: PDFKit

• Uses Apple’s PDFKit framework (built-in)
• Extracts text page-by-page with page separators
• Handles most standard PDFs

​

Android: PdfRenderer

• Uses Android’s native PdfRenderer API (Kotlin)
• Extracts text page-by-page with page separators
• Handles most standard PDFs

​

Limitations

• Image-based PDFs: Text extraction only works on PDFs with selectable text. Scanned documents (images) won’t extract text.
• Password-protected PDFs: Not supported (will fail gracefully)
• Corrupted PDFs: May fail to extract text

​

File Size Limits

• Max file size: 5MB per file
• Max text content: 50,000 characters per file
• Files larger than 5MB are rejected
• Text content longer than 50K characters is truncated with a note

​

Persistent Storage

​

Storage Location

{DocumentDirectory}/attachments/
  ├── 1701234567_document.pdf
  ├── 1701234890_code.py
  └── ...

​

Why Persistent?

• View attachments later (even if the original file is deleted)
• Export conversations with attachments intact
• Re-open documents from old messages

​

Cleanup

1. Delete conversations containing large attachments
2. Or manually delete files from {DocumentDirectory}/attachments/ (requires file manager access)

​

Use Cases

​

Code Review

1. Attach script.py
2. Ask: “Review this code and suggest improvements”
3. The AI reads the code and provides feedback

​

Data Analysis

1. Attach sales_2024.csv
2. Ask: “What are the top 3 products by revenue?”
3. The AI parses the CSV and answers your question

​

Document Q&A

1. Attach contract.pdf
2. Ask: “What is the termination policy?”
3. The AI extracts text from the PDF and finds the relevant section

​

Configuration Help

1. Attach config.yaml
2. Ask: “Why isn’t my database connection working?”
3. The AI reads the config and spots the issue

​

Log Analysis

1. Attach crash.log
2. Ask: “What caused this crash?”
3. The AI analyzes the stack trace and suggests fixes

​

Tips

​

Getting the Best Results

1. Attach only relevant files — Don’t attach entire codebases, just the file(s) related to your question
2. Ask specific questions — “Find the bug in line 47” vs. “Review this code”
3. Use small files — Under 10K characters works best
4. Combine with text — Explain what you’re looking for in your message
5. Try PDFs carefully — Ensure the PDF has selectable text, not just images

​

Optimizing Context Usage

• Increase context length (Settings → Model Settings → Context Length)
• Attach fewer files per message
• Truncate large files manually before attaching (copy relevant sections to a new file)
• Clear conversation history periodically (start a new chat)

​

Troubleshooting

• Check storage permissions (Android: Settings → Apps → Off Grid → Permissions → Storage)
• Ensure you have a file manager app installed

• Ensure the PDF has selectable text (not a scanned image)
• Try opening the PDF in another app to verify it’s not corrupted
• Convert the PDF to text manually (copy/paste into a .txt file)

• Files over 5MB are rejected
• Try compressing the file or extracting only the relevant section

• Ensure you asked a specific question about the document
• Check that the document was attached (badge should appear in the input area)
• Try rephrasing: “According to the attached file, what is…”

• The original file may have been moved or deleted
• Check if the file still exists in device storage
• Re-attach the document if needed

​

Technical Details

​

Content URI Resolution (Android)

1. Copying the file to a temp cache path
2. Reading the temp file
3. Copying to persistent storage
4. Cleaning up the temp file

​

Text Truncation

if (raw.length > maxChars) {
  return `${raw.substring(0, maxChars)}\n\n... [Content truncated due to length]`;
}

​

Document Viewer Integration

• iOS: QuickLook framework (supports PDF, text, images, Office docs, etc.)
• Android: Intent.ACTION_VIEW (opens with the system’s default app for that file type)

​

Privacy

• Your files never leave your device
• No cloud uploads
• No data collection
• Works completely offline (after model download)