File Attachments

Work with images and documents in your conversations.

Overview

Astronox supports attaching files to your messages, enabling:

• Image analysis and OCR
• Document processing
• Visual problem-solving
• Context-rich conversations

Supported File Types

Images

Formats:

• JPG/JPEG
• PNG
• GIF
• WebP
• SVG (limited)

Use cases:

• Screenshots (analyze UI, extract text)
• Photos (describe content, identify objects)
• Diagrams (explain architecture, debug)
• Error messages (read and troubleshoot)
• Charts/graphs (extract data)

Documents

Formats:

• PDF (text extraction)
• TXT (plain text)
• MD (Markdown)
• CSV (data processing)
• JSON (analysis)

Use cases:

• Document summarization
• Data extraction
• Format conversion
• Content analysis

Code Files

Any text-based code:

• .py, .js, .ts, .java, .cpp, etc.
• Configuration files
• Logs

Use cases:

• Code review
• Bug finding
• Refactoring suggestions
• Documentation generation

Attaching Files

Method 1: Click Attach Button

Steps:

1. Click 📎 paperclip icon (next to send button)
2. Select file(s) from picker
3. Preview appears below input
4. Type message (optional)
5. Click Send

Preview:

┌─────────────────────────────────────┐
│  📎 Attachments (2)                 │
│                                     │
│  🖼️ screenshot.png    800 KB   [✕] │
│  📄 document.pdf      2.4 MB   [✕] │
└─────────────────────────────────────┘

Click [✕] to remove before sending

Method 2: Drag and Drop

Steps:

1. Drag file(s) from Finder/Explorer
2. Drop onto chat input area
3. Preview appears
4. Type message and send

Visual feedback:

• Drop zone highlights when dragging
• Shows "Drop files here" message

Method 3: Paste from Clipboard

For images:

1. Copy image (screenshot, photo)
2. Click in input field
3. Cmd/Ctrl + V to paste
4. Image appears as attachment
5. Send

Great for:

• Quick screenshot sharing
• Copying images from web
• Pasting from image editors

Using Attachments

Image Analysis

Basic description:

[Attach screenshot.png]
"What's in this image?"

AI: This is a screenshot of a code editor showing:
    • Python code (Flask application)
    • Syntax error on line 47 (missing closing parenthesis)
    • File path: /Users/you/project/app.py
    • The error message suggests...

Text Extraction (OCR)

[Attach photo of whiteboard]
"Extract all text from this image"

AI: Text found:

    Project Timeline
    • Phase 1: Dec 20-31
    • Phase 2: Jan 1-15
    • Phase 3: Jan 16-31

    Notes:
    - Need approval before Phase 2
    - Budget review on Jan 5

Problem Solving

[Attach error screenshot]
"This error keeps happening. How do I fix it?"

AI: I see the error: "Module not found: react-router-dom"

    This means the package isn't installed. Fix:

    1. Run: npm install react-router-dom
    2. Or add to package.json dependencies
    3. Restart dev server

    The import on line 3 should work after this.

Diagram Explanation

[Attach architecture diagram]
"Explain this system architecture"

AI: This diagram shows a microservices architecture:

    1. Frontend (React)
       ↓
    2. API Gateway (Node.js)
       ↓ ↓ ↓
    3. Three services:
       • User Service (Python)
       • Payment Service (Go)
       • Notification Service (Node)
       ↓
    4. Databases (MongoDB, PostgreSQL)

    Flow: ...

Document Processing

[Attach invoice.pdf]
"Extract all line items and totals from this invoice"

AI: Invoice Analysis:

    Invoice #: INV-2025-0421
    Date: Dec 20, 2025

    Line Items:
    1. Web Hosting      $29.99
    2. Domain Renewal   $14.99
    3. SSL Certificate  $49.99

    Subtotal: $94.97
    Tax:      $8.55
    Total:    $103.52

Code Review

[Attach app.py]
"Review this code and suggest improvements"

AI: Code Review for app.py:

    ✅ Strengths:
    • Good function naming
    • Proper error handling
    • Clean structure

    ⚠️ Issues:
    • Line 47: SQL injection risk (use parameterized queries)
    • Line 103: Hardcoded API key (use env vars)
    • No input validation on user_input

    🔧 Suggestions:
    • Add type hints
    • Extract database logic to separate module
    • Add unit tests

    [Detailed recommendations follow...]

Attachment Limits

Size Limits

Per file:

• Maximum: 20 MB per file
• Recommended: <5 MB for faster processing

Per conversation:

• Total: 8 MB for all attachments combined
• Exceeding shows: "Attachment limit reached"

Why limits:

• API token constraints
• Processing time
• Memory usage

File Count

Per message:

• Maximum: 10 files
• More than needed? Combine or send multiple messages

Format Limitations

Not supported (yet):

• Video files (mp4, mov, etc.)
• Audio files (mp3, wav, etc.)
• Office files (docx, xlsx - upload as PDF instead)
• Archives (zip, tar, etc. - extract first)
• Executables

Best Practices

✅ Crop Screenshots

Before attaching:

• Crop to relevant area only
• Reduces file size
• AI focuses on important parts
• Faster processing

Tools:

• macOS: Cmd+Shift+4 (select area)
• Windows: Snipping Tool

✅ Use Appropriate Resolution

Images:

• Screenshots: 1x or 2x resolution (not 3x)
• Photos: Resize if >2 MB
• Don't need 4K for text extraction

Tools:

• macOS Preview: Tools → Adjust Size
• Windows Paint: Resize
• Online: TinyPNG, CompressJPEG

✅ Clear File Names

Good:

error-message-login-page.png
architecture-diagram-v2.png
invoice-dec-2025.pdf

Avoid:

Screen Shot 2025-12-20 at 10.43.23 AM.png
IMG_1234.jpg
download (3).pdf

Why: AI can use filename for context

✅ Provide Context

Don't just attach silently:

❌ [Attaches image without message]

✅ [Attaches] "This error appears when I try to login"
✅ [Attaches] "Explain this network diagram"
✅ [Attaches] "Extract all dates from this document"

✅ Multiple Related Files

Batch attach for comparison:

[Attach: before.png, after.png]
"What changed between these two screenshots?"

[Attach: invoice1.pdf, invoice2.pdf]
"Compare these invoices and find differences"

❌ Don't Attach Sensitive Data

Avoid:

• Screenshots with passwords visible
• Documents with SSN, credit cards
• Private API keys or tokens
• Confidential company information

Remember: Attachments sent to AI provider (Gemini/MintAI)

Troubleshooting

Attach Button Not Working

Known issue (currently under repair)

Workaround:

Paste images: Cmd/Ctrl + V
Or: Ask AI to read from file path

"File Too Large" Error

Solutions:

1. Compress/resize image
2. Convert to smaller format (PNG → JPG)
3. Crop to relevant section
4. Split into multiple files

"Unsupported Format"

Try:

1. Convert to supported format
   • DOCX → PDF
   • Video → Screenshot key frames
2. Extract text manually and paste
3. Upload to cloud, share link (future)

Image Not Analyzed Correctly

Improve results:

1. Higher contrast/brightness
2. Crop to relevant area
3. Remove background noise
4. Use better camera/screenshot
5. Provide context in message

Example:

Instead of: "What's this?"
Try: "This is a Python error. What's wrong?"

Attachment Limit Reached

8 MB total per conversation

Solutions:

1. Start new conversation
2. Remove old attachments (if possible - future feature)
3. Compress files
4. Use file paths instead (AI reads from disk)

Alternative: File Paths

Instead of Attaching

For local files:

"Read the file at /Users/me/Documents/report.pdf"
"Analyze the image at ~/Desktop/screenshot.png"
"Show me the content of ./config.json"

Benefits:

• No size limits
• Faster (no upload)
• Works with any file type AI can read

Limitations:

• File must be on local system
• AI doesn't "see" binary files (images need attach or vision)

Future Enhancements

Coming soon:

• Attachment manager: View/delete old attachments
• Drag multiple files at once
• Video frame extraction (analyze key frames)
• Audio transcription (speech-to-text)
• Office document support (direct DOCX, XLSX reading)
• Cloud file integration (Google Drive, Dropbox)
• Attachment compression (auto-optimize large files)

Platform-Specific Tips

macOS

Best screenshot tool:

Cmd + Shift + 4: Select area
Cmd + Shift + 3: Full screen
Cmd + Shift + 4, then Space: Window

Screenshots auto-save to Desktop

Windows

Snipping Tool:

Win + Shift + S: Select area
Opens snipping overlay

Or: Use built-in Snip & Sketch

Next: Learn Best Practices for optimal Astronox usage.