Limitations & Known Issues

What Astronox can and cannot do (as of Beta version).

Current Limitations

Internet & Web Browsing

❌ Cannot:

• Browse the web actively
• Search Google or other search engines
• Download files from URLs
• Scrape websites
• Access online APIs (except LLM providers)

✅ Can:

• Check for Astronox updates
• Validate license status
• Fetch API keys (Pro)
• Access LLM APIs (Gemini, MintAI)

Workaround:

• Download files manually → ask Astronox to process local file
• Copy web content → paste in chat for analysis

Communication

❌ Cannot:

• Send emails or messages
• Post to social media
• Make HTTP requests to arbitrary URLs
• Connect to external databases
• Access cloud storage directly (Dropbox, Google Drive, etc.)

Why: Security and privacy by design

Future: Planned integrations with user permission

Application Control

❌ Limited control over:

• Apps' internal state (can launch/quit, but not control features)
• Keyboard/mouse simulation (future: with permission)
• Other apps' windows/UI elements
• System preferences (some settings inaccessible)

✅ Can:

• Launch applications
• Quit/kill processes
• Monitor resource usage
• Manage files used by apps

Platform Differences

macOS Issues

Brightness control:

• M4 chips: Brightness commands may not work reliably
• Workaround: Manual adjustment or use alternative tools
• See: docs/OS-Settings/BRIGHTNESS_ISSUE_RESOLUTION.md

OS Settings panel:

• Some features cause UI issues
• Currently hidden from stable UI
• Power saver toggle: Can cause black screen (bug in Tauri)

Windows Issues

Performance:

• Slightly slower file operations vs macOS
• Windows Defender may slow scans

Permissions:

• May require Admin for system operations
• UAC prompts more frequent
• May need manual configuration

File Operations

❌ Cannot (yet):

• Modify files open in other apps reliably
• Access encrypted drives/volumes
• Recover deleted files from trash (shows, doesn't restore)
• Network drive operations (slower, may fail)

Size limits:

• Attachments: 20 MB per file, 8 MB total per conversation
• Text files: Can read very large files (GB+) but may be slow
• Binary files: Limited analysis

Voice & Audio

❌ Not yet implemented:

• Voice input (speech-to-text)
• Voice output (text-to-speech)
• Audio file transcription

Status: /api/transcribe endpoint returns 501 (not implemented)

Future: Planned for upcoming releases

Known Issues

UI Issues

Attach Button:

• Status: Non-functional (under repair)
• Workaround: Paste images with Cmd/Ctrl+V, or use file paths
• ETA: Fix in next release

Plan Hub:

• Issue: Sometimes unresponsive or doesn't refresh
• Workaround: Close and reopen panel
• Status: Investigating

Settings Panel:

• Issue: Occasionally doesn't save on first try
• Workaround: Click Save again, or restart app
• Status: Intermittent

Backend Issues

Intent Parser Tests:

• Status: 2 of 27 tests failing in intent_parser.rs
• Impact: Minimal (doesn't affect runtime)
• Priority: Low (will fix in next release)

Memory Persistence:

• Issue: Memory sometimes duplicates facts
• Workaround: Manually review and delete duplicates
• Status: Low priority

Clipboard History:

• Issue: Lost on app crash/restart (by design)
• Limitation: Not persisted to disk
• Future: Optional persistent history planned

Performance Issues

Large Conversations:

• Issue: Slow response times after ~100 messages
• Cause: Token context grows large
• Workaround: Start new conversation

File Search:

• Issue: Slow on very large directories (>10k files)
• Workaround: Provide specific path, not entire system
• Status: Optimizing

Memory Usage:

• Issue: High memory usage after extended use
• Workaround: Restart app periodically
• Status: Investigating leaks

API & Provider Limitations

Gemini Specific

Rate Limits:

• Free tier: 60 req/min, 1500 req/day (Flash/Flash Lite)
• Free tier: 10 req/min, 500 req/day (Pro)
• See Rate Limits Guide

Context Size:

• Flash/Lite: 128k tokens (~100k words)
• Pro: 2M tokens (~1.5M words)
• Exceeding causes truncation

Vision Limitations:

• May misidentify complex images
• OCR accuracy varies with image quality
• Some formats not fully supported

MintAI Specific

Availability:

• Requires Pro subscription
• Auto-fetch may fail if subscription inactive
• Manual key entry not supported (by design)

Models:

• Currently only Devstral 2 available
• Future: More models planned

Feature Completeness

Beta Status Features

Production Ready:

• ✅ File management
• ✅ System monitoring
• ✅ Text processing
• ✅ Code generation
• ✅ Basic automation

Beta Quality:

• ⚠️ Multi-step plans (mostly stable)
• ⚠️ Memory system (occasional duplicates)
• ⚠️ Clipboard history (not persisted)
• ⚠️ Image analysis (good but not perfect)

Experimental:

• 🧪 OS system settings (hidden due to bugs)
• 🧪 Process guard (limited testing)
• 🧪 Proactive suggestions (disabled)

Not Implemented:

• ❌ Voice transcription
• ❌ Web browsing
• ❌ Email integration
• ❌ Cloud sync

Security Limitations

By Design (For Your Protection):

Cannot access:

• Password managers (Keychain, 1Password, etc.)
• Encrypted volumes
• Other apps' internal data
• System keychain entries
• Browser stored passwords

Cannot modify:

• System critical files (protected by OS)
• Files you don't have permission for
• Running processes you don't own

Data Limitations

Conversation History

Per conversation:

• Message limit: ~1000 messages (before slowdown)
• Attachment limit: 8 MB total
• No cloud sync (local only)

Persistence:

• Conversations saved locally
• Lost if app data deleted
• No backup/restore (yet)

Memory System

Limits:

• Total facts: 1,000
• Fact size: 500 characters each
• No cloud sync
• Lost if memory.json corrupted

Automation

Saved automations:

• No limit on count
• Limited to local system
• No sharing/export (yet)
• No cloud sync

Accuracy Limitations

Code Generation

Expected accuracy:

• Simple scripts: ~98%
• Complex code: ~90-95%
• Production code: Always review!

Common issues:

• May use deprecated APIs
• Edge cases not always handled
• Newer language features may be missed

Always:

• Review generated code
• Test before deploying
• Add error handling

File Operations

Accuracy:

• Path resolution: 99%
• File detection: 98%
• Content analysis: 95%

Edge cases:

• Symlinks may confuse AI
• Hidden files sometimes skipped
• Very long paths (>250 chars) may fail

System Information

Generally accurate:

• Disk space, memory, CPU usage
• Running processes
• File metadata

May be imprecise:

• Network speed (varies moment to moment)
• Battery estimates (depends on usage)
• Temperature readings (sensor-dependent)

What Won't Be Fixed (By Design)

No Cloud Features (Privacy)

Won't add:

• Cloud conversation sync
• Online backup
• Telemetry/tracking
• Account system

Reason: Privacy-first design. Your data stays on your device.

No Direct Internet Access

Won't add:

• General web scraping
• Arbitrary HTTP requests
• Direct API calls (except LLMs)

Reason: Security. Prevents data leakage.

No Password/Secret Access

Won't add:

• Keychain access
• Password manager integration
• Encrypted volume mounting

Reason: Security. AI should never handle passwords.

Reporting Issues

How to Report

1. Check this doc first - Issue may be known
2. Check FAQ - May have solution
3. Enable debug mode - Settings → Advanced → Debug
4. Collect logs - ~/Library/Logs/Astronox/ (macOS)
5. Report via official channels - (contact info in app)

What to Include

Always provide:

• Astronox version (Settings → About)
• OS version
• Steps to reproduce
• Expected vs actual behavior
• Screenshots (if UI issue)
• Logs (if crash/error)

Workarounds Summary

Future Improvements

Planned for upcoming releases:

• Fix attach button
• Optimize large conversations
• Persistent clipboard history (opt-in)
• Voice input/output
• Web browsing (limited)
• Cloud sync (optional, encrypted)
• Mobile companion app
• Plugin system

Timeline: Check Release Notes for updates

Next: Review Rate Limits & Costs for usage planning.