Faster exports, saved images, and a smoother overall experience.

• Faster conversation loading. Capsula now reads ChatGPT's conversation JSON first,
  reducing the need to scroll through long chats. Page-based collection remains available
  when the API cannot be used.
• Richer thinking details. Exports preserve available thinking summaries, step labels,
  and timing information, alongside message timestamps and model information.
• Images that stay with your export. Available images are downloaded and embedded in
  HTML and Markdown files and clipboard copies, so successfully saved images remain after
  ChatGPT's links expire. Embedding is enabled by default in Settings → Behavior.
• Clear loading progress. A conversation overlay shows collection progress and message
  counts, and covers scroll movement when Capsula needs to collect the page.
• Move the export button. Drag it to a comfortable position. Capsula remembers it across
  sessions, keeps it inside the window on resize, and offers Reset to corner in Settings →
  Appearance.
• A more reliable export panel. Fixed panels disappearing after a quick reopen and
  closing during URL changes that keep the same conversation.

Image downloads may ask for optional access to ChatGPT's image hosts. If an image cannot be
embedded, its original link remains. Markdown viewers need data URI support to display
embedded images; JSON and GitHub/Notion exports continue to use image URLs.

All export processing stays in your browser. No telemetry or Capsula servers.

Capsula v1.3.0 - Enhanced Selection & Tutorial Release

New: Advanced message selection, interactive tutorials, and a modern SVG icon system!



What's New in 60 Seconds

Interactive Tutorial System

• Context-based help for first-time users
• Versioned tutorials that survive domain shifts
• Non-intrusive 2-step guides for GitHub/Notion features
• Dismiss with "Got it!" or "Don't show again"
• Reset option in Settings > Help & Tutorials

Enhanced Timeline with Multi-Selection

• Ctrl+Click - Toggle individual messages on/off
• Shift+Click - Extend selection from last anchor
• Ctrl+Shift+Click - Add range to existing selection
• Ctrl+Drag - Add/remove multiple messages while dragging
• Right-Click - Clear all selections
• Non-contiguous selections for complex filtering

Interactive Message Toggles

• Checkbox on each message in preview area
• Click to include/exclude from export
• Visual feedback (green checkmark = included)
• Bidirectional sync with timeline
• No keyboard shortcuts required

Smart Message Visibility

• Greyed-out display for excluded messages
• Auto-collapse to 60px height
• Click to expand/collapse for review
• Clear "(Not in export)" labels
• Preserved message alignment (user right, assistant left)

Modern SVG Icon System

• Professional SVG icons replace all emojis
• Automatic light/dark mode adaptation
• Better accessibility and cross-platform compatibility
• Cleaner, more modern interface
• Perfect rendering at all sizes

Technical

• Set-based selection model for O(1) lookups
• Tutorial state in browser.storage.local (survives domain shifts)
• Event propagation control for nested interactions
• RequestAnimationFrame for smooth rendering
• SVG icons use 'currentColor' for theme adaptation

Upgrade

Firefox: Auto-update from Add-ons Manager

Manual: Load new manifest.json from about:debugging

v1.2.0 – November 6th, 2025

🚀 New Integrations

• GitHub Gist Export: Create public/private gists from conversations with progress tracking, retries, and token validation.
• Notion Export: Send chats to Notion pages or databases with rich formatting, code blocks, and block-level progress tracking.
• Integration Settings: Manage GitHub/Notion tokens securely (local encryption, no telemetry), test connections, and control permissions.

⚙️ Infrastructure

• Background API Broker: CSP-safe background requests, proper permission prompts, and message-passing architecture.
• Privacy & Security: Tokens stored locally, requested only when needed, no data sent to external servers.

✨ Improvements

• Timeline and content filters now apply to all exports.
• Enhanced modal input focus and keyboard handling.
• Clear, user-friendly error messages with retry logic and rate-limit detection.

🧩 Fixes

• Modal focus loss during typing.
• CSP and JSON export issues.
• Filtered messages not respected in exports.
• Duplicate export button bug.

🎨 UI/UX

• New GitHub/Notion export buttons, icons, and progress bars.
• Success states with clickable URLs.
• Polished modal design and settings UI.

Changelog

[1.1.0] - 2025-11-03

Fixed

• Fixed role detection preventing user messages from incorrectly getting thinking labels
• Fixed thinking detection to use structural DOM selectors instead of text search
• Fixed thinking labels appearing in wrong location in exports (now properly placed with assistant messages)
• Fixed thinking time extraction (was showing 0s, now correctly parses "1m 38s" format)
• Fixed code block counting accuracy
• Fixed memory issues

Added

• Canvas artifact detection with title and type metadata
• Canvas artifact markers in Markdown exports
• Canvas artifact markers in HTML exports
• File attachment detection for user messages
• File attachment markers in Markdown exports
• File attachment markers in HTML exports
• Multi-stage thinking sequence preservation (captures all thinking stages in order)
• Dashboard metrics for canvas artifacts (count, types)
• Dashboard metrics for file attachments (count, by type: image/pdf/archive/document/code)
• Dashboard metrics for multi-stage thinking sequences
• Dashboard section for canvas artifacts
• Dashboard section for file attachments

Changed

• Improved thinking detection to only search structural divs (.relative.my-1.min-h-6)
• Enhanced role detection to check data-turn attribute first (most reliable)
• Updated message processing order: role → type-specific → thinking → content
• Improved export format with proper metadata placement