Add conversation branching system that allows creating and exploring alternate conversation paths from any message point. Each branch maintains its own complete message history.
Backend (Rust):
- Branch data structures with backward-compatible storage migration
- Tauri commands for create, switch, delete, rename, and list operations
- Automatic cleanup of child branches when parent is deleted
Frontend:
- Branch button on all messages for creating new branches
- Branch indicator badge in header showing active branch
- Branch manager modal with full branch list and controls
- Visual improvements to message action toolbar (more opaque, positioned above messages)
Branches are character-specific and persist across sessions.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Add context limit field to API settings that:
- Stores context limit in config (defaults to 200000)
- Allows users to set custom limits for different models
- Uses configured limit in token counter display
- Shows format like "2.5k / 200k tokens" using actual limit
This allows proper token tracking for non-Claude models with
different context windows.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Add support for using mes_example field from character cards to teach the AI the character's voice and writing style. Examples are parsed, processed with template variable replacement, and injected into the context at a configurable position.
Backend changes:
- Extended RoleplaySettings with examples_enabled and examples_position fields
- Implemented parse_message_examples() to parse <START>-delimited example blocks
- Added example injection in build_api_messages() with position control
- Integrated examples into token counter with accurate counting
- Created update_examples_settings command for saving settings
Frontend changes:
- Added Message Examples UI controls in Author's Note tab
- Checkbox to enable/disable examples
- Dropdown to select injection position (after_system/before_history)
- Save button with success/error feedback
- Token breakdown now shows examples token count
- Settings load/save integrated with roleplay panel
Message examples help the AI understand character personality, speaking patterns, and response style by providing concrete examples of how the character should respond.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Added comprehensive message control features for fine-grained conversation management:
Backend (Rust):
- Extended Message struct with 'pinned' and 'hidden' boolean fields
- Added delete_message_at_index() command for removing any message
- Added toggle_message_pin() command to mark important messages
- Added toggle_message_hidden() command to temporarily hide messages
- Added continue_message() command to append AI continuations
- Added regenerate_at_index() command to regenerate ANY message (not just last)
Frontend (JavaScript):
- Added delete, pin, hide buttons to all messages (user & assistant)
- Added continue button for assistant messages
- Updated regenerate to work on any message, not just the last one
- Implemented state persistence for pinned/hidden in chat history
- Added dynamic icon changes for hide/unhide states
- Integrated with token counter for real-time updates
UI/UX (CSS):
- Pinned messages: accent-colored left border with glow effect
- Hidden messages: 40% opacity with blur effect (70% on hover)
- Delete button: red hover warning (#ef4444)
- Active state indicators for pin/hide buttons
- Always-visible controls on hidden messages for quick access
Features:
- Delete any message with confirmation dialog
- Pin messages to always keep them in context
- Hide messages with visual blur (still in context but dimmed)
- Continue incomplete assistant responses
- Regenerate any assistant message (creates new swipe)
- All states persist in chat history JSON
This completes Phase 3.2 "Enhanced Message Controls" from the roadmap,
providing users with complete control over their conversation history.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Add comprehensive token counting functionality to provide visibility into
context usage:
Backend (Rust):
- Add tiktoken-rs dependency for OpenAI-compatible token counting
- Implement get_token_count command with detailed breakdown
- Count tokens for: system prompt, preset instructions, persona, world info,
author's note, message history, and current input
- Per-section token breakdown for optimization insights
Frontend (JavaScript/HTML/CSS):
- Add token counter widget in status bar
- Real-time updates as user types (debounced 300ms)
- Expandable breakdown tooltip showing per-section counts
- Automatic update when chat history loads or changes
- Clean, minimal UI with hover interactions
Features:
- Accurate token counting using cl100k_base tokenizer
- Debounced updates for performance
- Detailed breakdown by context section
- Visual indicator with total token count
- Click to expand/collapse detailed breakdown
- Auto-hide when no character is active
This completes the "Must-Have for Basic Roleplay" features from the roadmap:
✅ World Info/Lorebooks
✅ Author's Note
✅ Token Counter
- Message Examples Usage (next)
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Add ability to edit built-in presets (Default, Roleplay, Creative Writing,
Assistant) while preserving original defaults:
- Built-in presets are now fully editable (system additions, author's note,
instruction blocks)
- Modifications are saved as overrides in ~/.config/claudia/presets/
- "Modified" badge appears when built-in preset has been customized
- "Restore to Default" button removes overrides and restores originals
- Backend commands: is_builtin_preset_modified, restore_builtin_preset
- All instruction blocks support expand/collapse and drag-and-drop reordering
Also update ROADMAP.md to reflect completed features:
- World Info/Lorebook System ✅
- Author's Note ✅
- User Personas ✅
- Regex Scripts ✅
- Chat History Import/Export ✅🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
This commit implements Options B, C, and D for enhanced roleplay capabilities:
**Option B: Enhanced Roleplay Features**
- Add regex pattern matching for World Info keywords with fallback to literal matching
- Make scan depth configurable (default: 20, range: 1-100 messages)
- Make Author's Note insertion depth configurable (default: 3, range: 1-50 messages)
- Add World Info import/export functionality with merge/replace options
- Update WorldInfoEntry struct with use_regex field
- Update RoleplaySettings with scan_depth and authors_note_depth fields
- Modify build_roleplay_context() to return configurable note_depth
- Update all 4 chat functions to use configurable Author's Note depth
**Option C: Chat Export Enhancements**
- Add export_chat_as_markdown: Clean markdown format with bold role labels
- Add export_chat_as_text: Simple plain text format with decorative separator
- Add export_chat_as_html: Styled HTML with responsive design and color-coded messages
- All exports use native file dialogs with appropriate file filters
**Option D: Polish & Quality of Life**
- Add validate_regex_pattern command for real-time regex validation
- Add update_roleplay_depths command with range validation
- Enhance add_world_info_entry with input validation (regex, empty checks)
- Enhance update_world_info_entry with input validation
- Add detailed error messages for all validation failures
**Technical Details:**
- Add regex dependency to Cargo.toml
- Use #[serde(default)] attributes for backward compatibility
- Graceful error handling with fallback strategies
- HTML export includes proper escaping for XSS protection
- All new commands registered in Tauri invoke_handler
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Add full context injection system for World Info, Author's Note, and Persona:
- Add scan_for_world_info() function to detect keywords in last 20 messages
- Add build_roleplay_context() to assemble all roleplay additions
- Inject Persona into system prompt as [{{user}}'s Persona: name - desc]
- Inject activated World Info entries into system prompt by priority
- Inject Author's Note as system message before last 3 messages
- Apply context injection to all 4 chat functions:
- chat() - regular non-streaming
- chat_stream() - streaming
- generate_response_only() - regenerate non-streaming
- generate_response_stream() - regenerate streaming
All roleplay features now fully functional and affecting AI generation.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Backend changes:
- Add RoleplaySettings and WorldInfoEntry data structures
- Implement per-character roleplay settings storage in ~/.config/claudia/roleplay_{id}.json
- Add Tauri commands for CRUD operations on World Info entries
- Add commands for saving Author's Note and Persona settings
Frontend changes:
- Add World Info entry management UI with add/edit/delete functionality
- Implement keyword-triggered context injection system with priority ordering
- Add Author's Note textarea with enable toggle
- Add Persona name and description fields with enable toggle
- Load roleplay settings when opening the roleplay panel
- Add CSS styles for World Info entry cards with hover effects
Features:
- World Info entries support multiple keywords, priority levels, and enable/disable
- Settings are per-character and persist across sessions
- Entries sorted by priority (higher priority injected first)
- Clean UI with edit/delete buttons and visual feedback
Implemented full chat history import/export with JSON format:
- Export button saves current conversation to JSON file
- Import button loads conversation from JSON file
- File dialog integration using tauri-plugin-dialog
- Message count feedback on successful import
- Automatic history reload after import
- Preserves all message data including swipes and timestamps
- Smart error handling (ignores cancelled dialogs)
Backend (Rust):
- export_chat_history: Opens save dialog, writes JSON to selected path
- import_chat_history: Opens file picker, parses JSON, saves to current character
- Message migration for backward compatibility
- Returns helpful feedback (file path on export, message count on import)
Frontend (JavaScript):
- Export/import buttons in header with up/down arrow icons
- Status updates during operations
- Auto-reload chat view after import
- Error handling with user-friendly messages
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Added support for all v2/v3 character card fields in the character editor UI:
- Description (10-row textarea)
- Scenario
- Message Example
- Post-History Instructions
- Alternate Greetings (one per line)
- Tags (comma-separated)
- Creator
- Character Version
- Creator Notes
Updated frontend to load and save all fields, and backend update_character
command to accept all new parameters. This allows imported v3 character cards
to display their full details, especially the description field which contains
the main character information.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Added read_png_text_chunks() function that manually parses PNG file structure
- Reads PNG chunks directly instead of relying on png crate's Info struct
- More reliable extraction of tEXt chunks with "chara" keyword
- Added debug logging (eprintln) to help diagnose import issues
- Updated read_character_card_from_png() to use manual parser as primary method
- Fixes "missing 'chara' chunk" error even when chunk exists in file
- Successfully imports both v2 and v3 character cards
- Added CharacterCardV3 struct to handle v3 format
- Created From<CharacterCardV3> for CharacterCardV2Data conversion
- Updated read_character_card_from_png() to detect and parse both v2 and v3 specs
- V3 cards have top-level fields + nested data object
- Maintains full backward compatibility with v2 cards
- Export still uses v2 format for maximum compatibility
- Expanded Character struct with all v2 spec fields (description, scenario, mes_example, post_history_instructions, alternate_greetings, character_book, tags, creator, character_version, creator_notes, extensions)
- Created CharacterCardV2 serialization structs following spec at github.com/malfoyslastname/character-card-spec-v2
- Implemented PNG metadata utilities:
* read_character_card_from_png() - extracts and decodes character data from PNG tEXt chunks
* write_character_card_to_png() - embeds character data into PNG files
* create_placeholder_png() - generates gradient placeholder images for avatarless characters
- Added Tauri commands:
* import_character_card - opens file picker, imports PNG with automatic name conflict handling
* export_character_card - exports character as v2 PNG card with embedded metadata
- Added Import/Export buttons to character settings UI
- Full backward compatibility with existing characters using serde defaults
- Added dependencies: png 0.17, base64 0.21, image 0.24
- Added swipe system to Message struct with backward compatibility
- Implemented swipe navigation UI with left/right arrows and counter
- Added generate_response_only and generate_response_stream commands
- Swipes persist properly when navigating between alternatives
- Updated message rendering to support swipe controls
- Add avatar upload with file picker dialog
- Display avatars in header, chat messages, and settings
- Implement clickable avatars with full-screen zoom modal
- Enable asset protocol for local file access in Tauri config
- Add tauri-plugin-dialog for native file selection
- Store avatars in ~/.config/claudia/avatars/
- Support PNG, JPG, JPEG, and WEBP formats
- Modal closes on overlay click or ESC key
- Add server-sent events (SSE) streaming support for real-time token display
- Implement progressive markdown rendering during streaming
- Add stream toggle in API settings (defaults to disabled for compatibility)
- Add visual streaming indicator with pulsing animation
- Graceful fallback to non-streaming mode when disabled
- Fix character saving bug (camelCase parameter naming)
Backend changes:
- New chat_stream command with SSE parsing
- Added futures and bytes dependencies
- Emit chat-token events progressively to frontend
- Support for OpenAI-compatible and Anthropic streaming formats
Frontend changes:
- Dual code paths for streaming/non-streaming
- Real-time markdown and syntax highlighting during streaming
- Stream status indicator with animation
