v5.4.1

[5.4.1] - 2026-02-28

Fixes

• File Processor: Fixed an issue where multi-modal message extraction was producing empty request bodies for image inputs.
• Batch API: Corrected logic for generating Gemini Batch API requests from within the File Processor.
• Batch TTS: Fixed an API timeout behavior and guaranteed the TTS internal singleton initializes correctly when entering the terminal tool interface without a prior GUI instantiation.
• Build Configurations: Excluded zlib1.dll from packaged launcher exclusions.

Full Changelog: v5.4.0...v5.4.1

v5.4.0

[5.4.0] - 2026-02-26

New Features

• Chat Interface: Added right-click context menus and inline action buttons (edit, rerun, more) to individual messages. This allows for modifying sent messages, regenerating responses without creating new user prompts, copying text, deleting, or branching off into a completely new session from any point in the conversation history.
• Session Browser: Added a "Rename" button to the Session Browser, allowing custom titles to be set via a modal dialog.
• Text Selection Strategies: The Text Edit Tool now utilizes low-level Win32 SendInput and WM_COPY parallel algorithms simultaneously to capture highlighted text. This significantly increases compatibility across various Windows applications where standard clipboard polling or basic Ctrl+C injection fails.

Improvements

• Markdown Rendering: Implemented hanging indents for markdown bulleted and numbered lists. Wrapped lines now correctly align with the text content rather than wrapping back under the list marker. Also ensured that markdown code blocks are properly padded with newlines so they don't break when immediately following other text.
• UI & Themes: Applied user and assistant background colors to the chat dialog instead of relying on transparent backgrounds. Improved text selection visibility by ensuring highlighted text always appears above these background colors.
• Window Management: Implemented dark DWM titlebar overrides and used a withdraw-and-deiconify loading pattern for all modal windows to eliminate the bright white titlebar flash that occurs before custom scaling is applied.
• Emoji Rendering: Improved the internal emoji loading pipeline. Built-in Unicode emojis rendered on standard standard Windows GDI menus now use an alpha compositing strategy against native gray/dark-mode theme hues instead of creating white-box transparency artifacts.

Fixes

• File Processor: Fixed an issue where the directory structure was not preserved during recursive repository formatting or parsing scans.
• Audio Splitting: Improved FFmpeg chunk duration estimation to account for target compression bitrates and added an optimization to pass-through the audio stream directly without transcoding when formats match, decreasing chunking times.
• Internal Configurations: Centralized and fixed _IS_COMPILED application state detection checks.

Full Changelog: v5.3.2...v5.4.0

v5.3.2

[5.3.2] - 2026-02-23

Fixes

• File Processor: Fixed an issue where temporary file names were injected into the AI context instead of the original user-facing filenames when processing certain media modes. Also improved context injection for chunked audio (e.g., appended Part 1/3).

Full Changelog: v5.3.1...v5.3.2

v5.3.1

[5.3.1] - 2026-02-22

Fixes

• Launch on Startup: Fixed an issue where the standard Tkinter UI fallback setting was ignored when the app was launched automatically on Windows startup.
• Sounds: Fixed an issue where notification sounds would not play when using the standalone executable package.

Full Changelog: v5.3.0...v5.3.1

v5.3.0

[5.3.0] - 2026-02-21

New Features

• Snip Tool: Added a new "Type" response mode that types the AI's response directly into the active text field.

Improvements

• Settings UX: Clarified the auto_save_session setting logic in the UI and documentation to explicitly state that sessions are always saved upon receiving an AI response or sending a reply, regardless of the selected trigger mode.

Fixes

• Providers: Prevented automatic API key rotation when receiving empty responses from Gemini and OpenAI compatible endpoints. Empty responses are now handled via a standard delay-and-retry mechanism.
• Prompt Editor: Fixed an initialization issue where updating the live preview could cause attribute errors if the playground tab had not yet been loaded.

Full Changelog: v5.2.0...v5.3.0

v5.2.0

[5.2.0] - 2026-02-20

New Features

• Text Comparison: Introduced "Compare mode" for the Text Edit Tool. Now it's possible to select and capture a second piece of text to perform direct comparisons.
  • Added default comparative actions: "Compare Texts", "Find Differences", "Which is Better", and "Before/After".
  • Added a dedicated Compare button (🔀) directly within the popup's "Ask" input field.
• Batch TTS Processor: Added a new terminal tool for batch text-to-speech generation.
  • Supports automatic style generation via AI Director (single aggregated style or unique per-segment styles).
  • Robust progress tracking with checkpoints and the ability to resume or retry failed segments.
  • Includes an option to merge all generated segment WAV files into a single unified output.
• TTS Playground: Added a dedicated "TTS" mode to the Prompt Editor's playground area, enabling real-time testing of models, voices, and AI Director style generation without leaving the editor.
• Prompt Loading from File: The CLI File Processor now supports loading custom prompts directly from text files ([F] Load prompt from file).

Improvements

• Audio Exporting: Optimized FFmpeg conversion logic to use format-specific encoders (libmp3lame, libvorbis, flac, aac) for significantly better compression and quality when exporting in TTS window.
• File Processor: Added an option to inject the filename directly into the AI context for non-text file types (images, audio, documents), improving the model's awareness of the processed file. This preference is now saved and restored within checkpoints.
• Prompt Editor UX:
  • Relocated the "Save Action" button outside of the scrollable area, ensuring it is always visible regardless of the list length.
  • Added the "Compare mode" checkbox for both Text Edit Tool and Snip Tool configurations.
• Settings & Notifications:
  • Made the server settings safety unlock transient, preventing the unlocked state from persisting into config.ini across app restarts.
  • Improved toast notification handling with programmatic dismissal for faster workflows.
• Context Handling: Re-engineered chat requests to properly format multimodal messages, ensuring compare mode and follow-up prompts display flawlessly in the chat window.

Full Changelog: v5.1.1...v5.2.0

v5.1.1

[5.1.1] - 2026-02-18

Improvements

• Thinking Mode: Standardized "Thinking" behavior across all tools (Audio, Snip, TextEdit) to respect the global thinking_enabled setting instead of using hardcoded defaults.

Fixes

• Startup Manager: Reimplemented launcher detection with a more robust multi-path search strategy to fix the "Launch on startup" feature in Nuitka-compiled production builds.
• Chat Persistence: Implemented immediate session saving after user messages and modifications, ensuring chat history is preserved and retry/regeneration is possible even if an API error occurs.
• Regenerate Button: Fixed assistant message handling in the chat window to correctly support response regeneration and session state persistence.

Full Changelog: v5.1.0...v5.1.1

v5.1.0

[5.1.0] - 2026-02-17

New Features

• Launch on startup: Add new option to allow AIPromptBridge to run automatically on Windows boot. Accessible via a new toggle in the Settings window.

Changes

• Hotkeys: Modified default global hotkeys to use Ctrl+Alt instead of Ctrl+Shift to avoid conflicts with common Windows/app shortcuts:
  • Screen Snip: Ctrl+Alt+X
  • Audio Tool: Ctrl+Alt+A
  • TTS: Ctrl+Alt+T
• Settings Layout:
  • Moved "Windows Startup" configuration to the top of the Settings window for better visibility.
  • Relocated "Server Settings" to the bottom and added a safety lock that must be manually unchecked to edit host/port values.

Improvements

• Terminal Interface: Reorganized the console command menu into clearer categories (Features, Info & Toggles) and updated hotkeys for consistency:
  • [S] Sessions (was L)
  • [L] List Sessions (was S status)
  • [I] Info/Status (new)
  • [K] Thinking Toggle (was T)
  • [T] TTS Window (was Y)
• Settings UX: Added explicit warnings when modifying server settings that require a restart.
• Port Discovery: Implemented automatic port discovery and conflict resolution for the local web server. The application will now check 20 subsequent ports if the configured port is occupied.
• Single Instance: Changed the mechanism to prevent multiple instances of AIPromptBridge from running simultaneously.

Full Changelog: v5.0.0...v5.1.0

v5.0.0

[5.0.0] - 2026-02-17

New Features

🔊 Text-to-Speech (TTS)

• Gemini TTS Integration: Full support for Google's Gemini TTS models with streaming audio generation.
• Voice Selection: Access to 30 prebuilt voices with style and gender descriptors.
• AI Director: Intelligent style generation that analyzes text to provide expressive performance instructions (tone, pace, emotion) before synthesis.
• Multi-Format Export: Save generated audio as WAV, MP3, OGG, AAC, or FLAC (requires FFmpeg).
• Multi-Speaker: Support for assigning different voices to up to 2 distinct speakers in a single generation.
• Integration Points: Accessible via terminal [Y], system tray menu, and directly from input popups.
• Tkinter Support: Full functional fallback UI for standard Tkinter environments.

🎤 Audio Tool Enhancements

• File Upload: Added support for uploading existing audio files for analysis, alongside live recording.
• Unified Controls: Consolidated Record/Stop and Play/Pause buttons into dynamic toggle controls for a cleaner UI.

⚙️ Configuration & Playground

• Live Preview: Prompt Editor playground now updates in real-time as settings are modified.
• Playground Audio: Added ability to upload audio files directly in the Prompt Editor playground for testing.
• TTS Settings: Dedicated "TTS" tab in Settings Window and configuration section in Prompt Editor.

Improvements

• Session History: Increased default maximum sessions from 50 to 200.
• UI/UX:
  • Relocated TTS buttons in popups for better layout consistency.
  • Added visual input validation feedback in popups.
  • Added tooltips to TTS controls for better discoverability.
• Performance: Centralized FFmpeg utility detection to reduce redundant system calls.

Fixes

• Threading: Resolved thread-safety issues when accessing UI widgets from background threads in the TTS window.
• Prompts: Improved robustness of transcript detection to prevent text duplication in AI Director outputs.

Full Changelog: v4.3.2...v5.0.0

v4.3.2

[4.3.2] - 2026-02-15

Improvements

• Attachments: Implemented a background cleanup process to automatically remove orphaned attachment directories from deleted or missing sessions.

Fixes

• Console: Bugfixes and adjustments to make console work reliably.

Full Changelog: v4.3.1...v4.3.2