Changelog

・Local LLM models are now available both with and without BYOK activated.
・Resolved the frustrating bug that caused Triplo AI to forget the last selected model.

・Triplo AI Mobile can now connect to local LLMs operating on Desktop Clients (from 5.0.0 on).
・Enhanced UI/UX with a fresh design.
・Improved speed thanks to a new AI loader.
・Fixed the bug related to sharing with Triplo AI.

・Users can now leverage their Mobile Triplo AI to securely connect to local LLMs running on desktops via encrypted communication, starting with Mobile Clients v3.4.0.
・Local LLM models now support Ollama and LM Studio.
・Admin devices can rename other devices linked to the same key.
・Minor adjustments to the user interface.

・Incorporated OpenAI o1 and o3 models under BYOK.
・Ollama model endpoints are now compatible with OpenAI [BETA].
・Introduced a Spider toggle in the UI footer for inline scraping.
・Added hotkeys: Control+Shift+S to toggle inline scraping and Control+Shift+C to switch to chat mode.
・Enabled reasoning display for OpenRouter models that support it.
・Enhanced accuracy in token usage measurement, now sourced directly from the service provider.
・Improved display and error handling for slower models, including Deepseek R1 and other cognitive models.
・Resolved issues with the listing of Ollama models.
・Fixed Mac compatibility issues related to trigger and hotkey configurations.

・Token counting error has been fixed.
・Markdown rendering is now supported in both single prompt and chat modes.
・Minor bug fixes.

・A correction has been implemented to resolve an issue that was erroneously causing token exhaustion.

・Triplo AI can now access and process Local Models via Ollama!

・Major performance improvement
・Added switches for BYOK
・Bug fixes and general improvements

・Localized the interface to German (DE).
・Enhanced the device association reset process.
・Introduced tooltips for features in the Settings/Switches menu.
・Added direct links to Help and Documentation for key features.
・Implemented a confirmation prompt for deleting history/logs.
・Included token estimation for attachments during prompts.
・Masked API keys and license details to enhance privacy.
・Added a reset link in the Activation/Device Management screens.
・Included model names in history/logs.
・Introduced a "Shift+Backspace" toggle switch in settings.
・Added voice speed control in the Voice settings.
・Expanded the Voice feature (TTS) with new voice options.
・Implemented a "Re-Prompt" option in the main input interface.
・Resolved minor UI issues.
・Made security enhancements.

・Fixed Anthropic Integration
・Fixed Long Results Bug
・Security Improvements

・Models list can now be updated without requiring a new release
・Resolved an issue with the Minds on Mobile

・ Three hotkeys are back!!!
・ Increased limits for Hotkeys: Early adopters can configure up to 50 per device, Soft Plan users can set 5 per device, Medium Plan users can set 15 per device, Pro account holders can set 50 per device, and Free Plan users are limited to 1 per device.
・ Significant speed enhancement for both Mac and Windows.

・ You can now incorporate images [BETA] into your prompts by pasting them from the clipboard or attaching them directly.
・ Introducing Triplo AI Hotkeys, which can process pre-defined or inline SmartPrompts on highlighted text and automatically replace the original content.
・ Users now have the option to allow other devices to "Edit" their shared SmartPrompts, Automations, Instructions and Minds.
・ Minds can now be trained using multiple URLs simultaneously.
・ SmartPrompts can be organized and sorted by tags, complete with search and filtering options.
・ Various minor UI improvements have been implemented.
・ A bug that prevented file attachments when the UI wasn't pinned has been fixed.
・ The models list can now be updated without requiring a new release.
・ Triplo AI now allows you access to 60 models. Details here.
🔥🔥🔥

・ Added GPT-4o-mini, Jamba 1.5 Mini 256k, DBRX 132B Instruct 32k, Microsoft WizardLM-2 8x22B 65k, Mistral Nemo 128k, NVIDIA Nemotron 70B 131k, and Qwen2-VL 7B Instruct 32k under Allowance.
・ Users can now use their Anthropic key to access premium models, including Claude Haiku 3.5
・ Minor Bug Fixes

・ Introduced the FILES feature on the Training page and Prompts.
・ Updated model lineup, now including: AI21 Jamba 1.5 Mini 256k, DBRX 132B Instruct 32k, Microsoft WizardLM-2 8x22B 65k, Mistral Nemo 128k, NVIDIA Nemotron 70B 131k, and Qwen2-VL 7B Instruct 32k.
・ Removed outdated models: Mistral Open Orca 7B 8k, PPLX Online 7B 4k, and Yi 6B 4k.
・ Enhanced UI performance on the Training page.
・ Implemented minor UI/UX improvements.

・ Allowance tokens are now standardized at 2,000,000 tokens per device per month across all 15+ available models, including the GPT-4o Mini.
・ Resolved several bugs introduced in the latest Windows version, such as the black UI, single-word prompt issues, and JavaScript errors.
・ Fixed a problem that prevented users from copying results from the UI on Windows 11.
・ A new hotkey has been added to restore the UI display; simply press CTRL+SHIFT+P while Triplo AI is in focus.

・ Triplo AI is now equipped for multiple languages, featuring localized user interfaces in Arabic and Brazilian Portuguese.
・ We have enhanced the multilingual experience by adding a language selector in the main UI, allowing users to prompt models to respond in their preferred language.
・ Resolved an issue in the automation creation form where the 'suggest payload' button was non-functional for non-string parameters.
・ Security updates have also been improved for better protection.

・ Included GPT-4o Mini under the allowance of 1 million tokens per month per device and API.
・ Added an Anthropic API key so users can access Claude models directly.
・ Added Claude 3.5 (OpenRouter/Anthropic), LLaMA 3.1 (8B & 70B under allowance/API), and 405B (API).
・ Cleaned up deprecated OpenRouter models under allowance/API (Nous Hermes Llama2 70B, Zephyr 7B, Synthia 70B).
・ Added inline scraping and API key switches under settings.
・ Significantly improved speed on Macs.
・ Improved UX: On errors, pressing "ESC" returns you to prompting or halts the current generation; "SHIFT+BACKSPACE/DEL" clears the previous results.
・ Fixed issues with agents and the unexpected "vanishing" UI in Windows.
・ Fixed automation (webhooks) under trial.

・ Fixed UI clearing after the response when using Openrouter models.
・ Fixed the settings persistance for Chat Mode, Active Mind, and Active Custom Instruction.
・ Added the following models to OpenRouter (API): Gemini Pro 1.5 and Gemini Flash 1.5

・ Included OpenAI's new GPT-4o model.
・ Implemented security updates on the backend.
・ Minor bug fixes.

・ Fixed Settings "Save State".
・ Fixed Google Gemini cutting off results.
・ Adjusted Google models tokenization (different from other models).
・ Google Gemini Pro under allowance is now paired with Open AI's GPT 3.5 not the other OpenRouter models.
・ Fixed Chat history markdown crash for some markdown tables.
・ Added Llama3 8B to the Allowance (1/4 tokens cost model).
・ Added Llama3 70b to the API Key usage.

・ Expanded allowance for plans (500,000 GPT 3.5 and 2 million Mixtral/OpenRouter tokens per device per month)
・ Added multiple OpenRouter models to the allowance. Now you can use the following models: Dolphin Mixtral 8x7B 32k, Google Gemini Pro 131k, Meta Llama-2 13B 4k, Mistral Open Orca 7B 8k, Mistral Tiny 32k, Mistral-7B 8k, MythoMax 13B 4k, MythoMax 13B 8k, MythoMist 7B 32k, Nous Llama-2 13B 4k, Open Hermes 2.5 7B 4k, OpenChat 3.5 7B 8k, PPLX Online 7B 4k, Toppy M 7B 32k, Yi 6B 4k, Zephyr-7B 4k
・ Introduction of new models: Claude 3 200k Haiku/Sonnet/Opus, Mixtral 8x22B 65k, and the new GPT 4 Turbo
・ Introduction of a new lifetime FREE plan for Triplo AI
・ Clearer pop-ups and error messages on disabled/limited features
・ Bug fixes and minor UI improvements

・ Triplo AI UI can now be resized
・ The Status Dot (visual notification of processing status) is back!!!
・ You can now use TAB to select your SmartPrompts, Awareness commands etc
・ Added new models from OpenRouter: Airoboros L 2 70B, Code Llama 34B Instruct, Goliath 120B, Mistral Open Orca 7B, Mythalion 13B, MythoMist, Neural Chat 7B, Noromaid 20B, Nous Hermes 13B, Open Hermes 2.5, Phind Code Llama 34B, PPLX 7B, Psyfighter v2 13B, Synthia 70B, Yi 34B, Yi 6B [DONE]
・ GPTs 3 and 4 models updated to their latest versions (0125)
・ Minor UI improvements: fixed double Enter bug on Voice mode, added an alert for Automations under OpenRouter, corrected bug related to RTL languages.

• Added multiple new models under OpenRouter: Airoboros L 2 70B, Code Llama 34B Instruct, Goliath 120B, Mistral Open Orca 7B, Mythalion 13B, MythoMist, Neural Chat 7B, Noromaid 20B, Nous Hermes 13B, Open Hermes 2.5, Phind Code Llama 34B, PPLX 7B, Psyfighter v2 13B, Synthia 70B, Yi 34B, Yi 6B.
• Multiple small fixes for better UX using Voice Mode on Triplo AI Mobile
• Fixed bug that unexpectedly triggers inline SmartPrompts.

• Release of Voice Features [Beta]
• Changed Mixtral model under Allowance to Dolphin version
• New Models added under Open Router keys
• Added Multilingual Youtube Scraping features to Mobile
• Refined UI (better icon placement, fonts and other improvements)
• Improved allowance models selection

• Corrects bugs introduced on the latest version 2.5.5) related to the initial use of Mixtral/GPT allowance Models.

• Added voice features (Text to speech and speech to text) with keyboard shortcuts
• Replaced mixtral allowance from Mixtral 8x7B Instruct to Dolphin Mixtral 8x7B 32k
• Added Scrapemark SmartPrompt and inline markdown scrape with #url (Ex. #https://triplo.ai)
• Added the following models to OpenRouter: Perplexity 70B 4k, Dolphin Mixtral 8x7B 32k, OpenChat 3.5 7B 8k, MythoMax 13B 4k, MythoMax 13B 8k, Mistral Tiny 32k, Mistral Medium 32k, ReMM SLERP 13B 6k, Izlv 70B 4k.
• Fixed a wrong error message in inline scrape where it would show invalid URL instead of 'couldn't scrape this site' error
• Tweaking feature is now out of BETA

• Minds can now be used on Mobile
• Added Mixtral model on the Allowance Tokens
• Minor bug/UI fixes

• Minds can now be shared between devices (Desktop and Mobile)
• Added scrapehtml normal prompt and inline with @url (RAW HTML Scraping)
• Changed sync approach (a device can now see synced data from others even if it's sync is off)
• Added auto translate to YouTube scrape (AMAZING!)
• Dark/light mode control under settings
• UX Improvement - "esc" on secondary screens returns to the main UI
• Fixed app crash in history page when some chats have an error in inline scraping
• Fixed an issue on Windows and macOS where awareness did not work with Ctrl+A in some cases

・Released multi-URL scraping (you can now scrape multiple pages at once inline)
・Instructions can now be significantly longer (5000 chars)
・Improved Allowance Tokens process
・Added a debug info extractor under settings for better support

・Allowance has been updated to include Model Mixtral 8x7b 32k, allowing for up to 1.5 Million tokens per month per device

・Major performance fixes and improvements
・Added Mixtral 4x7B Instruct to OpenRouter models
・Improved error handling to inline scrape (Triplo AI warns of scraping problems even on inline prompting)
・Added tokens of each History entry
・Added PUT on Automation settings
・Bug fixes on Automation (confirmation sometimes caused the app to crash) and UI bug in History page if some titles were too long

・Implementing OpenRouter
・Improve device management functionality
・Improve chat title functionality
・Security improvements
・Bug fixes

・Expanded Training Minds to 200k tokens
・Reduced training tokens costs to 1/5th the chat/prompting costs
・Training Feature is not under Beta anymore
・Added ability to disable confirmation of Automations under settings
・Automation Feature is not under Beta anymore
・Added ability to rename chats, AI rename chats, and automatically naming chats on creation
・Added copy button in chat history
・Fixed a but were some new chats were not shown in history due to a refresh issue
・Improved license validation and device management UI/UX
・Improved error messages and error handling
・Fixed a bug of stop generating content
・Added 'Auto Select' for GPT3.6 16k and GPT4 128k under OpenAI/OpenRouter APIs
・Added Tobby M 7B 32k under OpenRouter Models
・Faster OpenAI and OpenRouter keys validation
・Improved Allowance Keys security

・Now mobile App can act as Admin Device
・Create/Edit/Sync Custom SmartPrompts
・Improvements on Instructions Syncing

・Allowance tokens now are GPT3.5 16k
・Added OpenRouter models: Hugging Face Zephyr 7B (beta), Hermes Llama2 13B (beta) and Hermes Llama2 70B (beta)
・Added Admin approval for new devices under Devices Management
・Added button to delete all chats
・Added confirmation screen on delete (chat history, custom prompts, instructions, automations, and training)
・Added “~” as a hotkey option
・Security improvements

・Code refactoring/cleaning. Preparation for future developments.

・Security fix

・Speed improvements
・Fixes for multiple keyboards/languages
・Security improvements and bug fixes

・Bug fixes

・Hotkeys Fix/Revision

・Server Upgrade
・Hotkey scheme change (2 keys)

・Server Upgrade

・Server Upgrade

・New Website/YouTube scraping (Scrape/Inline)
・Improved Awareness Support (apps with no available "share" function)
・Added Custom SmartPrompts sync
・Fixed Chat Mode on iOS
・Fixed iOS notifications issue
・Improved loading speed on the first login
・Changed the history titling prompt
・Added Welcome screen

・New Device Management dashboard (with sync settings)
・Fixed an issue with Youtube scraping
・Improved control of Automation (confirmation routine)
・Automatic text alignment for RTL/LTR
・Added Mistral 7B Instruct model to OpenRouter
・Better UX for Training setting
・Stopped showing window on startup
・Fixed autoscroll bug (it didn’t scroll all the way to the buttons)
・Disabled automations on OpenRouter (not supported)
・Added more choices in hotkey options
・Performance improvements
・General bug fixing

・Bug fix iOS old iPhone models

・Custom SmartPrompts syncing
・Bug fixes

・New Awareness Mode for Mobiles
・Fixed Chat history order
・Improved Chat persistency
・Many UX improvements (less clicks/touches)
・Bug fixes and Code improvements

・Added Instructions feature
・Much better/polished UI
・Several bug fixes

・Integration of OpenRouter API [BETA]
・Added Always On mode
・Fixed License Revoking Bug
・Fixed Windows Scaling/Multi-Monitor Bug

・Unified version Android/iPhone
・New polished User Interface
・Added new Chat mode support
・Added new History/Logging capabilities
・Added the GPT 3.5 4k tokens for licensed users
・Added model selection
・Bugs fixes and Improved stability

・New Tweaking [BETA] feature (Creativity/Diversity settings)
・Add option to set Youtube language for scraping under Settings
・Improved Automation [BETA] feature (better webhooks delivery)
・Triplo AI is now aware of the URLs addresses scraped by it
・History page pagination fix, and UI improvements
・Added an Active Devices counter under Settings

・Fixed a bug where listen to clipboard sometimes does not work on startup.
・Added ability to disable auto scroll in settings
・Chat Mode state is saved on app restart

・Fixed Licencensing Validation Bug

・Besides SmartPrompts, now Instructions and Automations are now synced across devices
・Instructions, Minds, and Automations now have sellectors directly from Triplo Bar
・You can now add multiple lines in the input with Shift+Enter
・Max number of Minds per device are now 10
・Max Tokens per Mind set to 32k (performance/costs testing under Beta Mode)
・Fixed a bug where some stored settings sometimes were missed on App launch
・Fixed a bug on "OOO" mode was recording tokens incorrectly
・Other minor UI improvements and bug fixes

・Noticebly faster and more responsive User Interface
・Increased limits for Training. "Minds" can now have 80,000 tokens
・Enhanced Training results (WOW! What an improvement!)
・Improved Automations UX. Automatic suggestion for JSON payloads
・Improved scraping (added more page elements)
・Added the ability to select and copy Triplo AI and OpenAI keys
・Fixed "Stop Generation" bug (persistent messages)
・Several minor bug fixes
・Security improvements

・Important bug fixes and security improvements

・Fixed licensing bug

・AWESOME Training Feature [Beta]
・AWESOME Automation Feature [Beta]
・Instructions
・Minor bug fixes

・Chat History
・Custom Instructions
・Free 300k/month GPT 3.5 4k tokens (for LTDs)
・Tokens Counter
・Fixed LinkedIn links for Scraper
・Security improvements and bug fixes

・Improved scraping capabilities
・Youtube links scraping and awareness
・Improved chat/stand-alone UI
・Send results to cliboard switch
・Selectable results
・Markdown support (code, table, text)

・New amazing UI (with progressive generation)
・Faster insights/content delivery
・AI Chat anywhere on your OS
・Easier model selection (Including the 16k new model)
・Refined UX for better experience
・Direct URL content capture in-prompt
・Scraping
・Custom SmartPrompts

・Ability to change model
・Ability to to share text from any app to Triplo

・Bug fixing (Windows Built)

・Bug fixing (UI issues, slow behavior on Windows, saving model selection)

・Clipboard access
・Awareness feature (Selection and Context)
・Bug fixes

・Bug fixes

・Minor UI adjustments
・Bug fixes

・Refined UI for a smoother experience.
・Introduced a sleek Dark Mode feature.
・Enabled users to reposition the Spotlight Bar on the screen.
・Displayed search results directly on the Spotlight Bar.
・Added customizable keyboard shortcuts to the Spotlight Bar.
・Provided the ability to select AI Models.
・Enabled users to control auto start of Triplo AI on boot.

・Initial Release Android/Chromebook Apps

・Correction of the 1.1.6 auto-update process on Mac
・Change of the versioning rules

・Release of the new UI (Spotlight)
・Change of the "status dot" behavior on Mac (click to Spotlight)
・Major bug fixes (keys management)

・Major bug fixes (key validation errors and "ooo " trigger fixes included)
・Security improvements
・Code refactoring
・Trial period implementation
・Improved SmartPrompts search

・Fixed issue where the indicator takes the focus on each request
・Adjusted minimize icon on Linux and Windows
・Added Disclaimer Agreement on 1st run
・Adjusted "How to use" link on UI
・Fixed issue where the license gets deleted on startup & when the internet disconnects

・Added custom triggers (others than "ooo")
・Added SmartPrompts Search (ooo smart + word)
・Added auto-update
・Added auto run at startup (macOS & Windows)
・Ability to revoke a device/license (switch devices)
・DMG package support for MacOS
・AppImage package for Linux
・General bug fixes

・ UX improvements (ability to disable the dot status when idle, better phrasing etc)
・ Added the ability for users to use OpenAI tokens
・ Added larger limits associated with OpenAI tokens use (multi-language capabilities, larger consumption of tokens/call)
・ Indicator of new versions on the UI
・ Better logging process (Server/Client sides)
・ Wide code refactoring (faster/cleaner)
・ 1st run of Harpy (little teaser for you)

・ Improved movable status notifications
・ Mac Silicon Pro/Max correction
・ Added onboarding process
・ Revision of Keys/Tokens association
・ Excluded Windows Portable Version

・ Development tests

・ Signed app for Windows
・ Multi-plans

・ Development tests

・ Stability improvements
・ UI improvements
・ Release of SmartPrompts documentation
・ Beta Version

・ Reinforced privacy standards
・ Multi-language management
・ Custom notifications
・ Signed App for Mac

・ Fix Connection stability
・ Licensing system implementation
・ Improvements on SmartPrompts dynamics

・ Inception post MVP