New features

• Added refresh model list for Google Gemini, and option to use v1beta API. If Gemini Pro 1.5 is available for your account, it'll show up
• Added support for Groq
• Added shortcuts when chatting: Stop Generate (Alt + S), Continue Generate (Alt + C), Regenerate (Alt + R)

Self-hosted

• Added Test email account
• Added Override AI Provider setting

Improvements

• Improve streaming message: fix issue when streaming data is incomplete
• Improve Tooltip positioning
• Improve calculate tokens
• Improve streaming message from Gemini models
• Fix can't enable sync data

...to be added

Claude-3 models, profile photo and name, improved layout and colors, and more

New features

• Better Doc Chat app: support conversation history, better finding information and response, improved layout
• Added Profile Settings. Set your name and profile photo to be used in conversation, available in Settings panel > Profile
• Added Claude-3 models (Sonnet, Opus, Haiku)
• Added Model dropdown on Chat window, with option to pin model favorate
• Added Minimal chat style
• Added temprature settings for custom character
• Various improvements for layout and colors
• Added bottom left menu (visible when the left panel is open)

A development version is available at dev.gasbyai.com. All changes will be pushed to this site before going live to our main page.

Self-hosted

• Improved build time (replace babel with esbuild)
• Added backup/restore database

Improvements

• Improved Settings Panel: replace top tab navigation with the left panel navigation, moved Default provider settings to Providers tab
• Improved create/edit custom character panel
• Prevent elements (ie Conversation settings, Stop/Continue/Regenerate) overlay by Chatbox
• Show error when published conversation link is invalid
• Fixed generate chat's title when using Claude models
• Fixed OpenAI fine-tuned models name are missing

... to be added

New features

• Added support for Perplexity models
• Add more Documentation content, improve layout

Improvements

• Show all Changlog content
• Show attached image when using Vision
• Add show/hide attachments in chat messages
• Fix issue when set chat Tone
• Fix Finetune tool not working when not select model from the dropdown
• Fix FFmpeg issue with Transcribe app

New features

• Use GPT Vision for model gpt-4-vision-preview
• Usage analytics: learn about your token usage, and chat sessions
• Added support for Google Gemini models (Gemini is still in beta, be expecting breaking changes)

Improvements

• Improve Settings page
• Replace checkbox with switch in setting pages
• Improve performance (use RctMarkdown to render messages x5 faster)

Major updates for Gasby Business and Enterprise tier codebase. This update will help manage the codebase easier, features available:

• Easier to setup
• Overview of token usage and number chat sessions
• Manage users, default settings including basic settings, styles, providers, etc
• Option to allow/disallow user to update basic settings, styles, provider and migrations
• System settings including System prompt (can be use to set response tone, prohibit topics), mailserver
• User will be using email/password (instead of login link)
• User data are sync across all devices

New features

• Added gasby/open, a free model provided by Gasby
• [App] Text-to-speech using OpenAI models

Improvements

• Improve Chat with Docs app to give better answers
• Added chat server in the US to improve performance for users in Americas (this apply to Gasby free models, and Gasby subscription users)

New features

• [plugin] Added perfect. Enter "perfect" at the beginning of your message to give better output
• Added ability to summarise messages when the content get too long
• Support transcribe long video/audio (automatically split into multiple files)
• Added GPT-style messages (instead of bubble message style). This will be the default, and the setting can be changed in Settings > Style > Bubble Message
• Generate Settings URL to import your settings in a different browser (available in Settings > Migration)
• Gasby subscription users will now have access to 20 models, including GPT, Claude, PaLM 2 Bison, Llama, hermes, Synthia, etc. Big models like GPT-4, Claude-2 will have limited access
• Vision is now available for all users. Free users will have limited access (to use it, attach the image into your message, and send your question)

Improvements

• Talk is now support OpenAI Whisper as the input provider (seems a little slower than if the browser support text-to-speech), custom Instruction, and Language settings
• Translate is now support stream, with UI improved, without content size limit, added Download feature
• Added Custom Prompts, Characters, API Provider settings for exported data
• Added Download message button (when hover on each message in the chat)
• Updated ChatGPT plugins list
• Performance improved: chat message will show the recent messages, scroll down to load more
• Improve websearch prompt to give better answer
• Fix whisper unable to transcribe file
• Improve light mode styles
• Update settings layout
• Token estimation will now only apply for the current conversation (visible when hover your mouse, between model name and current AI Character)

New features

• Create custom apps: Create your own apps to generate response with sequence of prompts and data input (see video below for a demo). Available on https://gasbyai.com/apps/create
• JSONL Generator: generate fine-tuning data file format from docs (pdf, docx, text file) for GPT 3.5 Turbo. Available on https://gasbyai.com/jsonl-generator
• Fine-tune is now using OpenAI new API. Available on https://gasbyai.com/oai-finetune

Improvements

• All models of configured providers is now visible on model command (using /model on the chat input)
• Fix tooltip still visible when navigate to a different page
• Improve response from some custom models/provider (ie delta not found)
• Improve performance for large list (ie list of plugins)
• Display response better (show table and minor styling improvements)
• Docs updated, added how to set Proivder, generate finetuning data, finetuning a model

New features

• Option to edit chat name
• Added Consensus search plugin
• Added support for OpenRouter, Claude APIs
• Added custom headers and body to support APIs that are not compatible with OpenAI API
• Added response tone for the conversation. Available in the Conversation settings panel (setting icon right above the chat input)

Improvements

• Added topics for plugins
• Fix OpenAI Key settings not found
• Fullscreen mode is now available for everyone
• Added Project Management Assistant (contributed by @Mistwalker)
• Improve extract content feature, when URL contain some special characters (ie URL contains &)
• Added Scroll to Bottom setting to automatically scroll to the latest message when chatting, or opening a new chat. Available in the Settings panel (settings icon on the top right)

New features:

• ChatGPT official plugins is in preview (still a work in progress, with limited plugins are usable). You can access plugin list from the star icon (on the left of the chatbox), or enter /plugin on the chatbox to find a plugin. You can try plugins that are not showing tag
• SQL AI (app): Ask AI to generate SQL queries. Available on https://gasbyai.com/sql-ai

Improvements:

• Support multiple URL inputs
• Support multiple file upload (docx/pdf for document, or mp3/mov/avi for transcription)
• Website and URLs will be shown as cards on the message
• Fix glitching response when generating resposne
• Fix pause response doesn't work in some cases

New features

• [app] Bulk Prompter: upload your CSV then generate new column values (https://gasbyai.com/bulk-prompter)
• [app] Talk: similar to Apple's Siri, powered by GPT with Web Voice API for Safari and Chrome (https://gasbyai.com/talk)

Improvements

• Update Google Drive permission
• Add logout to Partner Dashboard
• Copy/delete message is available on mobile
• Minor styles improvement for Chat with Docs app
• Use /clear command to clear messages in the current conversation
• Customize featured AI Characters on your home page (in the library popup, hover an AI Character and select Pin to homepage)
• Added photoURL for your custom AI Characters

There is some changes for Bussiness tier in this version. No need to run run command to install dependencies (just need a server that support NodeJS and should be ready to go)

New features

• Partnership: Join and earn 25% refer GasbyAI to a friend. Available on https://gasbyai.com/partner
• [app] Finetune tool: A tool to finetune and manage files for your OpenAI account. Available on https://gasbyai.com/oai-finetune
• [app] Chat with Docs: Ask question from pdf, docx, pptx. Not capable of sumarization but good at answer questions in the document. Not limit on the file size, doesn't use embeddings, and use very little token. Available on https://gasbyai.com/doc-chat
• GasbyAI subscription: Use Gasby without OpenAI Key. Early bird deals available in $9, $19, and $29 tiers

Improvements

• [app] Image: Add crop, and outpaint ability
• [settings] added better UI for choosing provider
• Speedup loading time on startup
• Improve code viewer for PHP (sometimes being cut off)
• Support reading products from Amazon URL
• Fix input width overflow when the link content is too long
• Add download button for Document viewer in messag

New features

• Image Generator: added edit (similar to inpaint), and variants (generate similar photos)
• Translate: added support to upload doc files (pdf, docx), copy output
• Added share chat (next to delete chat button)
• Added sync with Google Drive (experimental)

Improvements

• Improve background color and some styles for light mode
• Transcribe: added support to upload audio/video file
• Sync: remove merge settings (it'll priority newest data when syncing)
• Some minor visual improvements

New features

• Introduce Apps (features that are separated from the Chat, works with your OpenAI API), available at https://gasbyai.com/apps. Available apps are Generate image, transcribe, translate
• Added docs available on https://gasbyai.com/docs
• Added changelog available on https://gasbyai.com/changelog

Improvements

• Added show/hide doc attachments
• Improve plugins system to help produce more consistent results
• Fix menu not closed on the first click
• Fix message not being sent when chat from the homepage
• Update new models (gpt-3.5-turbo-16k, gpt-3.5-turbo-0613, gpt-3.5-turbo-16k-0613, gpt-4-0613, gpt-4-32k-0613)
• Some small features and styles improvements

New features

• Support message streams (ie. when you edit a message or regenerate response, it create a new stream, you can go back and forth between these streams)
• Use a new local data storage (indexedDB, can store up to x200 more, see Storage Usage on Settings > Migration)
• AI Character images (there are Human and Cute Creatures styles, you can choose it from Settings > Styles > Character style)

Improvements

• Fixed blank page when directly open gasbyai.com/chat/{chatid}
• Added Customer Email Support AI Character
• Cleaning up the prompt list (since some author added an instruction to add a link into the response)
• Improved Regenerate/Pause/Continue generate sometime unstable

A proper documentation is on the way

Note: Since supporting message streams, and started using the new data storage, the app will run a process to migrate the database automatically. You'll see a message to download a backup data file (download and keep this file in case something go wrong)

The changes mentioned above has been updated for self-hosted PRO and Team (please update it from lemonsqueezy). I'm still working on the schema changes for self-hosted Business tier.

New features

• Sync (each active device must enable separately)
• Web search: perform a search when your message start with "search"/"google", then use the content from search results to answer (require Google Programmable Search Engine API setup)
• Extract keywords: when enabled, it'll extract top keywords, relevant keywords, longtail keywords for any answer with 3 sentences or more

Improvements

• Added support multiple placeholders for custom prompts
• Added show prompt (when using a template)
• Added button to generate AI character prompt when creating (based on AI Character title)
• Added button to set a default AI Character
• Added support to import conversations from OpenAI exported data
• Added support background with color instead of image
• Added Chinese output language
• Not sending message when typing Enter + Shift if Chat submission behavior is Enter

New features

• Added manage active devices (Green button at the top right)
• Added Continue, Stop, Generate buttons
• Added custom prompt panel (Prompt library > New)
• Added option to set message background (instead of lime green by default) (Settings > Styles)
• Added support to import typingmind exported data (Settings > Migration)
• Added Code Anything Now AI Character

Improvements

• Folder: fix unstable, add edit button (no longer need long press)
• Message: trip markdown when copy, fix error when code highlighting doesn't support a language
• Fix app sometimes sluggish
• Fix arrow up/down on the message input
• Fix inconsistent when open/close menu
• Some UI improvements here and there
• Util Server (this server is used to extract content from URL, upload pdf, etc) was overload, I've upgrade twice in the past week

New features

• Added support to choose model, output language for a particular conversation (use input suggestion)
• Added support for upload and extract content for docx, xlsx, pptx files
• Add estimate usage (only for chat messages, exclude other services ie dall-e, whisper)
• Add more chat submission behavior (setting icon above the chat box): Choose how you send the message (on Enter, on 2nd Enter or on Shift+Enter)
• More input suggestion:
• /prompt to choose a prompt,
• /role to choose an AI character,
• /chat to navigate to a conversation,
• /model to change model for the current conversation
• /lang to switch output language for the current conversation

Improvements

• For chat suggestion, you can use arrow up/down to change options
• Added tooltip for features
• Move chats in/out of a folder (to rename a folder, press and hold for 2 seconds)

There was a major changes in the organize the code better

New features

• Custom API endpoint
• Showing estimate tokens
• Organize chats with Folders
• Duplicate/edit AI Character
• Show/hide Welcome screen (PRO)
• Bulk select to delete conversations
• Revoke license (settings > scroll to bottom)
• Chat display (will be used by default for Gasby browser extension)
• Use "/prompt", or "/role" in the message box to find access Prompts and AI Characters
• PWA (if you're on mobile, create a shortcut from your browser to access Gasby like a native app)

Improvements

• Fix text color for dropdown on Windows
• Mobile: make buttons bigger, fix some button need to click twice
• Generate image (please choose Image Generator from AI Characters, or start your message with "generate a photo of xxx")

New features

• Added Prompt library
• Added Voice input
• Added handle content from Youtube URL (PRO)
• Added New chat button (Ctrl + N to add new chat)
• Auto generate title (settings)
• Stream responses (settings)

Improvements

• Allow user to choose Output language (settings)
• Disable/enable blur background (styles)
• Hide welcome screen (styles, PRO)
• Improved handle text content from any URL without blocking (PRO, free will still have access to any website that is not blocking)
• Allow response to consider all messages in the conversation
• Edit/delete messages

WIP

• Upload pdf, docx, xlsx files in order to ask question, summarize, etc
• Make transcribe, generate image more flexible
• Remove npm command when installing self-hosted with manage account
• Sync conversations, settings
• Command palette (use / to access prompt, ai role, generate image, transcribe)

The self-hosted for Pro and Teams has been updated (available on your lemonqueeze purchase link), and will updating for Business license

I sent out an update last Monday:

• Self-host with managed account, white label is available in preview
• Updated self-host for PRO and Team (will try to do it every Monday)
• Domain change to gasbyai.com

I'll focus on feature requests and improve in the message presentation this week. If anything need help with setting up, please DM or email me