v1.18.0 (build 94)

Exciting Update: Temporary Chat, Memory Plugin, FLUX text-to-image model, Gemini 1.5 Experimental 0801 and more.

• New: Temporary Chat. Quickly start a new Temporary Chat with a keyboard shortcut. Go to Settings > Shortcuts to set yours.
• New: Memory Plugin. Persist information across conversations.
• New: Pull Google Gemini models from the server. Added support for the new model gemini-1.5-pro-exp-0801
• New: Added support for the new model FLUX from Black Forest Labs (via Replicate)
• New: Added support for Dall E hosted on Azure AI Service
• New: Add new plugin Whisper via Azure
• New: Customize more keyboard shortcuts. Go to Settings > Shortcut to start.
• New: Customize icon for Custom AI Servers
• New: Chat Developer Layout. Renders markdown for user messages.
• New: New markdown editor when editing messages
• New: Drag documents & images to the Dock icon or Status Menu Icon to start a new chat
• Fixed: Other bug fixes and improvements

Screenshots & demo videos

1. Memory Plugin:

Demo Video:

Once enabled, BoltAI will automatically remember details between chats. You can also explicitly tell BoltAI to remember certain information. For example:

"Remember I'm a solopreneur" "My name is Daniel Nguyen. Use it whenever you answer to an email"

In BoltAI, memory feature is available as a plugin, and requires the Function Calling to work.

Memory is disabled in Temporary Chat.

You can manage your memories in Settings > Plugins > Memory. Select tab Memories

2. Temporary Chat:

Demo Video:

3. Customize keyboard shortcuts

Go to Settings > Shortcut to start.

4. New markdown editor

5. New developer layout

v1.17.0 (build 91)

Big Update: Document Analysis, Ollama function calling, Anthropic tool use streaming and more.

• New: Document Analysis is here. Attach documents to your chat and ask AI about it. Learn more...
• New: You can use plugins with Ollama now (Ollama v0.3+ required)
• New: Keyboard shortcut for "New Chat Window"
• New: Add plugin "JavaScript Interpreter" using JavaScriptCore
• Improvement: More accurate tokens counting & cost estumation
• Improvement: Anthropic Tool Use now supports streaming. Faster and better!
• Improvement: Customize Base URL for Anthropic
• Experiment: Customize the strategy to fetch highlighted text for AI Inline
• Bug fixes & improvements

v1.16.8 "Power Users"

Big update for power users! (hello @ethanj801 👋)

• New: Gemini 1.5 Pro 2M context window
• New: You can use AI plugins with Gemini models now
• New: Added support for Code Execution for Gemini models
• New: Customize Safety Settings for Gemini models
• New: Search in plugin list dropdown
• New: Supports more advanced parameters: Top P, Top K, Max Tokens…
• New: AI Command supports alternative prompt & more advanced parameters…
• New: More powerful Context Limits management: All Messages / No Context / First n Messages / Last n Messages…
• New: Detach Chat Configuration panel to a new window
• New: Save chat configurations as default
• New: Sort chats by Edit Date, Create Date or Title
• New: Pin chats / Hide chats
• New: Customize chat message action toolbar
• New: Delete all subsequence messages action
• Improvement: Allow disabling the current time in the System Instruction
• Improvement: Hide windows on login
• Attempt to fix the issue where chat messages are gone after pressing Enter
• General bug fixes & improvements

v1.16.7 (build 85)

• New: Added Anthropic's latest model Claude 3.5 Sonnet
• New: WolframAlpha LLM plugin
• New: Whisper plugin (run on Groq)
• Improvement: Right-click and open images with BoltAI
• Improvement: Supports finetuned models for AI Command & AI Assistant
• Improvement: Start a new chat in the right folder
• Fix: Icon Reverts on App Update

v1.16.5 (build 84)

Exciting release. This release focuses on Image Generation & Read Aloud features.

• New: A brand new Audio Player for the Read Aloud feature with the option to change playback speed
• New: More customizations for Dall E: quality, size, style and number of images
• New: Better supports for Replicate: new models, customize aspect ratio and number of images
• New: Use Dall E and Replicate as an AI plugin
• New: View revised prompt for each generated image
• New: New Whisper plugin, transcribe audio files using OpenAI's whisper model (requires OpenAI API key)
• New: Quickly switch to a different AI Service/Model
• New: Quickly change chat configuration with the Status Bar
• New: Use a custom app icon (only PNG is supported)
• Fixed: Attempted to fix the issue where chat message gone after pressing Enter
• Fixed: Stop chat stream when clicking the stop button, or when deleting current chat
• Fixed: Fixed issue with perplexity & context limit setting
• Fixed: Cannot scroll the chat input field
• Fixed: Only fetch OpenRouter model list if OpenRouter service is enabled
• Fixed: Restart not working
• Fixed: Launch at Login not working
• Fixed: AI Command Alternative Profile doesn't follow the behavior setting
• Fixed: AI Plugins settings are not saved
• Fixed: App Icon is not changed after quitting the app

v1.16.0: 20+ new features!

Hi everyone. I'm super super excited to share what I've been working on the last few weeks.

After a mini burnout, I was able to get back to building and added a lot of new and useful features. I can't wait to talk about it.

TL;DR: Lots of new features!

• Inline Whisper: use Whisper in any text field
• 7 new plugins: Shell Access, FFmpeg, AppleScript, Kagi Search, Brave Search, Perplexity Search, You.com Search
• Custom App Icon
• Support Function Calling for Groq, MistralAI & other services
• Support local vision model (llava via Ollama)
• Customize AI Service & model for AI Assistant
• “Default” as an option for Command & Assistant
• Better UX for AI Commands & Assistants
• Instant AI Command: press a shortcut key then ask AI about anything
• Alternative Profile for AI Command
• AI Command supports custom System Instruction
• Import from ChatGPT
• Import/Export BoltAI data
• BoltAI Database Maintenance
• Add “official” support for LM Studio
• Regenerate using different AI Service/Model
• Dark theme for code blocks
• Remember sidebar & state
• Hide Models
• Highlight Search Keyword
• Launch at Login
• Huge performance improvement!!

--

There are so many new changes, I don't know where to start. Let's start with the most voted feature requests: Inline Whisper

Inline Whisper

The Inline Whisper feature allows you to activate OpenAI's whisper transcription within any textfield on your Mac. Learn more...

Here is a simple demo video:

7 New AI Plugins

AI Plugin has been my primary focus for a while now. In this release, I've added 7 more plugins:

• New Web Search Plugins: Kagi Search, Brave Search, Perplexity Search and You.com Search
• Shell Access Plugin: allows AI to use bash command. WARNING: Always double check the command.
• FFmpeg Plugin: allows you to edit audio & video using ffmpeg command.
• AppleScript Plugin: allows BoltAI to control your computer using AppleScript

In the next version, I will add support for custom plugin development. It's more challenging than I thought.

Follow this guide to learn more on using AI Plugin in a chat

Added "Function Calling" support for more AI models

The AI Plugin feature requires the AI model to support Function Calling capability.

Previously, this feature is limited to OpenAI's models only.

In this version, I've extended this capability to more AI services. Here is the list of services you can use AI plugins:

• OpenAI & Setapp
• Azure OpenAI Service
• Anthropic Claude 3
• Groq (recommended to use with llama3-70b only link)
• Mistral AI (limited to a few models)
• Together AI (limited to a few models)
• OpenRouter (only OpenAI's models are supported)

Features for Pro Users:

BoltAI's pro users love the AI Command & AI Inline features. These features integrate deeply with the OS and provide better user experience, but it also the most challenging to implement.

In this release, I've improved some of the feature to make your life easier.

Local Vision Model BoltAI now supports local vision model (only llava via Ollama for now).

Added official support for LM Studio

AI Inline supports custom AI Service & Model

This feature is long overdue. In the early days, BoltAI support OpenAI models only. And it becomes a technical debt.

I'm happy to share that I've added support for custom AI Service & Model, including a new option "Default", where it always follow the default service you set in Settings > Models.

Default AI Assistant

Set an assistant as default and new chats will always use this assistant.

Go to Settings > Assistants. Right-click on an AI Assistant and choose Set as Default.

AI Command's Alternative Profile

AI Command's Alternative Profile allows you to run the same command with a different configuration profile. Learn more...

Customize Instant Command

In BoltAI, an AI Command consists of pre-defined configuration, system instruction and behavior. An Instant Command* allows you to type your prompt and start chatting right away.

Learn more...

More customization

UI & UX has always been my top priority for BoltAI. In this version, I've added App Icon customization & better code block theme support.

Customize App Icon:

Customize Code Block theme & fonts:

Data import & export

Another most voted feature is to support iCloud sync. Unfortunately, it requires too much effort and I wasn't able to make much progress on this feature yet.

In the meantime, you can transfer your data from one device to another with this data import & export feature.

You can learn more about the feature here: https://boltai.com/docs/chat-ui/import-export

In this release, I also added the ability to import your ChatGPT data. Enjoy a seamless transition from ChatGPT to BoltAI! Read the step-by-step guide: https://boltai.com/docs/chat-ui/import-from-chatgpt

Improved user experience

I'm quite happy with the overall UI & UX of BoltAI. But there are certain areas where the experience is not quite good yet.

In this version, I'm trying to address some of them.

Easier to switch AI service & model mid-conversation

Quickly regenerate with another AI service/model configuration.

Hide unused AI models

Some AI services like OpenRouter & Together AI support a lot of models. It can be overwhelming to find the right model.

In this version, you can choose which model to show when starting a new chat, or when switching to another model.

Highlight search keyword

It's a minor feature but I find it super useful.

Easier to manage AI Commands/Assistants/Prompts

You can already do bulk update AI Commands & Assistants but it's a bit hard to discover. In this verison, I've added an "action button" to improve the discoverability.

Launch at Login

Another useful quality-of-life feature ✌️

Huge performance improvement

I spent a few days optimize the overall performance of BoltAI and I can say it's a lot faster now.

Closing Thoughts

After more than a year working on BoltAI, I started to understand what it is, and who it's for.

Recently OpenAI launched the official ChatGPT app for macOS. And just yesterday, Apple announced Apple Intelligence with system-wide AI integration.

I must admit, it did stress me quite a bit.

But on second thought, I believe this is actually a positive news. More people will be able to leverage AI in ther day-to-day job. And when they understand the limitation of an AI system, they will naturally find a better alternative that solve their niche problems better.

That's where BoltAI can fill the gaps I think.

And lastly, thank you all for your support messages. BoltAI is nothing without its fans 🫶

v1.15.25 (build 69 "nice")

• New model Gemini 1.5 Flash
• New app icon ⚡
• Temporarily disable the "Remember Chat Configuration" since it confuses many users
• Fixed an issue where the app requires a restart to continue

v1.15.18 (build 61)

• Fix a crash for some users when selecting a particular chat

v1.15.17 (build 60)

• Improvement: Auto switch the focus back to previous app when using activation hotkey
• Fix: fixed the crash when using Anthropic Claude 3 with plugins
• Fix: fixed the regression bug where font option does not apply to AI messages

v1.15.16 (build 59)

• New: Customize font size for the chat input
• Fixed an issue with license activation
• Fixed: Command N do not open a quick chat
• Fixed: BoltAI can't force stop, endpoint keeps streaming
• Fixed: New line inserted at the bottom rather than at the cursor position
• Fixed: Unable to select assistant in new chat
• Fixed: Creating a new command shows deleted models
• Fixed: Bulk edit provider in Command showing deleted API Provider
• Fixed: Chat not showing all chat content after search
• Improvement: Remove quotes from chat title