MindMac - Better privacy-first, native ChatGPT Desktop for macOS. New: GPT-4-Vision and Internet Browsing

At present, users can only upload image files to MindMac in order to utilize the GPT-4-Vision model and ask questions about the image, such as extracting content or writing code. However, I am currently working on expanding the support to include other file types, including csv. Please stay tuned for upcoming updates.

I just discovered this application, this looks great so far. Adding Brave Search api would be nice. One thing I would like is the ability to add function calls. I use function calling for extracting data, not really for actually calling functions. It would be a nice to have. I will be purchasing shortly.

1 Like

Thank you for supporting MindMac. Function calling will be added soon. Additionally, please follow the progress of adding Brave Search API at Add Brave Search API | Voters | MindMac.

1 Like

Hi everyone,

I just released MindMac v1.8.20 to improve overall app performance and support image vision for Ollama LLaVA. Now users can upload images and use local model to ask something about them. New OpenAI models also are available on MindMac from this version.

Please read the full change log below.

:rocket: New

  • Enhanced overall app performance to provide a smoother user experience, including improved scrolling, faster search functionality, and seamless conversation switching.
  • Added new OpenAI models: gpt-4-0125-preview, gpt-4-turbo-preview, and gpt-3.5-turbo-0125.
  • Added a new Perplexity model: mixtral-8x7b-instruct.
  • Introduced a quick action Diagram feature, enabling users to generate diagrams swiftly.
  • Modified the left sidebar to remember the last state it was in when reopening the app, ensuring a seamless user experience.
  • Introduced an add button to the left sidebar, which appears when it is narrow, allowing users to create new conversations or folders conveniently.
  • Enabled image vision support for the Ollama LLaVA model, allowing users to upload images and ask something about them using local model via Ollama.

:sparkles: Improved

  • Disabled the unread notification feature by default to prevent unnecessary notifications.

:wrench: Fixed

Resolved the issue causing crashes when moving files into folders.

Let’s check it out at here and share your thoughts. Thank you all!

1 Like

When will this be on the Mac App Store?

At this time, I do not have any intentions to submit the app to the Mac App Store due to Apple’s requirement to remove the Inline mode feature. However, I may consider submitting a lite version to the Mac App Store at a later date.

Happy Lunar New Year 2024. Wishing you success, good luck and good fortune this year of the Dragon.

To celebrate, we are giving away MindMac Personal lifetime license to 24 people. Just retweet this post and follow @MindMacApp to join. Winners will be announced on February 14, 2024.

1 Like

Hello everyone,

I’m thrilled to announce that MindMac version 1.8.26 has been released with new setting for Network Proxy and several improvements and bug fixes.

Please read the full change log below.

:rocket: New

  • Introduced network proxy setting, allowing configuration of HTTP or SOCKS5 proxy for routing MindMac traffic.
  • Added a new option to enable or disable auto-correction.

:sparkles: Improved

    • Updated the menubar icon for better width conformity.

:wrench: Fixed

  • Addressed a problem where the right sidebar showed blank fields in macOS Ventura 13.
  • Fixed a crash triggered by adjusting the width of the left sidebar in macOS Ventura 13.
  • Corrected an issue where the conversation title was not fully visible in certain situations.
  • Resolved a crash that occurred when an API endpoint was deleted and a new one was added immediately afterward.
  • Solved a problem where GPT4All returned an empty response.
  • Resolved crashing problem when attempting to delete user messages or conversations while receiving streaming response.
  • Fixed crashing problem during the cleanup process of conversations when launch the app.
  • Fixed a problem where the message view would scroll up and down immediately after completing a streaming response.
  • Corrected an issue where the summary line would display the incorrect context mode when switching between conversation and global profile.

Let’s check it out at here and share your feedback. Thank you all!

Hello everyone,

I’m excited to announce the addition of a document upload feature from MindMac v1.9.0, currently in beta. This allows you to upload files and inquire about their content. It seamlessly supports PDF, plaintext, and some code file formats. Preliminary testing with Microsoft Office files (Word, PowerPoint, Excel) has shown promise, and we encourage you to provide feedback on this functionality.

Please read the full change log below.

:rocket: New

  • Introduced document upload feature
  • Introduced new Anthropic Claude models: Claude 3 Opus & Claude 3 Sonnet.
  • Rolled out new Perplexity models, including Sonar Small Chat, Sonar Small Online, Sonar Medium Chat, Sonar Medium Online, and Codellama 70B Instruct.
  • Added new Large model from MistralAI, along with updates to existing models.
  • Added native support for Groq service.

:sparkles: Improved

  • Discontinued support for outdated Anthropic Claude and Perplexity models.

:wrench: Fixed

  • Addressed a keyboard shortcut issue where Command + Shift + Enter inadvertently activated Inline mode, despite Shift + Enter being the designated shortcut.
  • Addressed an issue where certain specific cases were causing blank fields to appear on the right sidebar.
  • Resolved an issue causing poor rendering quality in Internet Browsing results with some OpenAI models, improving the visual experience.
  • Fixed issue with alternating messages when using Perplexity & Claude models.

Let’s check it out at here and share your feedback. Thank you all!

Hi all,

I’m thrilled to announce that MindMac v1.9.5 has been released with many new models: Claude 3 Haiku, Gemini 1.5 Pro, Gemini Ultra and several improvements and bug fixes. In this version, you can also add llama.cpp natively, or add MLX via MLX server or PicoMLXServer.

Please read the full change log below.

:rocket: New

  • Added Claude 3 Haiku model.
  • Added native support for llama.cpp.
  • Introduced Gemini 1.5 Pro and Gemini Ultra models.
  • Implemented support for Anyscale models.
  • Integrated Anthropic Claude 3 models into Vertex AI.
  • Added support for image upload for Claude 3 models.
  • Introduced support for Kagi Search, enabling Internet browsing directly from the platform.
  • Added support for MLX. Users can now easily connect to a MindMac by starting an MLX server or PicoMLXServer.

:sparkles: Improved

  • Improved citation in file upload feature.
  • Improved performance when receiving response in stream mode.
  • Implemented cost display for Claude 3 models in stream mode.
  • The model dropdown list has been updated to hide embedding models for a cleaner interface.
  • Modified the default keyboard shortcut to activate “Bring MindMac to front” from Control + Space to Shift + Control + Space.

:wrench: Fixed

  • Enabled plain HTTP connection for local network.
  • Resolved errors occurring when using Claude 3 models.
  • The platform now remembers the editing prompt when users switch between conversations.

Let’s check it out at here and share your feedback. Thank you all!

1 Like

Am I right to assume that the app does not format mathematics like the browser version of chatGPT?

I’m concerned about the excessive API calls from this app and the use of multiple tokens. Can you clarify where the additional tokens are being utilized?

You’re correct. MindMac now does not support LaTeX but we already have plan to do so, please follow the progress at Support LaTeX rendering | Voters | MindMac.

Thank you for the question. If you enable the option “Auto generate conversation title”, regenerate conversation title or turn on web browsing feature, it may require extra API calls to support these functions.

1 Like

Hello everyone,

Today, MindMac v1.9.7 has been released with new Cohere models: Command R and Command R+. Especially, Command R+ is impressively fast with 104B parameters, 128K context length, and amazing prices ($3/1M input tokens and $15/1M output tokens). It’s definitely a game-changer.

In this version, you can also add Brave Search as search provider for Internet Browsing feature. Please read the full change log below.

:rocket: New

  • Introduced Command R & Command R+ models for Cohere.
  • Added support for Brave Search for Internet Browsing.
  • Integrated support for KeyMateAI Google Browse Assistant API.

:wrench: Fixed

  • Removed unnecessary quick text shortcuts for accessing prompt template list and occupation list.

Let’s check it out at here and share your feedback. Thank you all!

Hello everyone,

I’m excited to announce that MindMac v1.9.8 has been released today with new OpenAI GPT-4 Turbo model as well as a better search feature. Now you can search across all conversations easily.

In this version, we also integrated the Gemini 1.5 Pro model for Vertex AI users.

Please read the full change log below.

:rocket: New

  • Introduced GPT-4-Turbo model from OpenAI.
  • Integrated the Gemini 1.5 Pro model for Vertex AI users.

:sparkles: Improved

  • Enhanced search functionality now allows for content search across all conversations.
  • Updated the MindMac icon in the message view to reflect the icon selected in app settings.

Let’s check it out at here and share your feedback. Thank you all!

I have attempted to upload some relatively large PDF files, ranging from 3 to 5 MB, and MindMac consistently crashes. Is there a maximum file size restriction or a practical upload size limit?

Presently, file uploads are limited to 20MB. Would you kindly send the crash log to support[at]mindmac.app so that I can conduct a further investigation into the issue? Thank you!

Hello all,

I’m excited to announce that MindMac v1.9.11 has been released today with document upload capabilities improvements and other fixes. Please read the full change log below.

:rocket: New

  • Added custom model field for Ollama when creating new endpoint.

:sparkles: Improved

  • Updated model name for Cohere Command R+.
  • Included service logos next to each answer from AI.
  • Enhanced document upload capabilities, specifically for Excel and CSV file formats.
  • More detailed error messages now shown for Vertex Claude 3 models.

:wrench: Fixed

  • Resolved a bug causing crashes during document uploads.
  • Addressed a crash that occurred during the cleanup of conversations for certain users.

Let’s check it out at here and share your feedback. Thank you all!

Hello all,

Llama 3 8B and 7B models now are available on both Groq and Perplexity in MindMac v1.9.13. This new version also includes improved scrolling, better handling of long messages, and simplified multi-paragraph selection for copying. Please read the full change log below.

:rocket: New

  • Added support for Llama 3 models on Groq and Perplexity.
  • Introduced support for LaTeX formatting.

:sparkles: Improved

  • Optimized overall application performance, including improved scrolling, better handling of long messages, and simplified multi-paragraph selection for copying.
  • Disabled network proxy when using local inference sources such as Ollama, LMStudio, MLX.
  • Transitioned explanatory messages to popover form in specific settings fields.
  • Enhanced user interface for the main message view.

:wrench: Fixed

  • Resolved problems preventing some users from utilizing Claude 3 models on Vertex AI.

Let’s check it out at here and share your feedback. Thank you all!

1 Like