RecurseChat 17+

Personal AI Chat

Xiaoyi Chen

    • 4.6 • 17 Ratings
    • $9.99

Screenshots

Description

New: Bookmarks and folders, Llama 3.2, Gemma 2, Claude Sonnet, Chat with multiple PDF files locally and offline

RecurseChat is a personal AI app for chatting with local AI offline and privately. Chat with local documents, chat with images, import ChatGPT history, perform advanced text search, and much more.

Features:

• Local first, Offline capable
Talk to AI offline and privately without complicated setup. No subscription.

• Zero-config setup
You don't need technical knowledge to configure local models. RecurseChat handles model downloads and serving. Supports Meta Llama 3.1, Meta Llama 3, Gemma 2, Mistral, Llava, WizardCoder Python and GPT-4/GPT-4o/GPT-4o mini/Claude 3.5/Claude 3 as default models, as well as custom GGUF and OpenAI compatible endpoints. Detailed model support information: https://recurse.chat/faq/#what-ai-models-does-recursechat-support

• Local document chat
Chat with local PDF, markdown and txt documents, completely offline. Attach, or just drag and drop your local doc in the chat UI, ask question specifically about your documents, and get detailed answers including citations of the original document. RecurseChat implements Retrieval Augmented Generation (RAG), and uses local embedding model and vector database.

• Full text search
Blazingly fast full text search over thousands of chat messages.

• Import ChatGPT chat history
You can import all of your ChatGPT chat history, and continue your previous chats using either local models or GPT models. Ever want to find an old chat in ChatGPT? Just import the chat history and type in the search box.

• Beautifully designed
Chat with ease in a beautifully designed interface.

• Organize chats
Organize chats with bookmarks and folders. Supports nested folders.

• Edit and branch messages.
Edit messages and switch between different message branches freely. Imported ChatGPT messages also have branches preserved.

• Multiple models in the same chat session
You are not limited to one LLM in a chat session. Switch models in the middle of chat session.

• Multimodal chat
Supports image understanding model such as LLaVA.

• Use as a ChatGPT client
Not limited to local models. Supports bringing your own OpenAI API key. (* if you opt to use OpenAI key, you are subject to OpenAI's privacy policy. Details here: https://recurse.chat/privacy/). If you use the local document / RAG feature, you can still rely on the local embedding feature and only use the API for chat completion.

• Customizable
Customize your AI personality by changing the name. Adjust font family and font size to your preference.

• Bring your own GGUF
Link your local GGUF file and chat with it, or add models from Hugging Face Hub.

Support: https://recurse.chat/support/
EULA: https://www.apple.com/legal/macapps/stdeula/
Privacy policy: https://recurse.chat/privacy/

What’s New

Version 1.2.2

- Performance improvements

Thank you for choosing RecurseChat. We are committed to continue improving the app and make it more useful for you. For support requests and feature suggestions, please feel free to email us at support@recurse.chat, or tweet at @recursechat. If you enjoy the app, leaving an app store review is a great way to support us.

Ratings and Reviews

4.6 out of 5
17 Ratings

17 Ratings

UniqueAirStream ,

RC is keeping up with the ChatGPT and great PDF Support

I've had the pleasure of using the Llama model for some time now, and I must say that I'm thoroughly enjoying my experience with this chat history app! The automatic chat titling feature has been particularly impressive, as it allows me to easily keep track of old conversations. Additionally, the app has better accurate responses from LLMs and beter converstional and memory history of previous thing mentinoed in the converstation when compared to other ondevice LLM apps I've tried so far.

What's more, the PDF support within the app works seamlessly on device, which is a major plus for me since I prefer not to rely on cloud storage whenever possible. All in all, I believe that Llama is truly the best LLM (Large Language Model) app available for macOS - its features and functionality have exceeded my expectations.

Developer Response ,

Thank you so much for your comprehensive and enthusiastic review! We appreciate your support. If you have suggestion for helping make the app more useful for you, we would love to hear about it.

chad.mckenna ,

Super Functional

I've been using RecurseChat as my primary tool for interacting with LLMs and its been perfect. Its simple and to the point, has all the features one might need: interacting with local LLMs (like the ones provided), chatting with Ollama models through the custom option, great way to interact with OpenAIs models, also provides RAG on local data. I've never had an issue with the UI, it just works. The developer is also super friendly and very helpful with any of my questions.

Developer Response ,

Thank you so much for your comprehensive review! Appreciate your support. Glad that you are finding RecurseChat useful for your workflows.

mpmcfarlane ,

Excellent

Good privacy commitments, very good app overall. I like being able to use on device and commercial API models in the same app. Built in RAG is great. Would like some more transparency on how/when RAG is used. I like that I can use a custom GGUF, but adding a few larger default models would be good too, so I didn't have to bother with the GGUF myself. Would also love to see Latex formatting support.

Developer Response ,

Thank you so much for your review! We will continue to commit to privacy. We are thinking about ways to enhance RAG usage customizability and visibility, but it might take some time to get to it. Latex support is pretty high on the priority list. (Edit: LaTeX rendering support is added in v1.1.10.) For default models, please feel free to email us at support@recurse.chat with the ones you would like to see.

App Privacy

The developer, Xiaoyi Chen, indicated that the app’s privacy practices may include handling of data as described below. For more information, see the developer’s privacy policy.

Data Not Collected

The developer does not collect any data from this app.

Privacy practices may vary, for example, based on the features you use or your age. Learn More

Supports

  • Family Sharing

    Up to six family members can use this app with Family Sharing enabled.