Typing Mind

I am excited to see what plugins and extension people are creating.

Can we have conversations where we can choose multiple bots to talk to - similar to Silly Tavern? https://sillytavernai.com/ For example, then we could have brainstorming sessions with different bots with different skills, or learn languages in a room with multiple bots for more immersion. This will also open up the program to more people.

Claude recently started to show <thinking> </thinking> . Since we likely pay for this output when it's not essential, is it possible to not show/use this part of the tokens used to reduce cost?

If you have an agent do research and browse the web, it can go through a lot of content fast (each chat interaction can cost like $0.30+). option to NOT sent all this content back and forth for future messages in same chat. Example: it does some research and then uses it to response with an answer. once that's done, all content becomes (expensive!) dead weight for all future interactions.

There is a program called ChatAll that gives access to many models (free, not free, chatgpt, openAI API, Claude API, Groq API, more). https://github.com/sunner/ChatALL Typingmind has a great UI - beats everything else, however the limited number of models is constraining it's use. Could you please add native Groq API access (currently the API is free) so we can use llama3-70b (and others), and, if possible, other free options (eg. access to chatGPT(not the API)) - though free Llama3 access would be ideal. ChatAll includes many models in their (sparse) interface that are accessed via the web (not API) including copilot, huggingface models, Poe. Adding these would be a game changer, (I've found that llama3-70b is a very capable model, and for general queries is up there with claude3 and chatgpt4, so I'm using it for free (via huggingchat or groq.com ) as a main AI and then using claude3/chatgpt4 via TypingMind for any queries that I need second opinions for. Being able to use llama3 in TypingMind would be fantastic.

I would like to be able to just login into TypingMind cloud using Google on a new device and have all my history and API keys loaded so I don't have to bother with getting a new API key. It's not a big risk if API key is leaked because of DB hack, so I would like to see this implemented. (It is a bigger risk if DB hacks leaks our chats, those private chats are more important to me than extra $100 in bill because someone used my leaked key)

I want to be able to enter (something like) https://www.typingmind.com/#agent=MyAgent (and save it as a bookmark) instead of going to typingmind, and then selecting the agent. it would be particularly useful on mobile. I currently have a custom extension to do this, but native support would be much better and would work uniformly across all machines.

Add the cost. Additionally, it would be nice to sort and filter by cost and feature.

When will other file formats be supported? Oftentimes epub files are smaller to upload and perhaps even faster to go through?

When you pick a model and want to switch to another one of the same family (e.q Claude) you have to scroll manually to the selected one.