TypingMind

Claude has an option to enable/disable thinking in chat. In OpenAI models, to change reasoning effort, I need to go to the model option or use an agent with changed model config.

Presently, when starting a new chat without selecting an agent, there's no option to modify the default system prompt. This forces you to override it manually in the first message, which is inconvenient, especially because the existing "o3" prompt instructs the assistant to use "$$" around mathematical content. As a result, sometimes the assistant incorrectly uses "$$" around code snippets, breaking the formatting.

Add support for Gemini Live API. https://ai.google.dev/gemini-api/docs/live This provides users with the experience of natural, human-like voice conversations, and with the ability to interrupt the model's responses using voice commands. The model can process text, audio, and video input, and it can provide text and audio output. I believe there is other feedback requesting live or real-time API support for other AI model providers such as OpenAI and Anthropic. This would add the same for Gemini.

The idea is that the user can save a combination of tags as a set to eaily filter chats. Also, when the user clicks our of main chat area into, for example, Agents, then clicks back to chat, the filter should still be applied. Currently, the filter gets cleared and the user then has to reapply the filter.

The title describes it. Hide disabled models & Include icons for vision, plugins, and context length.

While hovering the mouse over a model show a hint with the input/output token costs per M+. Or maybe show the costs in a column next to the context size. Like ($2/$8). This would help selecting an appropriate model. i.e. Perplexity api has a couple models that are very expensive and it would be useful to see the costs beforehand.

In the Models tab in the sidebar, we should have more information there like: * Price. * Release Date.

Allowing users to select a custom icon instead of the default "folder" icon will significantly help organize projects and create an extra layer of contex to the project, allowing quick visually understanding the folder intention. The icons could be uploaded by user, use urls, select from a detaul set (supplied by TypingMind), or use Emojis. Any single, or combiniation of these is a good start. To enhance custom icons implementation further, folders could be filtered by, sorted by, and grouped by "icon" in the sidebar. These icons could serve as larger Projects concepts and have their own page with their own details (description, etc). The idea is to help organize content better. I find that to be a major flaw in TypingMind (and any LLM UI). Create a breakthrough in how chats / projects / groups etc. are organized. I have several other feedback entries that builds on this concept.

Hello, I’m a TypingMind user who accesses the platform via a screen reader. When I began using Claude 3.7 Sonnet, I discovered (after the fact) that Thinking Mode appears to be enabled by default, but there is no screen-reader-accessible indication of this. Unfortunately, this caused me to unknowingly rack up over $80 in charges within one week, due to increased token usage associated with Thinking Mode. I respectfully request: A clear, screen-reader-friendly toggle or checkbox indicating whether Thinking Mode is ON or OFF. An option to set Thinking Mode OFF by default across sessions and models. This will greatly improve accessibility and allow users to control costs transparently. TypingMind has been receptive to accessibility concerns in the past, and I trust you’ll give this proper consideration. Thank you!

1) Add better icons for Python and Javascript plugins. 2) Add a feature that lists all active plugins when you hover over that section. - The reason, when plugins are locked, you don't know what plugins are active.