Feature Requests

I find very annoying that every time I want to simply ask a question, I need to manually click on the "X" of the current app context to remove it (e.g. I open Kerlig while viewing a web page, and it shows me the web page as attached to the chat, even though I don't necessarily want it included). My ideal scenario is that I open it via a shortcut, and if I press delete (with the text input empty), it automatically removes the current context.

I've accumulated a bunch of presets that I constantly switch between. Yes, they can be switched with ⌘+P, but when there are too many, you end up scrolling through them in a loop, which sucks! I suggest adding the ability to change the order of presets in the settings and plus the ability to activate a specific preset by pressing ⌘+1, ⌘+2, ⌘+3, etc. Would be awesome! Thanks!

I know this is a big lift but it would be awesome if somehow Kerlig was my main MCP client with an option to choose the inference server like ollama or OpenAI

In Kerlig 2.3.0, we can add emails as context to the LLMs via the Kerlig interface. This is neat, but I'm not sure it's reading the URLs from the emails. So, can you confirm if the email content is read as html/rich text, or just plaintext (in which case the URLs are obviously lost). Can you enhance this feature to include the URL links from the emails, so we could build useful workflows?

When users upload a PDF in Kerlig, configure the OpenRouter API call to use plugins.pdf.engine: 'native' for models supporting native PDF input. This leverages the model's direct PDF understanding, improving accuracy and efficiency. Reference: https://openrouter.ai/docs/features/images-and-pdfs

I have my HyperKey set to Control+Option+Command, and I'm finding that when I attempt to use it to automatically trigger a custom action, it seems like three keys is all that the custom action will accept. So I'm unable to use Control+Option+Command+G, for example, to trigger an action because the moment I input that fourth key, it just clears the text field. Any plans to increase the amount of keys that can be used as modifiers for the custom actions? Thanks!

Hello! Cerebras is a ridiculously fast provider that has a very generous free tier, which would make it great for Spelling/Grammar fixes. It supports the OpenAI API spec partially ; the catch is that it does not support the following fields: frequency_penalty logit_bias presence_penalty parallel_tool_calls service_tier Those have to be stripped out of the API request that you make, otherwise it fails. Can you add Cerebras as a default provider and make the changes it needs?

Please add a URL scheme for Kerlig, so that apps like PopClip can access Kerlig actions using kerlig://. It is easier to integrate into workflows that manipulate text selection without needing to use a keyboard shortcut.

I often engage in lengthy discussions on various topics and would appreciate the ability to export full chat conversations as a single Markdown file. This feature would enhance usability, improve documentation of conversations, and simplify the process of tracking information without the need to sift through individual segments for one chat.

Tahoe now enables access to Apple's on-device LLM from within apps ( https://developer.apple.com/documentation/foundationmodels ). Integrating this model within Kerlig would be wonderful! Here is how someone has already experimented with that approach: https://github.com/gety-ai/apple-on-device-openai