v1.2.1

Improvements

• The logic of the workflow now allows to use the new Open AI model: chatgpt-4o-latest.
• A script action block has been incorporated in the workflow for it to auto-download new versions. To opt out of auto-updates, simply change the version of your downloaded workflow to something higher (like 5... or 10). You can also simply remove the OneUpdater block.

Some time in the future I am planning a rewrite of the documentation—separating it in different sections and creating walk-through videos for different settings & features. This workflow is very powerful, flexible, and I use it all the time, but I am aware it may also be hard to understand for users. If you have any questions you'd like me to address, if you wonder about how to use a feature of the workflow, or if you'd like more examples about something specific let me know and I'll keep this in mind when working on the new docs.

v1.2.0

GROQ SUPPORT

• Kiki now has a dedicated text field for users to input their Groq API Token.
• Models from Groq auto-direct requests to the correct API URL without any additional prefix on their models.
  • NOTE: The custom_ prefix is only necessary for models using the API Endpoint URL in the configuration. This may change in the future if the service providers opt for different naming conventions for their models.
• Kiki can now work with Groq's Whisper AI, which is faster than OpenAI's. There's an option to select this in the workflow configuration.
  • I am hoping to find an app which can integrate Whisper AI with Kiki locally via Applescript or a CLI (MacWhisper still can't do this). If you know of any that could work feel free to let me know.

IMPROVEMENTS

• The "All-in-One" trigger can now receive a fifth argument with tab-separated paths to audio files to be transcribed by Whisper AI. This allows you to more easily integrate the Whisper AI feature with other automation tasks. For example, you can now send a command like "direct,whisper,1,translate,path1 path2 path3" to Kiki via Applescript.

v1.1.9

Improvement

• Kiki can now take multiple audio files as input (using Alfred Universal actions) and it will transcribe them using Whisper's API. The resulting transcriptions will be in a list format.

v1.1.8

Improvement

• Users can now set an "API Endpoint Token" to use with the "API Endpoint URL". This allows Kiki to integrate with OpenAI-compatible API services, such as Groq, in addition to local LLM tools. Remember to use the "custom_" prefix when specifying models to use with your API Endpoint URL.

v1.1.7

Bug Fix

• Fixed a bug where if the user left the API endpoint URL blank, the AI request would fail.

v1.1.6

Improvements

• Whisper AI's "Paste" functionality now integrates with Alfred's Large Type placeholder, ensuring that the user won't move to a different window before inserting the results.
• Kiki's configuration now includes new "placeholder" options, allowing users to assign unique identifiers to reusable content across requests and presets.

v1.1.5

Bug Fixes

• Fixed a bug where the combination of the FN + Shift modifiers would trigger an error for 'incorrect preset'.
• Fixed a bug where Kiki would ignore the user's input when using the 'Add a Custom Prompt' preset on selected text with a single or multiple modifiers.

v1.1.4

Bug fix that was preventing Kiki from following up conversations with online models.

v1.1.3

IMPROVEMENTS

• Kiki can now have both, a custom API Endpoint URL & online services. Some modifications have been made so the user can make requests to both online and offline models interchangeably. The documentation has been updated to include THIS SECTION which explains further.
• If OpenAI API Token is not set, user won’t have the Whisper AI transcription option in the main menu, in this way avoiding error messages.

v1.1.2

CUSTOM API ENDPOINT
Kiki now supports a custom API endpoint, which allows it to work with local/offline/custom models.
Thanks to @iandol for the contribution!

It's important to note that with this update, Kiki supports any of the other APIs (as before) or a Custom API Endpoint if set, but not both... yet.