This is an Visual Studio 2022 extension that makes use of the OpenAI API, to generate and read out loud messages with a custom personality, whenever a project fails to build, or a runtime exception is hit.
Generate messages about build errors and exceptions whilst coding, using OpenAIs GPT-3.5-turbo model. Uses code snippets and error messages for increased context.
Read aloud the generated messages, using built in Windows Text To Speech.
Customize the personality of the messages with the personality editor. Specify a prompt for both Build errors and Exceptions. A default "Rude" personality is included in the extension, as well as the "Helpful" and "Maid" personality.
The editor is found under:
Tools > Options > VSCodingBuddy > Personalities
Limit number of tokens used using a number of strategies.
In order to use this extension, an OpenAI API key is needed.
For a guide to retrieve this key, please see the OpenAI Setup section.
An OpenAI key is needed in order to use the OpenAI API. This is not free, however OpenAI gives you a free 5$ for the first 3 months of usage. The pricing is also in general extremely cheap (<0.002$ for a messages). This number can vary depending on the Token limiter strategies used.
To retrieve an OpenAI key follow the steps below.
The VSCodingBuddy extension needs the OpenAI key in order to function. To inform the extension about this key, follow the below steps.
The extension is now setup and should function normally. Additional options can be found under the General and Personalities options page.
All configurable options can be found under the Tools > Options > VSCodingBuddy page.
Each option under the general page contains a brief description, which can be referred to.
This menu should be used, when editing, adding or deleting personalities.
Whenever the name or prompt fields are edited, the save button can be used to override the selected personalities values, or the new button can be used, to create a new personality, with the current name and prompts.
prompts can optionally be compressed, by pressing the compress button. This works by telling a GPT-3 model to summerise the prompt, which leads to a smaller number of tokens for the prompt. This compression strategy does not always produce a desired result, so make sure to make a backup of the original prompt.
Visual Studio Marketplace -> https://marketplace.visualstudio.com/items?itemName=zarstensen.vscb-1