Artifact-like Output Window for gptel

[guibor]

Description
Implement an artifact-like output window for gptel that displays substantial, standalone content generated by the AI in a dedicated buffer separate from the main conversation. This feature would enhance gptel's functionality by providing a more interactive and visually appealing way to work with large AI-generated outputs.

Proposed Functionality
Dedicated Output Buffer: Create a new buffer called "gptel-artifact" to display large AI responses.
Side-by-Side Display: Show the artifact buffer alongside the main gptel conversation buffer using display-buffer-in-side-window or a similar function.
Syntax Highlighting: Apply appropriate major modes to the artifact buffer based on content type (e.g., markdown-mode, org-mode, or language-specific modes for code).
Version Control: Implement a simple version history for artifact content, allowing users to revert changes or view previous versions.

[guibor]

I'm trying to replicate this flow with gptel-write and adding a buffer to the context, but it doesn't exactly match. I would like to be able to provide instructions in a gptel buffer, and get results in another buffer (e.g., a .py file)

[karthink]

I would like to be able to provide instructions in a gptel buffer, and get results in another buffer (e.g., a .py file)

gptel does not work like this out of the box, but it provides redirection options that can help:

1. In your chat window, set the system message to "generate only code, no comments or markdown, etc" -- it'll need to be more specific and longer.
2. Make sure that you've provided enough context, either by adding a buffer to the context or including it in your chat buffer.
3. In the python buffer make sure that the cursor is where you want code to appear.
4. In the chat window, type in your instructions and choose the "Response to any buffer (b)" option in the transient menu.
5. Hit Send in the transient menu.

For a dedicated tool that works this way out of the box, you can check out the Elysium package. There may also be other LLM clients that work like this.

[karthink]

Dedicated Output Buffer: Create a new buffer called "gptel-artifact" to display large AI responses.

I don't understand the need for this feature. Could you explain what a "large AI response" is, when you would see it, and how it would be identified?

[karthink]

Moving to discussions as this is currently not planned.