This package provides `gptel-inline', a way to interact with an LLM in a
persistent gptel session from anywhere in Emacs.  Responses appear in
an overlay at point rather than requiring the user to visit the chat
buffer.

`gptel-inline' displays a buffer for prompt entry (in a small window) and
injects selected text/surrounding context as reference material if you want.
When the query is sent, LLM responses are shown inline.  The primary chat
buffer remains in the background and provides the full conversation history
context -- you don't need to switch it.

By default, the chat buffer chosen for the conversation is one used
previously, or a chat buffer created by gptel-inline for the current project.
You can customize this by adding to the hook
`gptel-inline-find-chat-buffer-functions'.

References are context snippets selected by cycling through
"things at point" (region, sexp, defun, paragraph, window, buffer,
etc.) using `gptel-inline-cycle-reference'.  The reference type is
determined by `gptel-inline-reference-types'.

To send the request, use `gptel-inline-send' in the prompt window.  (This is
bound to whatever `gptel-send' is bound to in chat sessions.)

You can use `gptel-inline-append' to build up a query by appending multiple
prompts to the chat session without sending them.  You can use this for
reviewing multi-part code or prose generated by the LLM.

Call `gptel-inline-help' ("C-c ?" by default) for a quick overview of how to
use the inline window.