I tried to code a plugin for copilot but soon realized that your api doesn't allow low level control of scintilla or I'm too noob to realize how. Copilot only needs to grab context from current cursor position (say 200 chars before and after), then it will suggest autocompletion options. Btw, Copilot is not the only tool for the task, I plan to add and OpenAI backend that can be gracefully replaced by LocalAI for 100% private and off-line coding (so no IP issues). Devs can choose to train/finetune their own LLM if they wish, plug it into LocalAI and let the geany plugin take care of the rest.