HuggingFace backend for AI mode

Overview

This package provides a code autocomplete backend through HuggingFace Inference API or local Text Generation Interference server for ai-mode extension.

Installation

First of all, install ai-mode, if you haven't already done so.

Installation via MELPA

It's easiest/recommended to install from MELPA. Here's a minimal MELPA configuration for your ~/.emacs:

(package-initialize)
(add-to-list 'package-archives '("melpa" . "http://melpa.org/packages/"))

Afterwards, M-x package-install RET ai-mode-hf RET (you might want to M-x package-refresh-contents RET beforehand if you haven't done so recently).

Installation via GIT

Clone this repository somewhere

$ cd ~/.emacs.d/plugins
$ git clone --recursive https://github.com/ai-mode/ai-mode-hf

Add the following in your .emacs file:

(add-to-list 'load-path
              "~/.emacs.d/plugins/ai-mode-hf")
(require 'ai-mode-hf)

or you can use use-package

(use-package ai-mode-hf
  :load-path "~/.emacs.d/plugins/ai-mode-hf`"
  :config (progn
            (add-to-list 'ai-completions--backends '("HuggingFace" . ai-mode-hf--completion-backend))

            ;; Optionally, we are setting up HuggingFace as the default backend for AI completions.
            (setq ai-completions--current-backend #'ai-mode-hf--completion-backend)
            ))

Configuration

To use HuggingFace cloud backend, you need to obtain an authorization key. You can get the key on the page.

There are several ways to assign your key to the variable ai-mode-hf--api-key.

1. Put (setq ai-mode-hf--api-key "you key") in the configuration file.
2. Use the command M-x set-variable ai-mode-hf--api-key.
3. Use the command M-x customize-variable ai-mode-hf--api-key.

Configuration variable	Description
ai-mode-hf--request-timeout	Default request timeout
ai-mode-hf--base-url	API base
ai-mode-hf--completions--url	Completions api
ai-mode-hf--model	Used model
ai-mode-hf--stop-tokens	List of stop tokens
ai-mode-hf--default-max-tokens	Max number of tokens in response
ai-mode-hf--model-temperature	Sampling temperature
ai-mode-hf--top-p	Top-p sampling

Add the HuggingFace completion backend to ai-completions--backends:

(add-to-list 'ai-completions--backends '("HuggingFace" . ai-mode-hf--completion-backend))

Now you can choose available backends by calling the command ai-completions-change-backend.

If you want to change the default backend, you need to redefine the value of the variable ai-completions--current-backend in your Emacs config.