increa.el

Intelligent code completion for Emacs using LLM APIs. Provides GitHub Copilot-like ghost text completions.

Supports OpenAI-compatible APIs.

Installation

Doom Emacs

Add to packages.el:

(package! increa :recipe (:host github :repo "bencode/increa.el"))

Add to config.el:

(use-package! increa
  :config
  (setq increa-api-endpoint "https://dashscope.aliyuncs.com/compatible-mode/v1/chat/completions"
        increa-api-key (getenv "QWEN_API_KEY")  ; or hardcode your key
        increa-model "qwen3-coder-flash")
  (global-increa-mode 1))

Then run doom sync.

Straight.el

(straight-use-package
 '(increa :type git :host github :repo "bencode/increa.el"))

(require 'increa)
(setq increa-api-key "YOUR_API_KEY"
      increa-model "qwen3-coder-flash")
(global-increa-mode 1)

Manual Installation

1. Install dependency:

M-x package-install RET request RET

2. Clone and load:

git clone https://github.com/bencode/increa.el.git ~/.emacs.d/increa.el

(add-to-list 'load-path "~/.emacs.d/increa.el")
(require 'increa)

Configuration

Qwen-Coder

Tested and recommended. Get API key: https://help.aliyun.com/zh/model-studio/get-api-key

(setq increa-api-endpoint "https://dashscope.aliyuncs.com/compatible-mode/v1/chat/completions")
(setq increa-api-key "YOUR_API_KEY")
(setq increa-model "qwen3-coder-flash")
(global-increa-mode 1)

Available models via Alibaba endpoint:

• qwen3-coder-flash - Fast, recommended for code completion
• qwen3-coder-plus - More powerful but slower
• glm-4.5-air - Zhipu AI's GLM model

Moonshot (Kimi)

Tested. Get API key: https://platform.moonshot.cn/console/api-keys

(setq increa-api-endpoint "https://api.moonshot.cn/v1/chat/completions")
(setq increa-api-key "YOUR_API_KEY")
(setq increa-model "kimi-k2-0905-preview")
(global-increa-mode 1)

OpenAI

Tested. Get API key: https://platform.openai.com/api-keys

Note: May require HTTP proxy configuration in China.

(setq increa-api-endpoint "https://api.openai.com/v1/chat/completions")
(setq increa-api-key "YOUR_OPENAI_KEY")
(setq increa-model "gpt-4o-mini")
(global-increa-mode 1)

Other OpenAI-compatible APIs

Any OpenAI-compatible endpoint should work. Just set the three parameters: increa-api-endpoint, increa-api-key, increa-model.

Advanced Options

;; Control which modes to enable completion
(setq increa-enabled-modes '(prog-mode text-mode))  ; default
;; Or specific modes:
;; (setq increa-enabled-modes '(python-mode js-mode emacs-lisp-mode))
;; Or all modes:
;; (setq increa-enabled-modes nil)

;; Behavior
(setq increa-idle-delay 0.5)            ; trigger delay
(setq increa-context-max-chars 100000)  ; context size
(setq increa-max-tokens 500)            ; completion length
(setq increa-temperature 0.2)           ; randomness

;; Enable for specific modes only (alternative to increa-enabled-modes)
(add-hook 'python-mode-hook #'increa-mode)
(add-hook 'js-mode-hook #'increa-mode)

Usage

• Type code, wait 0.5s for gray ghost text
• C-<return> (Ctrl+Enter) to accept
• C-g to dismiss
• C-c i to trigger manually

Providing Hints to AI

Use ai: or ai? comments near the cursor to give hints:

def process_data(items):
    # ai: filter out None values and sort by price

The AI will see these hints and generate code accordingly.

License

MIT