Honyac Usage Guide

Requirements

• macOS 14 or later

Setup

1. Place Honyac.app in the /Applications folder
2. Launch it — a translation icon appears in the menu bar
3. Open Settings… from the menu
4. Pick a connection type (Ollama or an OpenAI-compatible API) and set the translation and explanation models

Keep Honyac.app in the Applications folder so it can register its macOS service for right-click translation.

Ollama is supported out of the box. Honyac connects to http://127.0.0.1:11434 by default.

Steps

1. Choose "Ollama" as the connection type in Settings
2. Press "Get models" to pick from the models installed in your Ollama
3. Select the translation model and the grammar explanation model
4. Run the connection test

After each translation or explanation, Honyac asks Ollama to unload the model (keep_alive: 0), so memory is not held while idle.

Any service that exposes an OpenAI-compatible chat API can be used as an endpoint.

Examples

Service	Notes
OpenClaw (VPS)	Service-side model aliases may be available
Hermes Agent	Exposes an OpenAI-compatible API
Open WebUI	Local UI/API server
LiteLLM	Proxy that fronts multiple backends with an OpenAI-compatible API
LM Studio	Local LLM runtime
vLLM	Fast inference server

Configuration

1. Choose "OpenAI-compatible API" as the connection type
2. Enter the API base URL and API key
3. Set the model names for translation and explanation — real model names, or service-side aliases such as OpenClaw’s where available
4. Verify the URL, authentication, and model names with "Connection test" before use

Security notes

• Enable HTTPS and authentication
• Never expose the API port without authentication

Where your text goes depends entirely on your connection settings. Refer to each service’s documentation for its own setup.

Honyac keeps two model roles and lets you assign them separately.

The two roles

• Translation model — called only when you translate
• Grammar explanation model — called by the "Explain grammar" button and the question chat

Example pairings

• A small, fast model for translation plus a model that explains well for grammar
• Start with the same model for both, then split them once you feel the difference

You manage the models and endpoints yourself; no cloud API key is required.

Both the translation prompt and the grammar explanation prompt are multi-line editable and saved.

What you edit

Prompt	Used by
Translation prompt	The translation model
Grammar explanation prompt	The explanation model (also used by the question chat)

Template variables

Prompts can reference the source text, the translation, and the language as template variables — useful for instructions like "explain this source text and its translation".

Reset

Both prompts have a reset button that instantly reapplies the current default prompt, so experiments are easy to undo.

What the default translation prompt protects

• No free interpretation, invention, or added information
• Preserves the meaning of emotion, evaluation, auxiliaries, negation, tense, idioms, phrasal verbs, sentence count, paragraphs, and line breaks
• Does not soften attitude nuances like be excited to into neutral wording

The language used for grammar explanations is configurable.

• It defaults to your macOS system language
• Choose Japanese and grammar and vocabulary notes appear in Japanese
• Question chat replies also follow the configured explanation language

The global shortcut can be changed to any combination you like from Settings.

Behavior

• The default is ⌥⌘H
• Any combination including ⌘, ⌥, or ⌃ can be registered
• Combinations already in use by other apps or macOS are detected as conflicts
• If registration fails, the previous shortcut is kept
• Check the current shortcut and its registration status in Settings

Select text in Safari or Chrome, choose right-click → Services → "Honyacで翻訳", and Honyac opens and translates the selection.

If the menu item does not appear

1. Make sure Honyac.app is in the Applications folder
2. Quit and restart both Honyac and the browser
3. Check "System Settings → Keyboard → Keyboard Shortcuts → Services" to confirm the service is enabled

OpenAI-compatible API settings include a "Connection test".

It checks that the base URL, authentication, and model names you entered work as expected. Try it whenever you add a new endpoint or change a model name.

• Translation is performed by the LLM you configure as the endpoint
• With a local setup such as Ollama, your text can be translated without sending it to an external cloud
• When using an OpenAI-compatible API, enable HTTPS and authentication, and never expose the API port without authentication
• The models and endpoints are chosen by you
• Actual network behavior depends on your connection settings

What are the system requirements?

macOS 14 or later.

Can I use it without Ollama?

Yes. Local setups that expose an OpenAI-compatible API (LM Studio, Open WebUI, vLLM, …) and VPS services like OpenClaw all work.

Can I pick separate translation and explanation models?

Yes. Assign them individually in Settings, or use the same model for both roles.

What happens if the shortcut conflicts with another app?

The conflict is detected, and if registration fails the previous shortcut is kept. You can check the status in Settings.

Can I change the explanation language?

Yes. It defaults to your macOS system language and can be changed in Settings.

Are more languages planned?

The initial release covers Japanese ⇄ English. Support for more languages is planned as a future paid activation (Pro) feature.