Verity JE Extra

Compatibility update for Verity JE 6, which restructured the classes this addon hooks into. Requires Verity JE 6.0+ and YACL 3.6+. Beta.1 no longer loads against Verity 6 at all, since its mixin targets were renamed.

Added

• Offline Voice now uses a native file picker instead of a typed path. The field is replaced by an "Import .onnx..." button that opens the operating system's own file dialog, filtered to .onnx files.
• The Offline Voice button reflects the current selection. With no voice chosen, it reads "Import .onnx..." in gray italics. Once a voice is selected, it displays the file name in plain white text instead.

Fixed

• All three overrides (Chat, Cloud Voice, Offline Voice) retargeted to Verity 6's new class layout. Verity 6 moved chat handling into a new LLMBuilder class, TTS handling into a restructured AiAPI, offline voice into LocalTTSBuilder, and the config screen into VerityConfigUI. Same three features as beta.1, same fields, rebuilt against the new classes.
• Chat override now forces Verity's LLMBuilder into its OpenAI dispatch branch. Verity 6 added its own AI Provider setting, but it is a fixed dropdown (Gemini, Groq, OpenRouter, Mistral, Ollama, OpenAI). For four of those six choices, Verity hardcodes the real endpoint internally and ignores whatever is typed into its own endpoint field. Only Ollama and OpenAI honor a free form endpoint, so this override forces the OpenAI branch and feeds it this addon's own Endpoint URL, API Key, and Model Name, giving a genuinely arbitrary OpenAI compatible endpoint regardless of Verity's dropdown.
• Cloud Voice override rebuilt against Verity 6's playEndpointTTS. Verity 6's own cloud voice options still do not support a keyed custom endpoint: Groq's endpoint is hardcoded, and its Kokoro option sends no API key at all (self hosted, keyless only). This override takes over the whole request when enabled, using its own TTS Endpoint URL, API Key, Model, and Voice fields.
• Fixed a game crash when opening the file picker. The initial implementation used a Swing file chooser, which conflicts with Minecraft's own window and rendering system on some setups. Replaced with LWJGL's native file dialog, which is already bundled with the game for this exact purpose.

Known limitations (beta)

• Same limitations as beta.1 below, now against Verity 6's equivalent settings and paths.
• Tied to Verity 6's internal class and method names. A future Verity update could rename or restructure these again. Mixins are set to defaultRequire: 1, so an incompatible update fails loudly at launch rather than silently ignoring your settings.

[1.0.0-beta.1] — 2026-07-28

First public release. Requires Verity JE 5.0+ and YACL 3.6+ (YACL is already required by Verity JE itself, so nothing extra to install if Verity is working).

Added

• Verity JE Extra tab, appended to the end of Verity's own config screen. Three groups, each gated behind its own tickbox — fields stay greyed out until enabled, and nothing is touched in verity-client.toml unless you turn a group on.

• Chat — Override Verity JE AI Settings Redirects Verity's chat requests to any OpenAI-compatible endpoint (OpenRouter, a local LiteLLM/Ollama server, a different hosted model, etc.) instead of the GROQ/OPENROUTER choice in Verity's own settings. Fields: Endpoint URL, API Key, Model Name.

• Cloud Voice — Override Verity JE TTS Settings Redirects Verity's cloud speech requests independently of the chat endpoint, with its own API key — cloud voice no longer has to share Verity's single apiKey with whichever chat provider you've pointed at. Defaults match Verity's stock Groq/Orpheus setup exactly, so turning this on with unmodified defaults changes nothing. Fields: TTS Endpoint URL, TTS API Key, TTS Model, TTS Voice.

• Offline Voice — Use A Custom Offline Voice Points Verity's built-in CPU speech engine at a Piper VITS model of your choosing, in place of the bundled en_US-ryan-medium. Fully offline — no server, no key, no network call. Expects the standard sherpa-onnx Piper layout (tokens.txt and espeak-ng-data/ alongside the .onnx). Falls back silently to Verity's bundled voice if the path is empty or anything's missing. Field: Voice Model Path.

Known limitations (beta)

• Voice input is not covered. Whisper transcription (audio/transcriptions) is built by Verity outside the code path this mod hooks, so it always uses Verity's own apiKey against Groq regardless of any override here.
• Cloud Voice only fires when Verity's own useNativeTts and useLocalTts are both off — that priority order is Verity's, not something this mod changes.
• Verity refuses cloud speech outright when aiProvider = "OPENROUTER" in its own settings, independent of anything this mod does. Leave aiProvider = "GROQ" if you want cloud voice; the Chat override still redirects the LLM regardless of that field.
• Tied to Verity 5.7.3's internal method names, including one synthetic lambda (lambda$playTTS$5) that a future Verity update could rename or restructure. Mixins are set to defaultRequire: 1, so an incompatible update fails loudly at launch rather than silently ignoring your settings — it will not fail quietly.
• API keys are stored in plain text in config/verityjeextra.json, same as Verity's own config storage.

Building

No Gradle, no ForgeGradle, no downloads — compiles directly against the jars an existing Modrinth Forge 1.20.1 install already has on disk. See README.md.