Skip to content

CLAUDE.md

Latest commit

 

History

History
110 lines (82 loc) · 3.93 KB

CLAUDE.md

File metadata and controls

110 lines (82 loc) · 3.93 KB

CLAUDE.md

This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.

ユーザとの対話は日本語で行うこと。

気をつけること

  • 本プロジェクトはDLL形式で動作する。メモリ解放、ヒープ管理の安全性に最新の注意を払うこと。
  • リファクタリングの際はデグレーションを防ぎつつも、パフォーマンスや安全性のための最適化を積極的に行うこと。

Project Overview

GhostSpeaker is a Rust-based SHIORI plugin for 伺か (Ukagaka) that enables text-to-speech functionality for ghost characters using various TTS engines. It's built as a Windows DLL (ghost_speaker.dll) that integrates with the 伺か ecosystem.

Build Commands

Standard Development

# Build in release mode (standard)
cargo build --release

# Build for 32-bit Windows (production target)
cargo build --release --target=i686-pc-windows-msvc

# Run tests
cargo test

# Format code (uses 2-space indentation, no hard tabs)
cargo fmt

# Check for clippy warnings
cargo clippy

Production Build

The project is configured to build 32-bit Windows DLLs for compatibility with 伺か systems:

  • Target: i686-pc-windows-msvc
  • Output: ghost_speaker.dll (copied to project root)
  • CI builds automatically update MD5 checksums via GitHub Actions

Architecture

Core Components

Plugin Interface (src/lib.rs)

  • SHIORI plugin entry points: load(), loadu(), unload(), request()
  • Handles UTF-8/ANSI encoding conversion for Japanese text
  • Manages plugin lifecycle and panic handling

Engine System (src/engine/)

  • Modular TTS engine support via trait system
  • Supported engines: COEIROINK (v1/v2), VOICEVOX, LMROID, SHAREVOX, ITVOICE, AivisSpeech, 棒読みちゃん
  • Each engine has dedicated modules with speaker.rs and predict.rs
  • HTTP-based communication with TTS services (except 棒読みちゃん)

Event Handling (src/events/)

  • menu.rs: Plugin menu system integration
  • periodic.rs: Timer-based events
  • other_ghost.rs: Inter-ghost communication
  • common.rs: Shared event utilities

Audio System

  • src/player.rs: Audio playback using rodio
  • src/queue.rs: Speech queue management
  • src/speaker.rs: Voice/speaker configuration

Configuration (src/variables/)

  • rawvariables.rs: Persistent settings storage
  • Saves to vars.yaml in plugin directory
  • Manages per-ghost voice assignments and engine settings

Key Design Patterns

Async Architecture

  • Heavy use of tokio for async HTTP requests to TTS engines
  • async-trait for engine abstraction
  • Queue system for managing concurrent speech requests

Plugin Integration

  • SHIORI protocol implementation for 伺か communication
  • Windows-specific memory management via shiori_hglobal
  • Japanese text encoding support (UTF-8/ANSI fallback)

Engine Abstraction

  • Trait-based design allows easy addition of new TTS engines
  • Each engine implements speaker discovery and speech synthesis
  • Consistent API despite different underlying protocols

Development Notes

Windows-Specific Considerations

  • Built for 32-bit Windows compatibility
  • Uses Windows API for process management (CreateProcessW)
  • Character encoding handling for Japanese text is critical

TTS Engine Integration

  • Most engines use REST APIs on localhost ports
  • 棒読みちゃん uses a different protocol
  • Engines must be running before plugin can connect
  • Each engine has unique speaker/voice identification schemes

Configuration Files

  • vars.yaml: User settings and voice assignments
  • descript.txt: Plugin metadata for 伺か
  • rustfmt.toml: Code formatting (2 spaces, no hard tabs)

Testing and CI

  • GitHub Actions builds on Windows
  • Auto-updates MD5 checksums for distribution
  • Builds target i686-pc-windows-msvc specifically