diff --git a/README.md b/README.md index b61a449cb..2bc2bf7c1 100644 --- a/README.md +++ b/README.md @@ -180,7 +180,7 @@ Every item below is one more layer sedimented into a default — a capability yo - 🎨 **Style Pack Marketplace** — browse, install, and like community **style packs** from the in-app Marketplace, and publish your own (custom system prompt per pack, switchable by hotkey). Backed by a moderated marketplace backend; uploads are reviewed before they go public. - ⚡ **Streaming insertion** — polished text is written to the cursor character by character to reduce perceived latency, with an automatic one-shot-paste fallback. Toggle in Settings → Recording. - **Toggle and push-to-talk** recording modes, plus a **MediaPlayPause trigger** so wired-earbud inline controls can start and stop recording. `Esc` cancels at any phase, including polish and insertion. -- **Cloud ASR**: Volcengine streaming ASR (bigasr), iFlytek realtime ASR (RTASR), Alibaba Cloud Bailian (classic realtime / Qwen3 realtime / Fun-ASR-Flash file transcription), StepFun StepAudio (batch + realtime), Zhipu GLM-ASR, Xiaomi MiMo ASR, ElevenLabs Scribe, OpenAI-compatible batch transcription (OpenAI Whisper / Groq / SiliconFlow SenseVoice / OpenRouter), and Apple Speech (macOS). +- **Cloud ASR**: Volcengine streaming ASR (bigasr), iFlytek realtime ASR (RTASR), Alibaba Cloud Bailian (classic realtime / Qwen3 realtime / Fun-ASR-Flash file transcription), StepFun StepAudio (batch + realtime), Zhipu GLM-ASR, Xiaomi MiMo ASR, ElevenLabs Scribe, OpenAI-compatible batch transcription (OpenAI Whisper / Groq / SiliconFlow SenseVoice / OpenRouter / ZenMux), and Apple Speech (macOS). - **Local ASR**: bundled Qwen3-ASR (0.6B / 1.7B) via vendored `Open-Less/qwen-asr` (macOS); Windows Foundry Local Whisper and sherpa-onnx (experimental) variants. - **Polish providers**: Ark (Volcengine), DeepSeek, OpenAI, Google Gemini, Codex OAuth, SiliconFlow, Atlas Cloud, Xiaomi MiMo, CometAPI, OpenRouter, Alibaba Cloud Coding Plan, CodingPlanX, MiniMax, and StepFun — plus any OpenAI-compatible endpoint you bring. - **Four output modes**: raw, light polish, structured (**AI-prompt mode**), and formal. Plus a **translation hotkey** that converts speech directly into the configured target language ([#43](../../issues/43)). @@ -338,7 +338,7 @@ Long-term reference rewrites are stored as `raw → polished → rule` triples a The dictionary handles your proper nouns, product names, names of people, and new words. Today it supports: - Manually adding the correct spelling, a category, and notes. You do not need to maintain misspellings or context hints. -- Enabled entries are sent to the ASR provider that supports hotwords (Volcengine `context.hotwords`, StepFun `hotwords`, Whisper-compatible `prompt`, Bailian vocabulary ID) so they are recognized correctly during transcription. iFlytek realtime ASR has no request-level hotword parameter — configure personalized hotwords in the iFlytek console instead. +- Enabled entries are sent to the ASR provider that supports hotwords (Volcengine `context.hotwords`, StepFun `hotwords`, Whisper-compatible `prompt` — except ZenMux, whose JSON protocol does not carry `prompt`/`hotwords`, Bailian vocabulary ID) so they are recognized correctly during transcription. iFlytek realtime ASR has no request-level hotword parameter — configure personalized hotwords in the iFlytek console instead. - Entries are also injected into the polish prompt: the model decides per sentence whether to substitute. If "Cloud" clearly refers to the AI product `Claude` in context, it is corrected; if it genuinely means cloud computing, it is left as is. - The app auto-learns candidate corrections such as `Claude`, `ChatGPT`, and `OpenLess` from your history and offers them later. diff --git a/README.zh.md b/README.zh.md index 034b20520..72ace04f2 100644 --- a/README.zh.md +++ b/README.zh.md @@ -180,7 +180,7 @@ OpenLess 只做一件事:**把语音变成可用的书面文字(尤其是 AI 提 - 🎨 **风格包市场**——在应用内的 Marketplace 浏览、安装、点赞社区**风格包**,并发布自己的(每个包一套自定义系统提示词,可用快捷键切换)。由经过审核的市场后端支撑;上传内容公开前会经过审核。 - ⚡ **流式插入**——润色后的文本逐字符写入光标以降低感知延迟,并带有自动的一次性粘贴回退。可在 设置 → 录音 中切换。 - **切换式与按住说话(push-to-talk)** 两种录音模式,外加 **MediaPlayPause 触发**,让有线耳机的线控也能开始 / 停止录音。`Esc` 可在任意阶段取消,包括润色与插入。 -- **云端 ASR**:Volcengine 流式 ASR(bigasr)、讯飞实时语音转写(RTASR)、阿里云百炼(经典实时 / Qwen3 实时 / Fun-ASR-Flash 录音文件)、阶跃星辰 StepAudio(批式 + 实时)、智谱 GLM-ASR、小米 MiMo ASR、ElevenLabs Scribe、OpenAI 兼容批量转写(OpenAI Whisper / Groq / 硅基流动 SenseVoice / OpenRouter),以及 Apple Speech(macOS)。 +- **云端 ASR**:Volcengine 流式 ASR(bigasr)、讯飞实时语音转写(RTASR)、阿里云百炼(经典实时 / Qwen3 实时 / Fun-ASR-Flash 录音文件)、阶跃星辰 StepAudio(批式 + 实时)、智谱 GLM-ASR、小米 MiMo ASR、ElevenLabs Scribe、OpenAI 兼容批量转写(OpenAI Whisper / Groq / 硅基流动 SenseVoice / OpenRouter / ZenMux),以及 Apple Speech(macOS)。 - **本地 ASR**:通过 vendored 的 `Open-Less/qwen-asr` 内置 Qwen3-ASR(0.6B / 1.7B)(macOS);Windows 上的 Foundry Local Whisper 与 sherpa-onnx(实验性)变体。 - **润色提供方**:Ark(火山方舟)、DeepSeek、OpenAI、Google Gemini、Codex OAuth、硅基流动、Atlas Cloud、小米 MiMo、CometAPI、OpenRouter、阿里云 Coding Plan、CodingPlanX、MiniMax、StepFun,以及你自带的任意 OpenAI 兼容端点。 - **四种输出模式**:原文、轻度润色、结构化(**AI 提示词模式**)、正式。另有一个**翻译快捷键**,将语音直接转换为所配置的目标语言([#43](../../issues/43))。 @@ -192,7 +192,7 @@ OpenLess 只做一件事:**把语音变成可用的书面文字(尤其是 AI 提 - **Beta 频道(可选加入)**——设置 → 关于 → 加入 Beta 频道,可下载最新预发布版本进行手动安装。Beta 版本绝不会自动推送给 Stable 用户(见[贡献流程](#贡献流程))。 - **分发渠道**——从 [Releases](../../releases) 直接下载 DMG/EXE、Homebrew Cask(`brew install --cask openless`)、Windows 安装包。 - **单实例锁**——防止两个 OpenLess 进程争抢同一个快捷键边沿。 -- 词典条目注入到支持热词的 ASR 提供方(Volcengine 的 `context.hotwords`、StepFun 的 `hotwords`、Whisper 兼容的 `prompt`、百炼的 vocabulary_id),并在润色时作为语义提示;命中次数按会话累计。讯飞实时语音转写标准版没有请求级热词参数,需在讯飞控制台配置个性化热词。 +- 词典条目注入到支持热词的 ASR 提供方(Volcengine 的 `context.hotwords`、StepFun 的 `hotwords`、Whisper 兼容的 `prompt`(ZenMux 除外——其 JSON 协议不携带 `prompt`/`hotwords`)、百炼的 vocabulary_id),并在润色时作为语义提示;命中次数按会话累计。讯飞实时语音转写标准版没有请求级热词参数,需在讯飞控制台配置个性化热词。 - 平台原生全局快捷键:macOS 上为 CGEventTap,Windows 上为低级键盘钩子(`WH_KEYBOARD_LL`)。 ## 下载与安装(终端用户) diff --git a/openless-all/app/src-tauri/src/asr/whisper.rs b/openless-all/app/src-tauri/src/asr/whisper.rs index a29380eb3..de74dc7a0 100644 --- a/openless-all/app/src-tauri/src/asr/whisper.rs +++ b/openless-all/app/src-tauri/src/asr/whisper.rs @@ -22,17 +22,28 @@ pub const PROMPT_CHAR_BUDGET: usize = 240; /// 区切り文字（ASCII）。Whisper のトークナイザはどの言語でも安定して扱える。 const PROMPT_SEPARATOR: &str = ", "; +/// ZenMux 聚合平台的默认端点与模型（issue #837）。与前端 `ASR_PRESETS` 的 +/// `zenmux` 条目保持一致；`read_whisper_credentials` 在 active 为 zenmux 且 +/// 用户未填时回退到这里。 +pub const ZENMUX_DEFAULT_ENDPOINT: &str = "https://zenmux.ai/api/v1"; +pub const ZENMUX_DEFAULT_MODEL: &str = "qwen/qwen3-asr-flash"; + /// `/audio/transcriptions` 请求体编码方式。 /// /// OpenAI 官方及多数兼容厂商用 `multipart/form-data`（file + model）。 /// OpenRouter 虽路径相同、也走 Bearer，但请求体是 `application/json`： /// `{model, input_audio:{data:, format:"wav"}}`（issue #582）。 +/// ZenMux 与 OpenRouter 同形（issue #837），另支持可选的 `language` 与 +/// `enable_itn`（数字归一化）字段，故单独一个变体，避免给 OpenRouter +/// 的请求体塞未知字段（部分实现会对未知字段 4xx）。 #[derive(Clone, Copy, PartialEq, Eq, Debug)] pub enum AsrRequestFormat { /// `multipart/form-data`（既有行为，默认）。 Multipart, /// OpenRouter `application/json` + base64 音频。 OpenRouterJson, + /// ZenMux `application/json` + base64 音频（`language` 可选、`enable_itn` 恒发）。 + ZenMuxJson, } pub struct WhisperBatchASR { @@ -52,6 +63,12 @@ pub struct WhisperBatchASR { verbose_json: bool, /// 请求体编码方式。默认 `Multipart`，OpenRouter 走 `OpenRouterJson`。 request_format: AsrRequestFormat, + /// ZenMux 的 `language` 字段（ISO 639-1，如 `zh`）。None = 不发送，服务端 + /// 自动检测。仅 `ZenMuxJson` 编码下生效。 + language: Option, + /// ZenMux 的 `enable_itn` 字段（数字/单位归一化）。默认 true，与 ZenMux + /// 文档示例及中文 ASR 预期一致；仅 `ZenMuxJson` 编码下生效。 + enable_itn: bool, /// 一等 `hotwords` 参数（JSON 数组字符串）。StepFun 等厂商不认 `prompt` /// （静默忽略），但提供专门的热词字段——用它词典才真正生效。空 = 不发。 hotwords: Vec, @@ -75,6 +92,8 @@ impl WhisperBatchASR { max_chunk_duration_ms, verbose_json, request_format: AsrRequestFormat::Multipart, + language: None, + enable_itn: true, hotwords: Vec::new(), buffer: Mutex::new(Vec::new()), } @@ -94,6 +113,20 @@ impl WhisperBatchASR { self } + /// 设置 ZenMux 的 `language` 字段（ISO 639-1 码）。None = 不发送（自动检测）。 + /// 仅 `ZenMuxJson` 编码下生效。 + pub fn with_language(mut self, language: Option) -> Self { + self.language = language; + self + } + + /// 设置 ZenMux 的 `enable_itn`（数字归一化）开关，默认 true。仅 `ZenMuxJson` + /// 编码下生效。 + pub fn with_enable_itn(mut self, enable_itn: bool) -> Self { + self.enable_itn = enable_itn; + self + } + /// Stop collecting audio, encode the buffer as WAV, and POST to the /// Whisper transcriptions endpoint. /// @@ -218,6 +251,29 @@ impl WhisperBatchASR { } request.json(&body) } + AsrRequestFormat::ZenMuxJson => { + // ZenMux /audio/transcriptions（issue #837）：application/json， + // 音频走标准 base64。语言跟随 OpenLess 工作语言映射；enable_itn + // 由设置页开关决定，恒显式发送。不带 multipart 专属字段。 + let mut body = serde_json::json!({ + "model": self.model, + "input_audio": { + "data": base64::engine::general_purpose::STANDARD.encode(&wav), + "format": "wav", + }, + "enable_itn": self.enable_itn, + }); + if let Some(language) = self.language.as_ref() { + if !language.trim().is_empty() { + body["language"] = serde_json::Value::String(language.trim().to_string()); + } + } + let mut request = client.post(&url); + if !self.api_key.trim().is_empty() { + request = request.header("Authorization", format!("Bearer {}", self.api_key)); + } + request.json(&body) + } }; let resp = request @@ -858,6 +914,118 @@ mod tests { server.join().unwrap(); } + #[tokio::test] + async fn zenmux_format_posts_json_with_language_and_itn() { + // issue #837：ZenMuxJson 走 application/json + input_audio.data(base64)， + // 语言有映射时发送、enable_itn 恒显式发送；响应仍按 {text} 解析。 + let listener = TcpListener::bind("127.0.0.1:0").unwrap(); + listener.set_nonblocking(true).unwrap(); + let addr = listener.local_addr().unwrap(); + let server = thread::spawn(move || { + let deadline = Instant::now() + Duration::from_secs(5); + let mut stream = loop { + match listener.accept() { + Ok((stream, _)) => break stream, + Err(err) if err.kind() == std::io::ErrorKind::WouldBlock => { + assert!( + Instant::now() < deadline, + "timed out waiting for ASR test request" + ); + thread::sleep(Duration::from_millis(10)); + } + Err(err) => panic!("accept ASR test request failed: {err}"), + } + }; + stream.set_nonblocking(false).unwrap(); + stream + .set_read_timeout(Some(Duration::from_secs(5))) + .unwrap(); + let request = read_http_request(&mut stream); + let request_text = String::from_utf8_lossy(&request); + let lower = request_text.to_ascii_lowercase(); + assert!(request_text.starts_with("POST /audio/transcriptions HTTP/1.1")); + assert!(lower.contains("content-type: application/json")); + assert!(lower.contains("authorization: bearer key")); + assert!(request_text.contains("input_audio")); + assert!(request_text.contains(r#""format":"wav""#)); + assert!(request_text.contains(r#""language":"zh""#)); + assert!(request_text.contains(r#""enable_itn":true"#)); + assert!(!lower.contains("multipart/form-data")); + write_json_response(&mut stream, r#"{"text":"zenmux ok"}"#); + }); + let base_url = format!("http://{}", addr); + + let asr = WhisperBatchASR::new( + "key".to_string(), + base_url, + "qwen/qwen3-asr-flash".to_string(), + None, + Some(30_000), + false, + ) + .with_request_format(AsrRequestFormat::ZenMuxJson) + .with_language(Some("zh".to_string())) + .with_enable_itn(true); + let pcm = vec![0u8; 32_000 * 2]; + asr.consume_pcm_chunk(&pcm); + + let transcript = asr.transcribe().await.unwrap(); + assert_eq!(transcript.text, "zenmux ok"); + server.join().unwrap(); + } + + #[tokio::test] + async fn zenmux_format_omits_language_when_unset_and_sends_false_itn() { + // language 无映射（None）时不发送该字段；enable_itn=false 显式发送 false。 + let listener = TcpListener::bind("127.0.0.1:0").unwrap(); + listener.set_nonblocking(true).unwrap(); + let addr = listener.local_addr().unwrap(); + let server = thread::spawn(move || { + let deadline = Instant::now() + Duration::from_secs(5); + let mut stream = loop { + match listener.accept() { + Ok((stream, _)) => break stream, + Err(err) if err.kind() == std::io::ErrorKind::WouldBlock => { + assert!( + Instant::now() < deadline, + "timed out waiting for ASR test request" + ); + thread::sleep(Duration::from_millis(10)); + } + Err(err) => panic!("accept ASR test request failed: {err}"), + } + }; + stream.set_nonblocking(false).unwrap(); + stream + .set_read_timeout(Some(Duration::from_secs(5))) + .unwrap(); + let request = read_http_request(&mut stream); + let request_text = String::from_utf8_lossy(&request); + assert!(request_text.contains(r#""enable_itn":false"#)); + assert!(!request_text.contains(r#""language""#)); + write_json_response(&mut stream, r#"{"text":"zenmux no-lang ok"}"#); + }); + let base_url = format!("http://{}", addr); + + let asr = WhisperBatchASR::new( + "key".to_string(), + base_url, + "qwen/qwen3-asr-flash".to_string(), + None, + None, + false, + ) + .with_request_format(AsrRequestFormat::ZenMuxJson) + .with_language(None) + .with_enable_itn(false); + let pcm = vec![0u8; 32_000 * 2]; + asr.consume_pcm_chunk(&pcm); + + let transcript = asr.transcribe().await.unwrap(); + assert_eq!(transcript.text, "zenmux no-lang ok"); + server.join().unwrap(); + } + #[tokio::test] async fn empty_api_key_omits_authorization_header() { // openai-compatible 允许 API Key 留空（LAN 无鉴权端点）：请求不得携带 diff --git a/openless-all/app/src-tauri/src/commands/providers.rs b/openless-all/app/src-tauri/src/commands/providers.rs index e0a652f4a..537c304a9 100644 --- a/openless-all/app/src-tauri/src/commands/providers.rs +++ b/openless-all/app/src-tauri/src/commands/providers.rs @@ -1,4 +1,5 @@ use super::*; +use base64::Engine; use std::collections::HashMap; #[derive(Serialize)] @@ -304,7 +305,10 @@ async fn validate_asr_provider() -> Result<(), String> { .map_err(|e| e.to_string())? .filter(|s| !s.trim().is_empty()) .ok_or_else(|| "asrModelMissing".to_string())?; - validate_asr_transcription(&config, model.trim()).await + // 验证请求体与真实转写保持一致：OpenRouter / ZenMux 走 JSON+base64， + // 其余 whisper 兼容厂商走 multipart——避免「测试连接」假阴性（issue #837）。 + let request_format = crate::coordinator::whisper_request_format(&active_asr); + validate_asr_transcription(&config, model.trim(), request_format).await } /// 讯飞 RTASR 验证：真连 + 500ms 静音 + 收尾。鉴权错误（10105 / 10110）在握手阶段 @@ -676,7 +680,11 @@ pub(crate) fn active_sherpa_asr_is_supported(provider: &str) -> bool { } } -async fn validate_asr_transcription(config: &ProviderConfig, model: &str) -> Result<(), String> { +async fn validate_asr_transcription( + config: &ProviderConfig, + model: &str, + request_format: crate::asr::whisper::AsrRequestFormat, +) -> Result<(), String> { const MAX_ASR_VALIDATE_BODY_BYTES: usize = 1024 * 1024; const MAX_ATTEMPTS: u32 = 6; let url = asr_transcriptions_url(&config.base_url)?; @@ -689,20 +697,57 @@ async fn validate_asr_transcription(config: &ProviderConfig, model: &str) -> Res let mut attempt: u32 = 0; let response = loop { attempt += 1; - let wav_part = reqwest::multipart::Part::bytes(wav.clone()) - .file_name("openless-asr-check.wav") - .mime_str("audio/wav") - .map_err(|e| format!("请求体构建失败: {e}"))?; - let form = reqwest::multipart::Form::new() - .part("file", wav_part) - .text("model", model.to_string()); - // `openai-compatible` 允许 API Key 留空：此时不带 Authorization 头， - // 避免空 Bearer 被服务端 401 拒绝（与 fetch_provider_models 一致）。 - let mut request = client.post(&url); - if !config.api_key.trim().is_empty() { - request = request.header("Authorization", format!("Bearer {}", config.api_key)); - } - match request.multipart(form).send().await { + let request = match request_format { + crate::asr::whisper::AsrRequestFormat::Multipart => { + let wav_part = reqwest::multipart::Part::bytes(wav.clone()) + .file_name("openless-asr-check.wav") + .mime_str("audio/wav") + .map_err(|e| format!("请求体构建失败: {e}"))?; + let form = reqwest::multipart::Form::new() + .part("file", wav_part) + .text("model", model.to_string()); + let mut request = client.post(&url); + if !config.api_key.trim().is_empty() { + request = request.header("Authorization", format!("Bearer {}", config.api_key)); + } + request.multipart(form) + } + crate::asr::whisper::AsrRequestFormat::OpenRouterJson => { + // OpenRouter：application/json + base64（issue #582），与真实 + // 转写请求同形；不带 multipart 专属字段。 + let body = serde_json::json!({ + "model": model, + "input_audio": { + "data": base64::engine::general_purpose::STANDARD.encode(&wav), + "format": "wav", + }, + }); + let mut request = client.post(&url); + if !config.api_key.trim().is_empty() { + request = request.header("Authorization", format!("Bearer {}", config.api_key)); + } + request.json(&body) + } + crate::asr::whisper::AsrRequestFormat::ZenMuxJson => { + // ZenMux：application/json + base64，enable_itn 与真实请求默认 + // 一致（true），language 留空走服务端自动检测（issue #837）。 + let body = serde_json::json!({ + "model": model, + "input_audio": { + "data": base64::engine::general_purpose::STANDARD.encode(&wav), + "format": "wav", + }, + "enable_itn": true, + }); + let mut request = client.post(&url); + if !config.api_key.trim().is_empty() { + request = request.header("Authorization", format!("Bearer {}", config.api_key)); + } + request.json(&body) + } + }; + match request.send().await + { Ok(resp) => break resp, Err(e) if e.is_timeout() => return Err("providerRequestTimeout".to_string()), Err(e) if (e.is_connect() || e.is_request()) && attempt < MAX_ATTEMPTS => { diff --git a/openless-all/app/src-tauri/src/coordinator.rs b/openless-all/app/src-tauri/src/coordinator.rs index aa9ce9f13..6915be840 100644 --- a/openless-all/app/src-tauri/src/coordinator.rs +++ b/openless-all/app/src-tauri/src/coordinator.rs @@ -77,6 +77,9 @@ mod silence_auto_stop; pub(crate) mod selection_polish; use asr_wiring::*; +// providers.rs 的 ASR 验证路径按 provider 的真实请求格式发送探针（issue #837）， +// 需要跨模块访问 whisper 兼容系的格式映射，显式再导出。 +pub(crate) use asr_wiring::whisper_request_format; use capsule_focus::*; use hotkey_loops::*; use polish_flow::*; @@ -444,9 +447,9 @@ pub(crate) fn derive_bailian_endpoint( fn batch_asr_chunk_limit_ms(provider_id: &str) -> Option { match provider_id { - // OpenRouter 把音频 base64 进 JSON body，体积比二进制大 ~33%，长录音易撞 - // body/时长上限，保守按 30s 切分（与 zhipu 同）。 - "zhipu" | "openrouter" => Some(30_000), + // OpenRouter / ZenMux 把音频 base64 进 JSON body，体积比二进制大 ~33%， + // 长录音易撞 body/时长上限，保守按 30s 切分（与 zhipu 同）。 + "zhipu" | "openrouter" | "zenmux" => Some(30_000), // 其余预设默认不分片；openai-compatible 可由用户高级配置覆盖。 _ => read_advanced_asr_config(provider_id).chunk_duration_ms, } @@ -457,13 +460,30 @@ fn batch_asr_chunk_limit_ms(provider_id: &str) -> Option { /// 与分片时长由用户按 provider 配置（存凭据 vault）。 pub(crate) const OPENAI_COMPATIBLE_ASR_PROVIDER_ID: &str = "openai-compatible"; -/// `openai-compatible` 预设的高级配置（per-provider 存于凭据 vault）。 -#[derive(Debug, Clone, Copy, PartialEq, Eq, Default)] +/// ZenMux ASR 预设 id（issue #837）。与前端 `ASR_PRESETS` 的 `zenmux` 条目一致， +/// 复用 Whisper 批式管线，但请求体走 JSON + base64（`ZenMuxJson`）。 +pub(crate) const ZENMUX_ASR_PROVIDER_ID: &str = "zenmux"; + +/// `openai-compatible` 与 `zenmux` 预设的高级配置（per-provider 存于凭据 vault）。 +#[derive(Debug, Clone, Copy, PartialEq, Eq)] pub(crate) struct AdvancedAsrConfig { /// 是否请求 `response_format=verbose_json`（配合幻听过滤；服务端不支持时保持 false）。 pub(crate) verbose_json: bool, /// 单次请求的音频分片时长；None = 不分片整段发送。 pub(crate) chunk_duration_ms: Option, + /// ZenMux `enable_itn`（数字/单位归一化）。默认 true，与 ZenMux 文档示例及 + /// 中文 ASR 预期一致；仅 zenmux 消费。 + pub(crate) enable_itn: bool, +} + +impl Default for AdvancedAsrConfig { + fn default() -> Self { + Self { + verbose_json: false, + chunk_duration_ms: None, + enable_itn: true, + } + } } /// 解析 per-provider 高级配置 JSON；缺失/非法一律回落保守默认。 @@ -492,20 +512,25 @@ fn parse_advanced_asr_config(raw: Option<&str>) -> AdvancedAsrConfig { .map(|ms| ms.floor() as u64) }) }), + // 缺失/非布尔回落默认 true（开启），与前端 advancedAsrConfig.ts 一致。 + enable_itn: value + .get("enableItn") + .and_then(|v| v.as_bool()) + .unwrap_or(true), } } -/// 按 provider id 决定高级配置：仅 `openai-compatible` 采用用户配置， -/// 命名厂商一律返回默认（保持硬编码行为）。 +/// 按 provider id 决定高级配置：`openai-compatible` 与 `zenmux` 采用用户配置， +/// 其余命名厂商一律返回默认（保持硬编码行为）。 fn advanced_asr_config_for(provider_id: &str, raw: Option<&str>) -> AdvancedAsrConfig { - if provider_id != OPENAI_COMPATIBLE_ASR_PROVIDER_ID { + if provider_id != OPENAI_COMPATIBLE_ASR_PROVIDER_ID && provider_id != ZENMUX_ASR_PROVIDER_ID { return AdvancedAsrConfig::default(); } parse_advanced_asr_config(raw) } -/// 读取某 ASR provider 的高级配置。仅 `openai-compatible` 读 vault；命名厂商 -/// 走硬编码行为（这里返回默认值），避免破坏已测通的路径。 +/// 读取某 ASR provider 的高级配置。仅 `openai-compatible` / `zenmux` 读 vault； +/// 其余命名厂商走硬编码行为（这里返回默认值），避免破坏已测通的路径。 fn read_advanced_asr_config(provider_id: &str) -> AdvancedAsrConfig { let raw = CredentialsVault::get_for_asr_provider( provider_id, @@ -2547,18 +2572,35 @@ fn read_whisper_credentials() -> (String, String, String) { .ok() .flatten() .unwrap_or_default(); + let active_asr = CredentialsVault::get_active_asr(); + let (default_endpoint, default_model) = whisper_credential_defaults(&active_asr); let base_url = CredentialsVault::get(CredentialAccount::AsrEndpoint) .ok() .flatten() - .unwrap_or_default(); + .filter(|s| !s.trim().is_empty()) + .unwrap_or(default_endpoint); let model = CredentialsVault::get(CredentialAccount::AsrModel) .ok() .flatten() - .filter(|s| !s.is_empty()) - .unwrap_or_else(|| "whisper-1".to_string()); + .filter(|s| !s.trim().is_empty()) + .unwrap_or(default_model); (api_key, base_url, model) } +/// whisper 兼容系 provider 的「空槽默认值」。zenmux 有厂商默认端点/模型 +/// （与前端 preset 一致）；其余预设沿用空 endpoint + `whisper-1`（默认值由 +/// 前端切换 provider 时写入 vault）。纯函数，便于单测。 +fn whisper_credential_defaults(provider_id: &str) -> (String, String) { + if provider_id == ZENMUX_ASR_PROVIDER_ID { + ( + crate::asr::whisper::ZENMUX_DEFAULT_ENDPOINT.to_string(), + crate::asr::whisper::ZENMUX_DEFAULT_MODEL.to_string(), + ) + } else { + (String::new(), "whisper-1".to_string()) + } +} + fn read_mimo_credentials() -> (String, String, String) { let api_key = CredentialsVault::get(CredentialAccount::AsrApiKey) .ok() @@ -3232,6 +3274,7 @@ mod tests { AdvancedAsrConfig { verbose_json: true, chunk_duration_ms: Some(30_000), + enable_itn: true, } ); // 命名厂商忽略该配置，保持硬编码行为。 @@ -3266,6 +3309,7 @@ mod tests { AdvancedAsrConfig { verbose_json: true, chunk_duration_ms: None, + enable_itn: true, } ); // 分片时长 0 或缺失 = 不分片。 @@ -3280,6 +3324,7 @@ mod tests { AdvancedAsrConfig { verbose_json: false, chunk_duration_ms: Some(30_000), + enable_itn: true, } ); // 浮点分片时长与前端一致向下取整：30000.9 → 30000。 @@ -3288,6 +3333,7 @@ mod tests { AdvancedAsrConfig { verbose_json: false, chunk_duration_ms: Some(30_000), + enable_itn: true, } ); // 负数 / 字符串分片时长 → 不分片。 @@ -3322,6 +3368,38 @@ mod tests { assert_eq!(batch_asr_chunk_limit_ms("openrouter"), Some(30_000)); } + #[test] + fn zenmux_credential_defaults_and_advanced_config() { + use crate::asr::whisper::{ZENMUX_DEFAULT_ENDPOINT, ZENMUX_DEFAULT_MODEL}; + // base64 进 JSON body 体积膨胀，与 OpenRouter 同按 30s 切分。 + assert_eq!(batch_asr_chunk_limit_ms("zenmux"), Some(30_000)); + // 空槽默认值：zenmux 回落厂商默认端点/模型（与前端 preset 一致）； + // 其余 whisper 兼容预设沿用空 endpoint + whisper-1。 + assert_eq!( + whisper_credential_defaults("zenmux"), + ( + ZENMUX_DEFAULT_ENDPOINT.to_string(), + ZENMUX_DEFAULT_MODEL.to_string() + ) + ); + assert_eq!( + whisper_credential_defaults("whisper"), + (String::new(), "whisper-1".to_string()) + ); + assert_eq!( + whisper_credential_defaults("openrouter"), + (String::new(), "whisper-1".to_string()) + ); + + // enable_itn 默认 true；用户配置显式 false 可覆盖；仅 openai-compatible / + // zenmux 读用户配置，其余命名厂商忽略（保持硬编码行为）。 + assert!(advanced_asr_config_for("zenmux", None).enable_itn); + assert!(advanced_asr_config_for("zenmux", Some(r#"{"verboseJson":true}"#)).enable_itn); + assert!(!advanced_asr_config_for("zenmux", Some(r#"{"enableItn":false}"#)).enable_itn); + assert!(advanced_asr_config_for("whisper", Some(r#"{"enableItn":false}"#)).enable_itn); + assert!(!advanced_asr_config_for("zenmux", Some(r#"{"enableItn":false}"#)).verbose_json); + } + #[test] fn stepfun_is_whisper_compatible_with_hotwords_vocab() { use crate::asr::whisper::AsrRequestFormat; diff --git a/openless-all/app/src-tauri/src/coordinator/asr_wiring.rs b/openless-all/app/src-tauri/src/coordinator/asr_wiring.rs index 342f58945..55dda7922 100644 --- a/openless-all/app/src-tauri/src/coordinator/asr_wiring.rs +++ b/openless-all/app/src-tauri/src/coordinator/asr_wiring.rs @@ -385,7 +385,7 @@ pub(super) fn build_apple_speech( pub(super) fn is_whisper_compatible_provider(id: &str) -> bool { matches!( id, - "whisper" | "siliconflow" | "zhipu" | "groq" | "openrouter" | "stepfun" + "whisper" | "siliconflow" | "zhipu" | "groq" | "openrouter" | "stepfun" | "zenmux" ) || id == OPENAI_COMPATIBLE_ASR_PROVIDER_ID } @@ -416,9 +416,11 @@ pub(super) fn whisper_vocab_for_provider( /// 该 provider 的请求体编码方式。OpenRouter 的 `/audio/transcriptions` 是 /// `application/json` + base64 音频（issue #582），其余兼容厂商沿用 multipart。 -pub(super) fn whisper_request_format(provider_id: &str) -> crate::asr::whisper::AsrRequestFormat { +/// ZenMux 同形但带 `language` / `enable_itn`（issue #837），单独走 `ZenMuxJson`。 +pub(crate) fn whisper_request_format(provider_id: &str) -> crate::asr::whisper::AsrRequestFormat { match provider_id { "openrouter" => crate::asr::whisper::AsrRequestFormat::OpenRouterJson, + "zenmux" => crate::asr::whisper::AsrRequestFormat::ZenMuxJson, _ => crate::asr::whisper::AsrRequestFormat::Multipart, } } @@ -437,11 +439,61 @@ pub(super) fn whisper_request_format(provider_id: &str) -> crate::asr::whisper:: pub(super) fn whisper_supports_verbose_json(provider_id: &str) -> bool { match provider_id { "whisper" | "groq" => true, + // ZenMux 的 JSON 请求体协议没有 response_format，恒关闭。 + "zenmux" => false, // openai-compatible 由用户高级配置决定；其余厂商保持关闭。 _ => read_advanced_asr_config(provider_id).verbose_json, } } +/// OpenLess 工作语言（原生名，见前端 `SUPPORTED_LANGUAGES`）→ ZenMux `language` +/// 字段值（ISO 639-1 码）。取 `working_languages` 主语言映射；未收录的语言返回 +/// None —— 请求体省略 `language`，由 ZenMux 服务端自动检测（issue #837）。 +pub(super) fn zenmux_language_code(native_name: &str) -> Option { + let code = match native_name.trim() { + "简体中文" | "繁体中文" => "zh", + "English" => "en", + "日本語" => "ja", + "한국어" => "ko", + "Français" => "fr", + "Deutsch" => "de", + "Español" => "es", + "Italiano" => "it", + "Português" => "pt", + "Русский" => "ru", + "العربية" => "ar", + "Tiếng Việt" => "vi", + "ไทย" => "th", + "हिन्दी" => "hi", + _ => return None, + }; + Some(code.to_string()) +} + +/// 当前 prefs 的主工作语言 → ZenMux `language`（None = 不发送，自动检测）。 +pub(super) fn zenmux_language_for_prefs(prefs: &crate::types::UserPreferences) -> Option { + prefs + .working_languages + .first() + .and_then(|name| zenmux_language_code(name)) +} + +/// 构造完的 `WhisperBatchASR` 上注入 zenmux 专属选项（`language` 跟随工作语言、 +/// `enable_itn` 读用户高级配置）。非 zenmux 原样返回，保持现有行为；QA 与听写 +/// 两处构造点共用，避免重复逻辑。 +pub(super) fn apply_zenmux_asr_options( + builder: crate::asr::whisper::WhisperBatchASR, + active_asr: &str, + inner: &Arc, +) -> crate::asr::whisper::WhisperBatchASR { + if active_asr != ZENMUX_ASR_PROVIDER_ID { + return builder; + } + builder + .with_language(zenmux_language_for_prefs(&inner.prefs.get())) + .with_enable_itn(read_advanced_asr_config(ZENMUX_ASR_PROVIDER_ID).enable_itn) +} + pub(super) fn is_bailian_provider(id: &str) -> bool { id == crate::asr::bailian::PROVIDER_ID } @@ -750,7 +802,7 @@ pub(super) async fn build_qa_asr_start( let label = AsrCallLabel::new(effective_asr.clone(), Some(model.clone())); let (whisper_prompt, hotwords) = whisper_vocab_for_provider(active_asr, enabled_phrases(inner)); - let whisper = Arc::new( + let whisper = Arc::new(apply_zenmux_asr_options( WhisperBatchASR::new( api_key, base_url, @@ -761,7 +813,9 @@ pub(super) async fn build_qa_asr_start( ) .with_request_format(whisper_request_format(active_asr)) .with_hotwords(hotwords), - ); + active_asr, + inner, + )); let active = ActiveAsr::Whisper(Arc::clone(&whisper)); let consumer: Arc = whisper; Ok((QaAsrStart::Ready { active, consumer }, label)) @@ -798,6 +852,47 @@ pub(super) async fn build_qa_asr_start( mod tests { use super::*; + #[test] + fn zenmux_is_whisper_compatible_json_provider() { + use crate::asr::whisper::AsrRequestFormat; + // issue #837：ZenMux 走 whisper 兼容路由，请求体 JSON+base64（ZenMuxJson）， + // 与 OpenRouter 共用 30s 切分；JSON 协议不吃 response_format / hotwords。 + assert!(is_whisper_compatible_provider("zenmux")); + assert_eq!( + active_asr_provider_kind("zenmux"), + ActiveAsrProviderKind::WhisperCompatible + ); + assert_eq!( + whisper_request_format("zenmux"), + AsrRequestFormat::ZenMuxJson + ); + assert!(!whisper_supports_verbose_json("zenmux")); + assert!(!whisper_uses_hotwords("zenmux")); + } + + #[test] + fn zenmux_language_code_covers_supported_languages_and_omits_unknown() { + // 覆盖前端 SUPPORTED_LANGUAGES 的全部 15 种语言；未收录 → None（自动检测）。 + assert_eq!(zenmux_language_code("简体中文").as_deref(), Some("zh")); + assert_eq!(zenmux_language_code("繁体中文").as_deref(), Some("zh")); + assert_eq!(zenmux_language_code("English").as_deref(), Some("en")); + assert_eq!(zenmux_language_code("日本語").as_deref(), Some("ja")); + assert_eq!(zenmux_language_code("한국어").as_deref(), Some("ko")); + assert_eq!(zenmux_language_code("Français").as_deref(), Some("fr")); + assert_eq!(zenmux_language_code("Deutsch").as_deref(), Some("de")); + assert_eq!(zenmux_language_code("Español").as_deref(), Some("es")); + assert_eq!(zenmux_language_code("Italiano").as_deref(), Some("it")); + assert_eq!(zenmux_language_code("Português").as_deref(), Some("pt")); + assert_eq!(zenmux_language_code("Русский").as_deref(), Some("ru")); + assert_eq!(zenmux_language_code("العربية").as_deref(), Some("ar")); + assert_eq!(zenmux_language_code("Tiếng Việt").as_deref(), Some("vi")); + assert_eq!(zenmux_language_code("ไทย").as_deref(), Some("th")); + assert_eq!(zenmux_language_code("हिन्दी").as_deref(), Some("hi")); + assert_eq!(zenmux_language_code(""), None); + assert_eq!(zenmux_language_code("Esperanto"), None); + assert_eq!(zenmux_language_code(" "), None); + } + #[test] fn require_openai_compatible_fields_errors_on_missing_endpoint_or_model() { // endpoint 缺失（含纯空白）→ 明确报错，绝不静默回落 whisper-1。 diff --git a/openless-all/app/src-tauri/src/coordinator/dictation.rs b/openless-all/app/src-tauri/src/coordinator/dictation.rs index 0e607e056..672e5d834 100644 --- a/openless-all/app/src-tauri/src/coordinator/dictation.rs +++ b/openless-all/app/src-tauri/src/coordinator/dictation.rs @@ -2153,7 +2153,7 @@ pub(super) async fn begin_session_as( let (whisper_prompt, hotwords) = whisper_vocab_for_provider(&active_asr, enabled_phrases(inner)); let asr_call_label = AsrCallLabel::new(effective_asr.clone(), Some(model.clone())); - let whisper = Arc::new( + let whisper = Arc::new(apply_zenmux_asr_options( WhisperBatchASR::new( api_key, base_url, @@ -2164,7 +2164,9 @@ pub(super) async fn begin_session_as( ) .with_request_format(whisper_request_format(&active_asr)) .with_hotwords(hotwords), - ); + &active_asr, + inner, + )); store_asr_for_session( inner, current_session_id, diff --git a/openless-all/app/src/i18n/en.ts b/openless-all/app/src/i18n/en.ts index a21c84899..5149714c1 100644 --- a/openless-all/app/src/i18n/en.ts +++ b/openless-all/app/src/i18n/en.ts @@ -868,6 +868,7 @@ export const en: typeof zhCN = { asrGroq: 'Groq Whisper-large-v3', asrWhisper: 'OpenAI Whisper (compatible)', asrOpenrouter: 'OpenRouter Whisper', + asrZenmux: 'ZenMux', asrOpenAiCompatible: 'Custom OpenAI-compatible', asrXiaomiMimo: 'Xiaomi MiMo ASR', asrIflytek: 'iFlytek Realtime ASR', @@ -878,11 +879,14 @@ export const en: typeof zhCN = { asrAppleSpeech: 'Apple Speech (macOS)', }, elevenLabsUploadNotice: 'ElevenLabs uploads recorded audio to the configured endpoint for batch transcription.', - asrAdvancedNote: 'Advanced options below only affect the Custom OpenAI-compatible preset; named provider presets keep their built-in behavior.', + zenmuxVocabularyNote: 'ZenMux uses a JSON transcription protocol and does not receive dictionary hotwords (prompt/hotwords); the dictionary still feeds the polish step but does not bias speech recognition.', + asrAdvancedNote: 'Advanced options below only affect the Custom OpenAI-compatible and ZenMux presets; other named provider presets keep their built-in behavior.', asrAdvancedVerboseJsonLabel: 'Segment metrics (verbose_json)', asrAdvancedVerboseJsonHint: 'Requests segment metrics for hallucination filtering when the server supports it; keep off for self-hosted servers that do not.', asrAdvancedChunkLabel: 'Chunk duration (ms)', asrAdvancedChunkHint: '0 = no chunking, send the whole clip at once. Chunked requests suit long recordings or servers with per-request duration limits.', + asrAdvancedEnableItnLabel: 'Number normalization (enable_itn)', + asrAdvancedEnableItnHint: 'Normalizes spoken numbers/units into Arabic numerals (e.g. “twenty twenty-six” → “2026”). Turn off to keep the raw text.', volcengineAppKeyLabel: 'APP ID', volcengineAccessKeyLabel: 'Access Token', volcengineApiKeyLabel: 'API Key', diff --git a/openless-all/app/src/i18n/ja.ts b/openless-all/app/src/i18n/ja.ts index d990bdef2..c577a0da0 100644 --- a/openless-all/app/src/i18n/ja.ts +++ b/openless-all/app/src/i18n/ja.ts @@ -870,6 +870,7 @@ export const ja: typeof zhCN = { asrGroq: 'Groq Whisper-large-v3', asrWhisper: 'OpenAI Whisper（互換）', asrOpenrouter: 'OpenRouter Whisper', + asrZenmux: 'ZenMux', asrOpenAiCompatible: 'カスタム OpenAI 互換', asrXiaomiMimo: 'Xiaomi MiMo ASR', asrIflytek: 'iFlytek リアルタイム音声認識', @@ -880,11 +881,14 @@ export const ja: typeof zhCN = { asrAppleSpeech: 'Apple 音声認識 (macOS)', }, elevenLabsUploadNotice: 'ElevenLabs は録音音声を設定済みのエンドポイントへアップロードしてバッチ文字起こしします。', - asrAdvancedNote: '以下の詳細オプションは「カスタム OpenAI 互換」プリセットのみに影響します。名前付きプロバイダーのプリセットは内蔵動作のままです。', + zenmuxVocabularyNote: 'ZenMux は JSON 文字起こしプロトコルを使用し、辞書ホットワード（prompt/hotwords）は送信されません。辞書は依然として潤色段階には渡りますが、音声認識のバイアスには使用されません。', + asrAdvancedNote: '以下の詳細オプションは「カスタム OpenAI 互換」と「ZenMux」のプリセットのみに影響します。その他の名前付きプロバイダーのプリセットは内蔵動作のままです。', asrAdvancedVerboseJsonLabel: 'セグメント指標 (verbose_json)', asrAdvancedVerboseJsonHint: 'サーバーが対応する場合に幻聴フィルタ用の segments 指標を要求します。非対応の自前サーバーではオフのままにしてください。', asrAdvancedChunkLabel: '分割時間 (ms)', asrAdvancedChunkHint: '0 = 分割なし（全体を一括送信）。長い録音や1リクエストの時間制限があるサーバー向けに分割送信できます。', + asrAdvancedEnableItnLabel: '数字正規化 (enable_itn)', + asrAdvancedEnableItnHint: '口頭の数字・単位を算用数字に正規化します（例：「にせんにじゅうろく」→「2026」）。オフにすると元の表記を保持します。', volcengineAppKeyLabel: 'APP ID', volcengineAccessKeyLabel: 'Access Token', volcengineApiKeyLabel: 'API Key', diff --git a/openless-all/app/src/i18n/ko.ts b/openless-all/app/src/i18n/ko.ts index d761444e7..ad3139b68 100644 --- a/openless-all/app/src/i18n/ko.ts +++ b/openless-all/app/src/i18n/ko.ts @@ -870,6 +870,7 @@ export const ko: typeof zhCN = { asrGroq: 'Groq Whisper-large-v3', asrWhisper: 'OpenAI Whisper(호환)', asrOpenrouter: 'OpenRouter Whisper', + asrZenmux: 'ZenMux', asrOpenAiCompatible: '커스텀 OpenAI 호환', asrXiaomiMimo: 'Xiaomi MiMo ASR', asrIflytek: 'iFlytek 실시간 음성 인식', @@ -880,11 +881,14 @@ export const ko: typeof zhCN = { asrAppleSpeech: 'Apple 음성 (macOS)', }, elevenLabsUploadNotice: 'ElevenLabs는 녹음 오디오를 설정된 엔드포인트에 업로드해 일괄 전사합니다.', - asrAdvancedNote: '아래 고급 옵션은「커스텀 OpenAI 호환」프리셋에만 적용됩니다. 명명된 공급자 프리셋은 내장 동작을 유지합니다.', + zenmuxVocabularyNote: 'ZenMux는 JSON 전사 프로토콜을 사용하며 사전 핫워드(prompt/hotwords)를 보내지 않습니다. 사전은 여전히 다듬기 단계에 전달되지만 음성 인식 편향에는 사용되지 않습니다.', + asrAdvancedNote: '아래 고급 옵션은「커스텀 OpenAI 호환」및「ZenMux」프리셋에만 적용됩니다. 다른 명명된 공급자 프리셋은 내장 동작을 유지합니다.', asrAdvancedVerboseJsonLabel: '세그먼트 지표 (verbose_json)', asrAdvancedVerboseJsonHint: '서버가 지원할 때 환각 필터링용 segments 지표를 요청합니다. 지원하지 않는 자체 호스팅 서버에서는 꺼두세요.', asrAdvancedChunkLabel: '분할 시간 (ms)', asrAdvancedChunkHint: '0 = 분할 없음(전체를 한 번에 전송). 긴 녹음이나 요청당 시간 제한이 있는 서버에 적합합니다.', + asrAdvancedEnableItnLabel: '숫자 정규화 (enable_itn)', + asrAdvancedEnableItnHint: '구어 숫자/단위를 아라비아 숫자로 정규화합니다(예: "이천이십육" → "2026"). 끄면 원문을 유지합니다.', volcengineAppKeyLabel: 'APP ID', volcengineAccessKeyLabel: 'Access Token', volcengineApiKeyLabel: 'API Key', diff --git a/openless-all/app/src/i18n/zh-CN.ts b/openless-all/app/src/i18n/zh-CN.ts index b4515d57a..5086e86b8 100644 --- a/openless-all/app/src/i18n/zh-CN.ts +++ b/openless-all/app/src/i18n/zh-CN.ts @@ -866,6 +866,7 @@ export const zhCN = { asrGroq: 'Groq Whisper-large-v3', asrWhisper: 'OpenAI Whisper（兼容）', asrOpenrouter: 'OpenRouter Whisper', + asrZenmux: 'ZenMux', asrOpenAiCompatible: '自定义 OpenAI 兼容', asrXiaomiMimo: '小米 MiMo ASR', asrIflytek: '讯飞实时语音转写', @@ -876,11 +877,14 @@ export const zhCN = { asrAppleSpeech: 'Apple 语音（macOS）', }, elevenLabsUploadNotice: 'ElevenLabs 会将录音上传到所配置的端点进行批量转写。', - asrAdvancedNote: '以下高级选项仅影响「自定义 OpenAI 兼容」预设；命名厂商预设保持内置行为。', + zenmuxVocabularyNote: 'ZenMux 走 JSON 转写协议，不发送词典热词（prompt/hotwords）；词典仍会进入润色链路，但不会参与语音识别偏置。', + asrAdvancedNote: '以下高级选项仅影响「自定义 OpenAI 兼容」与「ZenMux」预设；其余命名厂商预设保持内置行为。', asrAdvancedVerboseJsonLabel: '分段指标 (verbose_json)', asrAdvancedVerboseJsonHint: '服务端支持时返回 segments 指标，用于幻听过滤；自建服务若不支持请保持关闭。', asrAdvancedChunkLabel: '分片时长 (ms)', asrAdvancedChunkHint: '0 = 不分片，整段发送；按片段多次请求，适合长录音或服务端单次请求时长受限。', + asrAdvancedEnableItnLabel: '数字归一化 (enable_itn)', + asrAdvancedEnableItnHint: '把口语数字/单位归一化为阿拉伯数字（如“二零二六年”→“2026年”）。关闭后保留原始文字。', volcengineAppKeyLabel: 'APP ID', volcengineAccessKeyLabel: 'Access Token', volcengineApiKeyLabel: 'API Key', diff --git a/openless-all/app/src/i18n/zh-TW.ts b/openless-all/app/src/i18n/zh-TW.ts index 490c9a9ce..9d18afe58 100644 --- a/openless-all/app/src/i18n/zh-TW.ts +++ b/openless-all/app/src/i18n/zh-TW.ts @@ -868,6 +868,7 @@ export const zhTW: typeof zhCN = { asrGroq: 'Groq Whisper-large-v3', asrWhisper: 'OpenAI Whisper（兼容）', asrOpenrouter: 'OpenRouter Whisper', + asrZenmux: 'ZenMux', asrOpenAiCompatible: '自訂 OpenAI 相容', asrXiaomiMimo: '小米 MiMo ASR', asrIflytek: '訊飛即時語音轉寫', @@ -878,11 +879,14 @@ export const zhTW: typeof zhCN = { asrAppleSpeech: 'Apple 語音（macOS）', }, elevenLabsUploadNotice: 'ElevenLabs 會將錄音上傳至已設定的端點進行批次轉寫。', - asrAdvancedNote: '以下進階選項僅影響「自訂 OpenAI 相容」預設；具名廠商預設維持內建行為。', + zenmuxVocabularyNote: 'ZenMux 走 JSON 轉寫協定，不傳送詞典熱詞（prompt/hotwords）；詞典仍會進入潤色鏈路，但不會參與語音辨識偏置。', + asrAdvancedNote: '以下進階選項僅影響「自訂 OpenAI 相容」與「ZenMux」預設；其餘具名廠商預設維持內建行為。', asrAdvancedVerboseJsonLabel: '分段指標 (verbose_json)', asrAdvancedVerboseJsonHint: '服務端支援時回傳 segments 指標，用於幻聽過濾；自建服務若不支援請保持關閉。', asrAdvancedChunkLabel: '分片時長 (ms)', asrAdvancedChunkHint: '0 = 不分片，整段傳送；按片段多次請求，適合長錄音或服務端單次請求時長受限。', + asrAdvancedEnableItnLabel: '數字正規化 (enable_itn)', + asrAdvancedEnableItnHint: '把口語數字/單位正規化為阿拉伯數字（如「二零二六年」→「2026年」）。關閉後保留原始文字。', volcengineAppKeyLabel: 'APP ID', volcengineAccessKeyLabel: 'Access Token', volcengineApiKeyLabel: 'API Key', diff --git a/openless-all/app/src/lib/advancedAsrConfig.test.ts b/openless-all/app/src/lib/advancedAsrConfig.test.ts index 9ff67346d..832542397 100644 --- a/openless-all/app/src/lib/advancedAsrConfig.test.ts +++ b/openless-all/app/src/lib/advancedAsrConfig.test.ts @@ -11,54 +11,64 @@ function assertEqual(actual: unknown, expected: unknown, name: string) { assertEqual( parseAdvancedAsrConfig(null), - { verboseJson: false, chunkDurationMs: null }, + { verboseJson: false, chunkDurationMs: null, enableItn: true }, 'null raw falls back to conservative defaults', ) assertEqual( parseAdvancedAsrConfig(''), - { verboseJson: false, chunkDurationMs: null }, + { verboseJson: false, chunkDurationMs: null, enableItn: true }, 'empty raw falls back to conservative defaults', ) assertEqual( parseAdvancedAsrConfig('not-json'), - { verboseJson: false, chunkDurationMs: null }, + { verboseJson: false, chunkDurationMs: null, enableItn: true }, 'invalid JSON falls back to conservative defaults', ) assertEqual( parseAdvancedAsrConfig('[1,2]'), - { verboseJson: false, chunkDurationMs: null }, + { verboseJson: false, chunkDurationMs: null, enableItn: true }, 'non-object JSON falls back to conservative defaults', ) assertEqual( parseAdvancedAsrConfig('{"verboseJson":true}'), - { verboseJson: true, chunkDurationMs: null }, + { verboseJson: true, chunkDurationMs: null, enableItn: true }, 'missing chunkDurationMs stays null', ) assertEqual( parseAdvancedAsrConfig('{"chunkDurationMs":0}'), - { verboseJson: false, chunkDurationMs: null }, + { verboseJson: false, chunkDurationMs: null, enableItn: true }, 'zero chunk duration means no chunking', ) assertEqual( parseAdvancedAsrConfig('{"chunkDurationMs":30000.9,"verboseJson":false}'), - { verboseJson: false, chunkDurationMs: 30000 }, + { verboseJson: false, chunkDurationMs: 30000, enableItn: true }, 'chunk duration is floored to integer', ) assertEqual( parseAdvancedAsrConfig('{"verboseJson":"yes"}'), - { verboseJson: false, chunkDurationMs: null }, + { verboseJson: false, chunkDurationMs: null, enableItn: true }, 'non-boolean verboseJson falls back to false', ) assertEqual( - parseAdvancedAsrConfig(serializeAdvancedAsrConfig({ verboseJson: true, chunkDurationMs: 30000 })), - { verboseJson: true, chunkDurationMs: 30000 }, + parseAdvancedAsrConfig('{"enableItn":false}'), + { verboseJson: false, chunkDurationMs: null, enableItn: false }, + 'explicit false enableItn is honored', +) + +assertEqual( + parseAdvancedAsrConfig(serializeAdvancedAsrConfig({ + verboseJson: true, + chunkDurationMs: 30000, + enableItn: false, + })), + { verboseJson: true, chunkDurationMs: 30000, enableItn: false }, 'serialize/parse round-trip preserves config', ) diff --git a/openless-all/app/src/lib/advancedAsrConfig.ts b/openless-all/app/src/lib/advancedAsrConfig.ts index 6f3c21016..d2ad0ba04 100644 --- a/openless-all/app/src/lib/advancedAsrConfig.ts +++ b/openless-all/app/src/lib/advancedAsrConfig.ts @@ -1,25 +1,27 @@ -// 通用 OpenAI 兼容 ASR（openai-compatible）的高级配置。 +// 通用 OpenAI 兼容 ASR（openai-compatible）与 ZenMux（zenmux）的高级配置。 // JSON 形状与后端 coordinator.rs::AdvancedAsrConfig 保持一致： -// {"verboseJson": bool, "chunkDurationMs": number|null}。 -// 解析策略与后端一致：缺失 / 非法一律回落保守默认（不发 response_format、不分片）。 +// {"verboseJson": bool, "chunkDurationMs": number|null, "enableItn": bool}。 +// 解析策略与后端一致：缺失 / 非法一律回落保守默认（不发 response_format、不分片、 +// enable_itn 默认开启）。 export interface AdvancedAsrConfig { verboseJson: boolean chunkDurationMs: number | null + enableItn: boolean } export function parseAdvancedAsrConfig(raw: string | null): AdvancedAsrConfig { if (!raw) { - return { verboseJson: false, chunkDurationMs: null } + return { verboseJson: false, chunkDurationMs: null, enableItn: true } } let value: unknown try { value = JSON.parse(raw) } catch { - return { verboseJson: false, chunkDurationMs: null } + return { verboseJson: false, chunkDurationMs: null, enableItn: true } } if (typeof value !== 'object' || value === null) { - return { verboseJson: false, chunkDurationMs: null } + return { verboseJson: false, chunkDurationMs: null, enableItn: true } } const record = value as Record const rawMs = record.chunkDurationMs @@ -30,6 +32,8 @@ export function parseAdvancedAsrConfig(raw: string | null): AdvancedAsrConfig { return { verboseJson: record.verboseJson === true, chunkDurationMs, + // 缺失 / 非布尔回落默认开启（与后端 parse_advanced_asr_config 一致）。 + enableItn: record.enableItn !== false, } } @@ -37,5 +41,6 @@ export function serializeAdvancedAsrConfig(config: AdvancedAsrConfig): string { return JSON.stringify({ verboseJson: config.verboseJson, chunkDurationMs: config.chunkDurationMs, + enableItn: config.enableItn, }) } diff --git a/openless-all/app/src/pages/settings/ProvidersSection.test.ts b/openless-all/app/src/pages/settings/ProvidersSection.test.ts index 7d14e7fa2..371d5e4f2 100644 --- a/openless-all/app/src/pages/settings/ProvidersSection.test.ts +++ b/openless-all/app/src/pages/settings/ProvidersSection.test.ts @@ -26,3 +26,17 @@ if (openAiCompatiblePreset.baseUrl !== '' || openAiCompatiblePreset.model !== '' `Custom OpenAI-compatible ASR preset must have no defaults (got baseUrl=${openAiCompatiblePreset.baseUrl}, model=${openAiCompatiblePreset.model})`, ); } + +const zenmuxPreset = ASR_PRESETS.find(p => p.id === 'zenmux'); + +if (!zenmuxPreset) { + throw new Error('ZenMux ASR preset is missing'); +} + +if (zenmuxPreset.baseUrl !== 'https://zenmux.ai/api/v1') { + throw new Error(`unexpected ZenMux base URL: ${zenmuxPreset.baseUrl}`); +} + +if (zenmuxPreset.model !== 'qwen/qwen3-asr-flash') { + throw new Error(`unexpected ZenMux default model: ${zenmuxPreset.model}`); +} diff --git a/openless-all/app/src/pages/settings/ProvidersSection.tsx b/openless-all/app/src/pages/settings/ProvidersSection.tsx index 2b10c6f44..40ca6caed 100644 --- a/openless-all/app/src/pages/settings/ProvidersSection.tsx +++ b/openless-all/app/src/pages/settings/ProvidersSection.tsx @@ -659,9 +659,14 @@ export function ProvidersSection({ kind = 'all' }: ProvidersSectionProps = {}) { {t('settings.providers.elevenLabsUploadNotice')} )} + {committedAsrProvider === 'zenmux' && ( +
+ {t('settings.providers.zenmuxVocabularyNote')} +
+ )} {/* 统一百炼「拉取模型」只写 model，不覆盖用户选择的区域或工作空间 endpoint。 */} setAsrModelRevision(v => v + 1)} /> - {committedAsrProvider === 'openai-compatible' && ( + {(committedAsrProvider === 'openai-compatible' || committedAsrProvider === 'zenmux') && ( )} @@ -672,12 +677,14 @@ export function ProvidersSection({ kind = 'all' }: ProvidersSectionProps = {}) { ); } -// 通用 OpenAI 兼容 ASR 的高级选项：仅对 openai-compatible 预设显示。 -// 命名厂商预设的怪癖开关（verbose_json / 分片等）是测过的硬编码行为，不开放。 +// ASR 高级选项：openai-compatible 与 zenmux 两个预设显示。 +// openai-compatible 暴露 verbose_json / 分片时长（其余命名厂商保持硬编码行为）； +// zenmux 暴露 enable_itn（数字归一化）开关，verbose_json / 分片对其无意义。 function AsrAdvancedOptions({ provider }: { provider: string }) { const { t } = useTranslation(); const [verboseJson, setVerboseJson] = useState(false); const [chunkDraft, setChunkDraft] = useState(''); + const [enableItn, setEnableItn] = useState(true); const [status, setStatus] = useState<'idle' | 'saving' | 'error'>('idle'); const [error, setError] = useState(''); @@ -692,6 +699,7 @@ function AsrAdvancedOptions({ provider }: { provider: string }) { const config = parseAdvancedAsrConfig(raw); setVerboseJson(config.verboseJson); setChunkDraft(config.chunkDurationMs ? String(config.chunkDurationMs) : ''); + setEnableItn(config.enableItn); } catch (err) { if (!cancelled) { setStatus('error'); @@ -710,7 +718,11 @@ function AsrAdvancedOptions({ provider }: { provider: string }) { return Math.floor(value); }; - const save = async (partial: { verboseJson?: boolean; chunkDurationMs?: number | null }) => { + const save = async (partial: { + verboseJson?: boolean + chunkDurationMs?: number | null + enableItn?: boolean + }) => { setStatus('saving'); setError(''); const next: AdvancedAsrConfig = { @@ -719,11 +731,13 @@ function AsrAdvancedOptions({ provider }: { provider: string }) { partial.chunkDurationMs !== undefined ? partial.chunkDurationMs : parseChunkDraft(chunkDraft), + enableItn: partial.enableItn ?? enableItn, }; try { await setCredential('asr.advanced_config', serializeAdvancedAsrConfig(next), provider); setVerboseJson(next.verboseJson); setChunkDraft(next.chunkDurationMs ? String(next.chunkDurationMs) : ''); + setEnableItn(next.enableItn); setStatus('idle'); } catch (err) { setStatus('error'); @@ -744,31 +758,42 @@ function AsrAdvancedOptions({ provider }: { provider: string }) { > {t('settings.providers.asrAdvancedNote')} - - void save({ verboseJson: next })} /> - - - setChunkDraft(e.target.value)} - onBlur={() => void save({ chunkDurationMs: parseChunkDraft(chunkDraft) })} - onKeyDown={(e) => { - if (e.key === 'Enter') (e.target as HTMLInputElement).blur(); - }} - style={inputStyle} - /> - + {provider === 'zenmux' ? ( + + void save({ enableItn: next })} /> + + ) : ( + <> + + void save({ verboseJson: next })} /> + + + setChunkDraft(e.target.value)} + onBlur={() => void save({ chunkDurationMs: parseChunkDraft(chunkDraft) })} + onKeyDown={(e) => { + if (e.key === 'Enter') (e.target as HTMLInputElement).blur(); + }} + style={inputStyle} + /> + + + )} {status === 'error' && (
{t('common.operationFailed')}: {error} diff --git a/openless-all/app/src/pages/settings/shared.tsx b/openless-all/app/src/pages/settings/shared.tsx index 1f6ad5b79..231d170b2 100644 --- a/openless-all/app/src/pages/settings/shared.tsx +++ b/openless-all/app/src/pages/settings/shared.tsx @@ -235,6 +235,10 @@ export const ASR_PRESETS = [ // OpenRouter 的 /audio/transcriptions 走 application/json + base64（issue #582）， // 后端 coordinator.rs::whisper_request_format 对该 id 切换到 OpenRouterJson 编码。 { id: 'openrouter', nameKey: 'asrOpenrouter', baseUrl: 'https://openrouter.ai/api/v1', model: 'openai/whisper-large-v3-turbo' }, + // ZenMux 聚合平台的 /audio/transcriptions 同为 application/json + base64 + // （issue #837），后端 whisper_request_format 对该 id 切 ZenMuxJson； + // 语言跟随工作语言，enable_itn 开关见 AsrAdvancedOptions（仅该预设显示）。 + { id: 'zenmux', nameKey: 'asrZenmux', baseUrl: 'https://zenmux.ai/api/v1', model: 'qwen/qwen3-asr-flash' }, // 通用 OpenAI 兼容端点（自建 / 局域网 llama.cpp 等）：无默认 endpoint/model， // 用户必填；verbose_json 与分片时长由高级选项按 provider 配置（见 // ProvidersSection 的 AsrAdvancedOptions），默认行为最保守。