feat(asr): 支持 MiniMax 语音识别服务接入 (fixes #1080)

openless-all/app/crates/openless-core/src/asr/whisper.rs

@@ -57,6 +57,8 @@ pub struct WhisperBatchASR {
     /// 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<String>,
+    /// 自定义端点路径（默认 `/audio/transcriptions`；MiniMax 等厂商为 `/speech_to_text`）。
+    endpoint_path: Option<String>,
     buffer: Mutex<Vec<u8>>,
 }
 
@@ -80,10 +82,17 @@ impl WhisperBatchASR {
             language: None,
             enable_itn: true,
             hotwords: Vec::new(),
+            endpoint_path: None,
             buffer: Mutex::new(Vec::new()),
         }
     }
 
+    /// 设置自定义端点路径（例如 `"/speech_to_text"`）。
+    pub fn with_endpoint_path(mut self, path: impl Into<String>) -> Self {
+        self.endpoint_path = Some(path.into());
+        self
+    }
+
     /// 设置请求体编码方式（默认 `Multipart`）。OpenRouter 需 `OpenRouterJson`。
     /// 用 builder 而非给 `new()` 加参数，避免改动既有 4 处构造点的签名。
     pub fn with_request_format(mut self, request_format: AsrRequestFormat) -> Self {
@@ -166,7 +175,7 @@ impl WhisperBatchASR {
             .map(|chunk| i16::from_le_bytes([chunk[0], chunk[1]]))
             .collect();
         let wav = encode_wav_16k_mono(&samples);
-        let url = transcription_url(&self.base_url)?;
+        let url = transcription_url(&self.base_url, self.endpoint_path.as_deref())?;
         let client = crate::net::http();
 
         let request = match self.request_format {
@@ -275,6 +284,19 @@ impl WhisperBatchASR {
         }
 
         let json: serde_json::Value = resp.json().await.context("parse Whisper response")?;
+        if let Some(base_resp) = json.get("base_resp") {
+            let status_code = base_resp
+                .get("status_code")
+                .and_then(|c| c.as_i64())
+                .unwrap_or(0);
+            if status_code != 0 {
+                let msg = base_resp
+                    .get("status_msg")
+                    .and_then(|m| m.as_str())
+                    .unwrap_or("unknown error");
+                anyhow::bail!("MiniMax API error {}: {}", status_code, msg);
+            }
+        }
         if self.verbose_json {
             // verbose_json：セグメントのメタデータで幻聴を除いた本文を組む。
             // segments が無い応答では内部で従来どおり text にフォールバック。
@@ -402,16 +424,36 @@ pub fn split_pcm_by_duration(pcm: &[u8], max_chunk_duration_ms: Option<u64>) ->
     pcm.chunks(bytes_per_chunk).collect()
 }
 
-fn transcription_url(base_url: &str) -> Result<String> {
+fn transcription_url(base_url: &str, endpoint_path: Option<&str>) -> Result<String> {
     let parsed = reqwest::Url::parse(base_url.trim()).context("parse Whisper base URL")?;
     let mut url = parsed.clone();
     let path = parsed.path().trim_end_matches('/');
-    let next_path = if path.ends_with("/audio/transcriptions") {
+    let next_path = if let Some(target) = endpoint_path {
+        let target = target.trim();
+        if path.ends_with(target) {
+            path.to_string()
+        } else {
+            format!(
+                "{path}{}",
+                if target.starts_with('/') {
+                    target.to_string()
+                } else {
+                    format!("/{target}")
+                }
+            )
+        }
+    } else if path.ends_with("/audio/transcriptions") || path.ends_with("/speech_to_text") {
         path.to_string()
     } else if path.ends_with("/audio") {
         format!("{path}/transcriptions")
     } else if let Some(prefix) = path.strip_suffix("/chat/completions") {
         format!("{prefix}/audio/transcriptions")
+    } else if parsed
+        .host_str()
+        .map_or(false, |h| h.ends_with("minimaxi.com") || h.ends_with("minimax.chat") || h.contains("minimax"))
+        || path.contains("minimax")
+    {
+        format!("{path}/speech_to_text")
     } else {
         format!("{path}/audio/transcriptions")
     };
@@ -745,24 +787,54 @@ mod tests {
     #[test]
     fn transcription_url_accepts_base_audio_or_full_endpoint() {
         assert_eq!(
-            transcription_url("https://open.bigmodel.cn/api/paas/v4").unwrap(),
+            transcription_url("https://open.bigmodel.cn/api/paas/v4", None).unwrap(),
             "https://open.bigmodel.cn/api/paas/v4/audio/transcriptions"
         );
         assert_eq!(
-            transcription_url("https://open.bigmodel.cn/api/paas/v4/audio").unwrap(),
+            transcription_url("https://open.bigmodel.cn/api/paas/v4/audio", None).unwrap(),
             "https://open.bigmodel.cn/api/paas/v4/audio/transcriptions"
         );
         assert_eq!(
-            transcription_url("https://open.bigmodel.cn/api/paas/v4/audio/transcriptions").unwrap(),
+            transcription_url(
+                "https://open.bigmodel.cn/api/paas/v4/audio/transcriptions",
+                None
+            )
+            .unwrap(),
             "https://open.bigmodel.cn/api/paas/v4/audio/transcriptions"
         );
         assert_eq!(
             transcription_url(
-                "https://open.bigmodel.cn/api/paas/v4/audio/transcriptions?api-version=2026-01-01"
+                "https://open.bigmodel.cn/api/paas/v4/audio/transcriptions?api-version=2026-01-01",
+                None
             )
             .unwrap(),
             "https://open.bigmodel.cn/api/paas/v4/audio/transcriptions?api-version=2026-01-01"
         );
+        // MiniMax & speech_to_text tests
+        assert_eq!(
+            transcription_url("https://api.minimaxi.com/v1", None).unwrap(),
+            "https://api.minimaxi.com/v1/speech_to_text"
+        );
+        assert_eq!(
+            transcription_url("https://api.minimax.chat/v1", None).unwrap(),
+            "https://api.minimax.chat/v1/speech_to_text"
+        );
+        assert_eq!(
+            transcription_url("https://custom-proxy.com/v1/speech_to_text", None).unwrap(),
+            "https://custom-proxy.com/v1/speech_to_text"
+        );
+        assert_eq!(
+            transcription_url("https://custom-proxy.com/v1", Some("/speech_to_text")).unwrap(),
+            "https://custom-proxy.com/v1/speech_to_text"
+        );
+        assert_eq!(
+            transcription_url(
+                "https://custom-proxy.com/v1/speech_to_text",
+                Some("/speech_to_text")
+            )
+            .unwrap(),
+            "https://custom-proxy.com/v1/speech_to_text"
+        );
     }
 
     #[test]

openless-all/app/crates/openless-core/src/cloud_providers.rs

@@ -60,6 +60,7 @@ pub const SHARED_CLOUD_ASR_PROVIDER_TYPES: &[&str] = &[
     "openrouter",
     "orcarouter",
     "zenmux",
+    "minimax",
     "openai-compatible",
     "xiaomi-mimo-asr",
     "iflytek",
@@ -539,6 +540,9 @@ async fn build_cloud_transcription_session(
                 ),
             )
             .with_request_format(crate::provider_rules::whisper_request_format(provider_type));
+            if provider_type == "minimax" {
+                provider = provider.with_endpoint_path("/speech_to_text");
+            }
             if crate::provider_rules::whisper_uses_hotwords(provider_type) {
                 provider = provider.with_hotwords(context.polish.hotwords.clone());
             }

openless-all/app/crates/openless-core/src/provider_rules.rs

@@ -32,6 +32,7 @@ const ASR_PROVIDER_TYPES: &[(&str, &str)] = &[
     ("siliconflow", "asrSiliconflow"),
     ("stepfun", "asrStepfun"),
     ("zhipu", "asrZhipu"),
+    ("minimax", "asrMinimax"),
     ("groq", "asrGroq"),
     ("whisper", "asrWhisper"),
     ("openrouter", "asrOpenrouter"),
@@ -353,6 +354,7 @@ fn static_models(kind: ProviderKind, provider_type: &str) -> &'static [&'static
         (ProviderKind::Asr, "xiaomi-mimo-asr") => &[crate::asr::mimo::DEFAULT_MODEL],
         (ProviderKind::Asr, "bailian-fun-asr-flash") => DASHSCOPE_MODELS,
         (ProviderKind::Asr, "elevenlabs") => &[crate::asr::elevenlabs::DEFAULT_MODEL],
+        (ProviderKind::Asr, "minimax") => &["asr-1.0"],
         (ProviderKind::Llm, crate::polish::CODEX_OAUTH_PROVIDER_ID) => &[
             crate::polish::CODEX_DEFAULT_MODEL,
             "gpt-5.3-codex",
@@ -555,6 +557,7 @@ pub fn default_asr_endpoint(provider_type: &str) -> Option<&'static str> {
         "siliconflow" => Some("https://api.siliconflow.cn/v1"),
         "stepfun" => Some("https://api.stepfun.com/v1"),
         "zhipu" => Some("https://open.bigmodel.cn/api/paas/v4"),
+        "minimax" => Some("https://api.minimaxi.com/v1"),
         "groq" => Some("https://api.groq.com/openai/v1"),
         "whisper" => Some("https://api.openai.com/v1"),
         "openrouter" => Some("https://openrouter.ai/api/v1"),
@@ -574,6 +577,7 @@ pub fn default_asr_model(provider_type: &str) -> Option<&'static str> {
         "siliconflow" => Some("FunAudioLLM/SenseVoiceSmall"),
         "stepfun" => Some("stepaudio-2.5-asr"),
         "zhipu" => Some("glm-asr-2512"),
+        "minimax" => Some("asr-1.0"),
         "groq" => Some("whisper-large-v3-turbo"),
         "whisper" => Some("whisper-1"),
         "openrouter" => Some("openai/whisper-large-v3-turbo"),
@@ -787,7 +791,7 @@ pub fn is_tencent_cloud_provider(id: &str) -> bool {
 pub fn is_whisper_compatible_provider(id: &str) -> bool {
     matches!(
         id,
-        "whisper" | "siliconflow" | "zhipu" | "groq" | "openrouter" | "stepfun" | "zenmux"
+        "whisper" | "siliconflow" | "zhipu" | "groq" | "openrouter" | "stepfun" | "zenmux" | "minimax"
     ) || id == OPENAI_COMPATIBLE_ASR_PROVIDER_ID
 }
 
@@ -1259,6 +1263,26 @@ mod tests {
         assert!(llm.supported_request_formats.is_empty());
     }
 
+    #[test]
+    fn minimax_asr_supplies_defaults_and_whisper_compatibility() {
+        assert!(crate::cloud_providers::SHARED_CLOUD_ASR_PROVIDER_TYPES.contains(&"minimax"));
+        let asr = provider_descriptor(ProviderKind::Asr, "minimax").unwrap();
+        assert_eq!(asr.label_key, "asrMinimax");
+        assert_eq!(
+            asr.default_endpoint.as_deref(),
+            Some("https://api.minimaxi.com/v1")
+        );
+        assert_eq!(asr.default_model.as_deref(), Some("asr-1.0"));
+        assert_eq!(asr.static_models, vec!["asr-1.0"]);
+        assert_eq!(asr.auth_requirement, AuthRequirement::ApiKey);
+        assert_eq!(asr.validation_probe, ValidationProbe::AsrSilence);
+        assert!(is_whisper_compatible_provider("minimax"));
+        assert_eq!(
+            active_asr_provider_kind("minimax"),
+            ActiveAsrProviderKind::WhisperCompatible
+        );
+    }
+
     #[test]
     fn descriptors_are_the_single_source_for_defaults_auth_and_probes() {
         let compatible = provider_descriptor(ProviderKind::Asr, "openai-compatible").unwrap();

openless-all/app/src/i18n/de.ts

@@ -1331,6 +1331,7 @@ export const de: typeof zhCN = {
         asrSiliconflow: 'SiliconFlow SenseVoice',
         asrStepfun: 'StepFun StepAudio ASR',
         asrZhipu: 'Zhipu GLM-ASR',
+        asrMinimax: 'MiniMax ASR',
         asrGroq: 'Groq Whisper-large-v3',
         asrWhisper: 'OpenAI Whisper (kompatibel)',
         asrOpenrouter: 'OpenRouter Whisper',

openless-all/app/src/i18n/en.ts

@@ -1297,6 +1297,7 @@ export const en: typeof zhCN = {
         asrSiliconflow: 'SiliconFlow SenseVoice',
         asrStepfun: 'StepFun StepAudio ASR',
         asrZhipu: 'Zhipu GLM-ASR',
+        asrMinimax: 'MiniMax ASR',
         asrGroq: 'Groq Whisper-large-v3',
         asrWhisper: 'OpenAI Whisper (compatible)',
         asrOpenrouter: 'OpenRouter Whisper',

openless-all/app/src/i18n/es.ts

@@ -1323,6 +1323,7 @@ export const es: typeof zhCN = {
         asrSiliconflow: 'SiliconFlow SenseVoice',
         asrStepfun: 'StepFun StepAudio ASR',
         asrZhipu: 'Zhipu GLM-ASR',
+        asrMinimax: 'MiniMax ASR',
         asrGroq: 'Groq Whisper-large-v3',
         asrWhisper: 'OpenAI Whisper (compatible)',
         asrOpenrouter: 'OpenRouter Whisper',

openless-all/app/src/i18n/fr.ts

@@ -1340,6 +1340,7 @@ export const fr: typeof zhCN = {
         asrSiliconflow: 'SiliconFlow SenseVoice',
         asrStepfun: 'StepFun StepAudio ASR',
         asrZhipu: 'Zhipu GLM-ASR',
+        asrMinimax: 'MiniMax ASR',
         asrGroq: 'Groq Whisper-large-v3',
         asrWhisper: 'OpenAI Whisper (compatible)',
         asrOpenrouter: 'OpenRouter Whisper',

openless-all/app/src/i18n/ja.ts

@@ -1284,6 +1284,7 @@ export const ja: typeof zhCN = {
         asrSiliconflow: 'SiliconFlow SenseVoice',
         asrStepfun: 'StepFun StepAudio ASR',
         asrZhipu: 'Zhipu GLM-ASR',
+        asrMinimax: 'MiniMax ASR',
         asrGroq: 'Groq Whisper-large-v3',
         asrWhisper: 'OpenAI Whisper（互換）',
         asrOpenrouter: 'OpenRouter Whisper',

openless-all/app/src/i18n/ko.ts

@@ -1276,6 +1276,7 @@ export const ko: typeof zhCN = {
         asrSiliconflow: 'SiliconFlow SenseVoice',
         asrStepfun: 'StepFun StepAudio ASR',
         asrZhipu: 'Zhipu GLM-ASR',
+        asrMinimax: 'MiniMax ASR',
         asrGroq: 'Groq Whisper-large-v3',
         asrWhisper: 'OpenAI Whisper(호환)',
         asrOpenrouter: 'OpenRouter Whisper',

openless-all/app/src/i18n/zh-CN.ts

@@ -1235,6 +1235,7 @@ export const zhCN = {
         asrSiliconflow: '硅基流动 SenseVoice',
         asrStepfun: '阶跃星辰 StepAudio',
         asrZhipu: '智谱 GLM-ASR',
+        asrMinimax: 'MiniMax ASR',
         asrGroq: 'Groq Whisper-large-v3',
         asrWhisper: 'OpenAI Whisper（兼容）',
         asrOpenrouter: 'OpenRouter Whisper',

openless-all/app/src/i18n/zh-TW.ts

@@ -1237,6 +1237,7 @@ export const zhTW: typeof zhCN = {
         asrSiliconflow: '硅基流動 SenseVoice',
         asrStepfun: '階躍星辰 StepAudio',
         asrZhipu: '智譜 GLM-ASR',
+        asrMinimax: 'MiniMax ASR',
         asrGroq: 'Groq Whisper-large-v3',
         asrWhisper: 'OpenAI Whisper（兼容）',
         asrOpenrouter: 'OpenRouter Whisper',