@circuitwall/jarela 2.2.1 → 2.3.0

package/.next/standalone/.next/server/app/api/v1/voice/transcribe/route.js

@@ -638,7 +638,229 @@ module.exports = require("node:fs");
 
 /***/ }),
 
-/***/ 73809:
+/***/ 75919:
+/***/ ((module) => {
+
+module.exports = require("node:worker_threads");
+
+/***/ }),
+
+/***/ 76760:
+/***/ ((module) => {
+
+module.exports = require("node:path");
+
+/***/ }),
+
+/***/ 77598:
+/***/ ((module) => {
+
+module.exports = require("node:crypto");
+
+/***/ }),
+
+/***/ 79428:
+/***/ ((module) => {
+
+module.exports = require("buffer");
+
+/***/ }),
+
+/***/ 79766:
+/***/ ((__unused_webpack_module, __webpack_exports__, __webpack_require__) => {
+
+/* harmony export */ __webpack_require__.d(__webpack_exports__, {
+/* harmony export */   DM: () => (/* binding */ resolveGoogleApiKey),
+/* harmony export */   R$: () => (/* binding */ geminiTts),
+/* harmony export */   Tm: () => (/* binding */ geminiStt)
+/* harmony export */ });
+/* harmony import */ var _lib_env_config__WEBPACK_IMPORTED_MODULE_0__ = __webpack_require__(67500);
+/* harmony import */ var _lib_utils_google_api__WEBPACK_IMPORTED_MODULE_1__ = __webpack_require__(67118);
+// Gemini voice helpers — TTS via the *-preview-tts models and STT by
+// feeding inline audio bytes to a multimodal Gemini model. Both call the
+// native REST API (the OpenAI-compat proxy doesn't expose AUDIO modality
+// or inline_data parts). API key is resolved from the "google" integration,
+// same source as lib/tools/generate_image.ts.
+
+
+const ENDPOINT = "https://generativelanguage.googleapis.com/v1beta/models";
+function resolveGoogleApiKey() {
+    return (0,_lib_utils_google_api__WEBPACK_IMPORTED_MODULE_1__/* .resolveGoogleApiKey */ .D)();
+}
+/**
+ * Synthesize speech with Gemini TTS. Gemini returns PCM L16 (raw 16-bit
+ * little-endian, 24 kHz mono); we wrap it with a WAV header so any
+ * <audio> element can play it.
+ *
+ * Style/tone is steered by prepending a natural-language instruction to
+ * the spoken text (Gemini's recommended pattern), e.g. style="Say warmly
+ * and slowly" produces "Say warmly and slowly: <text>". For multi-speaker
+ * scenes, pass `speakers` with up to 2 entries and write the text as
+ * `Name: line ...` per turn — Gemini will switch voices accordingly.
+ */ async function geminiTts(opts) {
+    const url = `${ENDPOINT}/${encodeURIComponent(opts.model)}:generateContent?key=${encodeURIComponent(opts.apiKey)}`;
+    const style = opts.style?.trim();
+    const spoken = style ? `${style.replace(/:\s*$/, "")}: ${opts.text}` : opts.text;
+    const speakers = (opts.speakers ?? []).filter((s)=>s.name && s.voiceName).slice(0, 2);
+    const speechConfig = speakers.length >= 2 ? {
+        multiSpeakerVoiceConfig: {
+            speakerVoiceConfigs: speakers.map((s)=>({
+                    speaker: s.name,
+                    voiceConfig: {
+                        prebuiltVoiceConfig: {
+                            voiceName: s.voiceName
+                        }
+                    }
+                }))
+        }
+    } : {
+        voiceConfig: {
+            prebuiltVoiceConfig: {
+                voiceName: opts.voiceName
+            }
+        }
+    };
+    const body = {
+        contents: [
+            {
+                role: "user",
+                parts: [
+                    {
+                        text: spoken
+                    }
+                ]
+            }
+        ],
+        generationConfig: {
+            responseModalities: [
+                "AUDIO"
+            ],
+            speechConfig
+        }
+    };
+    const res = await fetch(url, {
+        method: "POST",
+        headers: {
+            "Content-Type": "application/json"
+        },
+        body: JSON.stringify(body),
+        signal: (0,_lib_utils_google_api__WEBPACK_IMPORTED_MODULE_1__/* .timeoutSignal */ ._)((0,_lib_env_config__WEBPACK_IMPORTED_MODULE_0__/* .getConfig */ .z)().voiceTimeoutMs)
+    });
+    const json = await res.json();
+    if (!res.ok) {
+        throw new Error(`Gemini TTS ${res.status}: ${json.error?.message ?? "request failed"}`);
+    }
+    const parts = json.candidates?.[0]?.content?.parts ?? [];
+    for (const p of parts){
+        const inline = p.inlineData ?? p.inline_data;
+        const data = inline?.data;
+        const mime = inline?.mimeType ?? inline?.mime_type ?? "";
+        if (data) {
+            const pcm = Buffer.from(data, "base64");
+            const rate = parseSampleRate(mime) ?? 24000;
+            return {
+                wav: pcmToWav(pcm, rate, 1, 16),
+                mime: "audio/wav"
+            };
+        }
+    }
+    const text = parts.map((p)=>p.text).filter(Boolean).join(" ").trim();
+    throw new Error(text ? `Gemini TTS returned no audio: ${text}` : "Gemini TTS returned no audio");
+}
+/**
+ * Transcribe audio via a Gemini multimodal model. Returns the verbatim
+ * transcript, preserving the speaker's original language (including code-
+ * switched English/Chinese).
+ */ async function geminiStt(opts) {
+    const url = `${ENDPOINT}/${encodeURIComponent(opts.model)}:generateContent?key=${encodeURIComponent(opts.apiKey)}`;
+    const body = {
+        contents: [
+            {
+                role: "user",
+                parts: [
+                    {
+                        text: "Transcribe the following audio verbatim. Preserve the speaker's " + "original language, including mixed English and Chinese. Output ONLY " + "the transcript text — no quotes, no commentary, no language tags."
+                    },
+                    {
+                        inline_data: {
+                            mime_type: opts.mimeType,
+                            data: opts.audio.toString("base64")
+                        }
+                    }
+                ]
+            }
+        ]
+    };
+    const res = await fetch(url, {
+        method: "POST",
+        headers: {
+            "Content-Type": "application/json"
+        },
+        body: JSON.stringify(body),
+        signal: (0,_lib_utils_google_api__WEBPACK_IMPORTED_MODULE_1__/* .timeoutSignal */ ._)((0,_lib_env_config__WEBPACK_IMPORTED_MODULE_0__/* .getConfig */ .z)().voiceTimeoutMs)
+    });
+    const json = await res.json();
+    if (!res.ok) {
+        throw new Error(`Gemini STT ${res.status}: ${json.error?.message ?? "request failed"}`);
+    }
+    const parts = json.candidates?.[0]?.content?.parts ?? [];
+    const text = parts.map((p)=>p.text).filter(Boolean).join("").trim();
+    if (!text) throw new Error("Gemini STT returned no transcript");
+    return {
+        text
+    };
+}
+// `audio/L16;codec=pcm;rate=24000` → 24000
+function parseSampleRate(mime) {
+    const m = /rate=(\d+)/i.exec(mime);
+    if (!m) return null;
+    const n = parseInt(m[1], 10);
+    return Number.isFinite(n) && n > 0 ? n : null;
+}
+// Build a RIFF/WAVE header for raw PCM. Standard 16-bit, mono unless told
+// otherwise. No fancy chunks — just enough to satisfy browser audio.
+function pcmToWav(pcm, sampleRate, channels, bitsPerSample) {
+    const byteRate = sampleRate * channels * bitsPerSample / 8;
+    const blockAlign = channels * bitsPerSample / 8;
+    const dataSize = pcm.length;
+    const header = Buffer.alloc(44);
+    header.write("RIFF", 0);
+    header.writeUInt32LE(36 + dataSize, 4);
+    header.write("WAVE", 8);
+    header.write("fmt ", 12);
+    header.writeUInt32LE(16, 16); // fmt chunk size
+    header.writeUInt16LE(1, 20); // PCM
+    header.writeUInt16LE(channels, 22);
+    header.writeUInt32LE(sampleRate, 24);
+    header.writeUInt32LE(byteRate, 28);
+    header.writeUInt16LE(blockAlign, 32);
+    header.writeUInt16LE(bitsPerSample, 34);
+    header.write("data", 36);
+    header.writeUInt32LE(dataSize, 40);
+    return Buffer.concat([
+        header,
+        pcm
+    ]);
+}
+
+
+/***/ }),
+
+/***/ 80099:
+/***/ ((module) => {
+
+module.exports = require("node:sqlite");
+
+/***/ }),
+
+/***/ 80481:
+/***/ ((module) => {
+
+module.exports = require("node:readline");
+
+/***/ }),
+
+/***/ 80747:
 /***/ ((__unused_webpack_module, __webpack_exports__, __webpack_require__) => {
 
 __webpack_require__.r(__webpack_exports__);
@@ -1080,228 +1302,6 @@ async function handler(req, res, ctx) {
 //# sourceMappingURL=app-route.js.map
 
 
-/***/ }),
-
-/***/ 75919:
-/***/ ((module) => {
-
-module.exports = require("node:worker_threads");
-
-/***/ }),
-
-/***/ 76760:
-/***/ ((module) => {
-
-module.exports = require("node:path");
-
-/***/ }),
-
-/***/ 77598:
-/***/ ((module) => {
-
-module.exports = require("node:crypto");
-
-/***/ }),
-
-/***/ 79428:
-/***/ ((module) => {
-
-module.exports = require("buffer");
-
-/***/ }),
-
-/***/ 79766:
-/***/ ((__unused_webpack_module, __webpack_exports__, __webpack_require__) => {
-
-/* harmony export */ __webpack_require__.d(__webpack_exports__, {
-/* harmony export */   DM: () => (/* binding */ resolveGoogleApiKey),
-/* harmony export */   R$: () => (/* binding */ geminiTts),
-/* harmony export */   Tm: () => (/* binding */ geminiStt)
-/* harmony export */ });
-/* harmony import */ var _lib_env_config__WEBPACK_IMPORTED_MODULE_0__ = __webpack_require__(67500);
-/* harmony import */ var _lib_utils_google_api__WEBPACK_IMPORTED_MODULE_1__ = __webpack_require__(67118);
-// Gemini voice helpers — TTS via the *-preview-tts models and STT by
-// feeding inline audio bytes to a multimodal Gemini model. Both call the
-// native REST API (the OpenAI-compat proxy doesn't expose AUDIO modality
-// or inline_data parts). API key is resolved from the "google" integration,
-// same source as lib/tools/generate_image.ts.
-
-
-const ENDPOINT = "https://generativelanguage.googleapis.com/v1beta/models";
-function resolveGoogleApiKey() {
-    return (0,_lib_utils_google_api__WEBPACK_IMPORTED_MODULE_1__/* .resolveGoogleApiKey */ .D)();
-}
-/**
- * Synthesize speech with Gemini TTS. Gemini returns PCM L16 (raw 16-bit
- * little-endian, 24 kHz mono); we wrap it with a WAV header so any
- * <audio> element can play it.
- *
- * Style/tone is steered by prepending a natural-language instruction to
- * the spoken text (Gemini's recommended pattern), e.g. style="Say warmly
- * and slowly" produces "Say warmly and slowly: <text>". For multi-speaker
- * scenes, pass `speakers` with up to 2 entries and write the text as
- * `Name: line ...` per turn — Gemini will switch voices accordingly.
- */ async function geminiTts(opts) {
-    const url = `${ENDPOINT}/${encodeURIComponent(opts.model)}:generateContent?key=${encodeURIComponent(opts.apiKey)}`;
-    const style = opts.style?.trim();
-    const spoken = style ? `${style.replace(/:\s*$/, "")}: ${opts.text}` : opts.text;
-    const speakers = (opts.speakers ?? []).filter((s)=>s.name && s.voiceName).slice(0, 2);
-    const speechConfig = speakers.length >= 2 ? {
-        multiSpeakerVoiceConfig: {
-            speakerVoiceConfigs: speakers.map((s)=>({
-                    speaker: s.name,
-                    voiceConfig: {
-                        prebuiltVoiceConfig: {
-                            voiceName: s.voiceName
-                        }
-                    }
-                }))
-        }
-    } : {
-        voiceConfig: {
-            prebuiltVoiceConfig: {
-                voiceName: opts.voiceName
-            }
-        }
-    };
-    const body = {
-        contents: [
-            {
-                role: "user",
-                parts: [
-                    {
-                        text: spoken
-                    }
-                ]
-            }
-        ],
-        generationConfig: {
-            responseModalities: [
-                "AUDIO"
-            ],
-            speechConfig
-        }
-    };
-    const res = await fetch(url, {
-        method: "POST",
-        headers: {
-            "Content-Type": "application/json"
-        },
-        body: JSON.stringify(body),
-        signal: (0,_lib_utils_google_api__WEBPACK_IMPORTED_MODULE_1__/* .timeoutSignal */ ._)((0,_lib_env_config__WEBPACK_IMPORTED_MODULE_0__/* .getConfig */ .z)().voiceTimeoutMs)
-    });
-    const json = await res.json();
-    if (!res.ok) {
-        throw new Error(`Gemini TTS ${res.status}: ${json.error?.message ?? "request failed"}`);
-    }
-    const parts = json.candidates?.[0]?.content?.parts ?? [];
-    for (const p of parts){
-        const inline = p.inlineData ?? p.inline_data;
-        const data = inline?.data;
-        const mime = inline?.mimeType ?? inline?.mime_type ?? "";
-        if (data) {
-            const pcm = Buffer.from(data, "base64");
-            const rate = parseSampleRate(mime) ?? 24000;
-            return {
-                wav: pcmToWav(pcm, rate, 1, 16),
-                mime: "audio/wav"
-            };
-        }
-    }
-    const text = parts.map((p)=>p.text).filter(Boolean).join(" ").trim();
-    throw new Error(text ? `Gemini TTS returned no audio: ${text}` : "Gemini TTS returned no audio");
-}
-/**
- * Transcribe audio via a Gemini multimodal model. Returns the verbatim
- * transcript, preserving the speaker's original language (including code-
- * switched English/Chinese).
- */ async function geminiStt(opts) {
-    const url = `${ENDPOINT}/${encodeURIComponent(opts.model)}:generateContent?key=${encodeURIComponent(opts.apiKey)}`;
-    const body = {
-        contents: [
-            {
-                role: "user",
-                parts: [
-                    {
-                        text: "Transcribe the following audio verbatim. Preserve the speaker's " + "original language, including mixed English and Chinese. Output ONLY " + "the transcript text — no quotes, no commentary, no language tags."
-                    },
-                    {
-                        inline_data: {
-                            mime_type: opts.mimeType,
-                            data: opts.audio.toString("base64")
-                        }
-                    }
-                ]
-            }
-        ]
-    };
-    const res = await fetch(url, {
-        method: "POST",
-        headers: {
-            "Content-Type": "application/json"
-        },
-        body: JSON.stringify(body),
-        signal: (0,_lib_utils_google_api__WEBPACK_IMPORTED_MODULE_1__/* .timeoutSignal */ ._)((0,_lib_env_config__WEBPACK_IMPORTED_MODULE_0__/* .getConfig */ .z)().voiceTimeoutMs)
-    });
-    const json = await res.json();
-    if (!res.ok) {
-        throw new Error(`Gemini STT ${res.status}: ${json.error?.message ?? "request failed"}`);
-    }
-    const parts = json.candidates?.[0]?.content?.parts ?? [];
-    const text = parts.map((p)=>p.text).filter(Boolean).join("").trim();
-    if (!text) throw new Error("Gemini STT returned no transcript");
-    return {
-        text
-    };
-}
-// `audio/L16;codec=pcm;rate=24000` → 24000
-function parseSampleRate(mime) {
-    const m = /rate=(\d+)/i.exec(mime);
-    if (!m) return null;
-    const n = parseInt(m[1], 10);
-    return Number.isFinite(n) && n > 0 ? n : null;
-}
-// Build a RIFF/WAVE header for raw PCM. Standard 16-bit, mono unless told
-// otherwise. No fancy chunks — just enough to satisfy browser audio.
-function pcmToWav(pcm, sampleRate, channels, bitsPerSample) {
-    const byteRate = sampleRate * channels * bitsPerSample / 8;
-    const blockAlign = channels * bitsPerSample / 8;
-    const dataSize = pcm.length;
-    const header = Buffer.alloc(44);
-    header.write("RIFF", 0);
-    header.writeUInt32LE(36 + dataSize, 4);
-    header.write("WAVE", 8);
-    header.write("fmt ", 12);
-    header.writeUInt32LE(16, 16); // fmt chunk size
-    header.writeUInt16LE(1, 20); // PCM
-    header.writeUInt16LE(channels, 22);
-    header.writeUInt32LE(sampleRate, 24);
-    header.writeUInt32LE(byteRate, 28);
-    header.writeUInt16LE(blockAlign, 32);
-    header.writeUInt16LE(bitsPerSample, 34);
-    header.write("data", 36);
-    header.writeUInt32LE(dataSize, 40);
-    return Buffer.concat([
-        header,
-        pcm
-    ]);
-}
-
-
-/***/ }),
-
-/***/ 80099:
-/***/ ((module) => {
-
-module.exports = require("node:sqlite");
-
-/***/ }),
-
-/***/ 80481:
-/***/ ((module) => {
-
-module.exports = require("node:readline");
-
 /***/ }),
 
 /***/ 81630:
@@ -1339,7 +1339,7 @@ module.exports = import("@huggingface/transformers");;
 var __webpack_require__ = require("../../../../../webpack-runtime.js");
 __webpack_require__.C(exports);
 var __webpack_exec__ = (moduleId) => (__webpack_require__(__webpack_require__.s = moduleId))
-var __webpack_exports__ = __webpack_require__.X(0, [2970,7377,1813,7981,4992,46], () => (__webpack_exec__(73809)));
+var __webpack_exports__ = __webpack_require__.X(0, [2970,7377,1813,7981,4992,46], () => (__webpack_exec__(80747)));
 module.exports = __webpack_exports__;
 
 })();