@wirecat/cli-core 0.20.0 → 0.262.0

package/dist/speech/models.js

@@ -0,0 +1,130 @@
+import { CliError } from "@wirecat/cli-base";
+const huggingFace = (repository, commit) => (name) => `https://huggingface.co/csukuangfj/${repository}/resolve/${commit}/${name}`;
+const gigaam = huggingFace("sherpa-onnx-nemo-transducer-punct-giga-am-v3-russian-2025-12-16", "a6039be7cee829a9044a69ac0ebaf1c191217c97");
+const gigaamCtc = huggingFace("sherpa-onnx-nemo-ctc-punct-giga-am-v3-russian-2025-12-16", "4fb5407ff028a69fec516cdf4c10fac9ddea7c16");
+const parakeet = huggingFace("sherpa-onnx-nemo-parakeet-tdt-0.6b-v3-int8", "2bda32ec70b097a55adaa07d9a7173915b43cc78");
+export const MODELS = [
+    {
+        id: "gigaam-v3",
+        title: "GigaAM v3, with punctuation",
+        languages: "Russian — the best of the three for Russian",
+        featureDim: 64,
+        files: [
+            {
+                name: "encoder.int8.onnx",
+                url: gigaam("encoder.int8.onnx"),
+                sha256: "369f35a71bf288d3b8e0391fabd8dba5f2314088d440bca474056b7b4b6e66bf",
+                bytes: 224_570_820,
+            },
+            {
+                name: "decoder.onnx",
+                url: gigaam("decoder.onnx"),
+                sha256: "38fc7475443ea2a26f63211ca350f73ac50fff824ab7a3876ee2bd610c53bbc4",
+                bytes: 4_600_132,
+            },
+            {
+                name: "joiner.onnx",
+                url: gigaam("joiner.onnx"),
+                sha256: "602ff7017a93311aad34df1437c8d7f49911353c13d6eae7a6ee7b041339465c",
+                bytes: 2_712_896,
+            },
+            {
+                name: "tokens.txt",
+                url: gigaam("tokens.txt"),
+                sha256: "39abae20e692998290c574e606f11a9edef2902a1995463fcff63d1490cf22b7",
+                bytes: 13_354,
+            },
+        ],
+        config: (path) => ({
+            transducer: { encoder: path("encoder.int8.onnx"), decoder: path("decoder.onnx"), joiner: path("joiner.onnx") },
+            tokens: path("tokens.txt"),
+            modelType: "nemo_transducer",
+        }),
+    },
+    {
+        id: "gigaam-v3-ctc",
+        title: "GigaAM v3 CTC, with punctuation — a little faster",
+        languages: "Russian — a little faster, rougher with capital letters",
+        featureDim: 64,
+        files: [
+            {
+                name: "model.int8.onnx",
+                url: gigaamCtc("model.int8.onnx"),
+                sha256: "d5fea8df94263c285e54b21e5774b707c707192d3bdbeffd7b1eb07fb6743b35",
+                bytes: 224_893_661,
+            },
+            {
+                name: "tokens.txt",
+                url: gigaamCtc("tokens.txt"),
+                sha256: "142de7570b3de5b3035ce111a89c228e80e6085273731d944093ddf24fa539cd",
+                bytes: 2_007,
+            },
+        ],
+        config: (path) => ({ nemoCtc: { model: path("model.int8.onnx") }, tokens: path("tokens.txt") }),
+    },
+    {
+        id: "parakeet-v3",
+        title: "NVIDIA Parakeet TDT 0.6B v3",
+        /** The model card's list: huggingface.co/nvidia/parakeet-tdt-0.6b-v3, read 2026-09-24. */
+        languages: "25: bg cs da de el en es et fi fr hr hu it lt lv mt nl pl pt ro ru sk sl sv uk",
+        featureDim: 80,
+        files: [
+            {
+                name: "encoder.int8.onnx",
+                url: parakeet("encoder.int8.onnx"),
+                sha256: "acfc2b4456377e15d04f0243af540b7fe7c992f8d898d751cf134c3a55fd2247",
+                bytes: 652_184_281,
+            },
+            {
+                name: "decoder.int8.onnx",
+                url: parakeet("decoder.int8.onnx"),
+                sha256: "179e50c43d1a9de79c8a24149a2f9bac6eb5981823f2a2ed88d655b24248db4e",
+                bytes: 11_845_275,
+            },
+            {
+                name: "joiner.int8.onnx",
+                url: parakeet("joiner.int8.onnx"),
+                sha256: "3164c13fc2821009440d20fcb5fdc78bff28b4db2f8d0f0b329101719c0948b3",
+                bytes: 6_355_277,
+            },
+            {
+                name: "tokens.txt",
+                url: parakeet("tokens.txt"),
+                sha256: "d58544679ea4bc6ac563d1f545eb7d474bd6cfa467f0a6e2c1dc1c7d37e3c35d",
+                bytes: 93_939,
+            },
+        ],
+        config: (path) => ({
+            transducer: {
+                encoder: path("encoder.int8.onnx"),
+                decoder: path("decoder.int8.onnx"),
+                joiner: path("joiner.int8.onnx"),
+            },
+            tokens: path("tokens.txt"),
+            modelType: "nemo_transducer",
+        }),
+    },
+];
+/** Cuts speech at pauses. Every model needs it: five minutes in one piece crash the WebAssembly build. */
+export const VAD = {
+    name: "silero_vad.onnx",
+    url: "https://github.com/k2-fsa/sherpa-onnx/releases/download/asr-models/silero_vad.onnx",
+    sha256: "9e2449e1087496d8d4caba907f23e0bd3f78d91fa552479bb9c23ac09cbb1fd6",
+    bytes: 643_854,
+};
+/**
+ * Parakeet first: it hears 25 languages, and a messenger's CLI whose people mostly speak one language
+ * puts that language's model first instead (`Messenger.speechModels`) — max-cli puts GigaAM first.
+ */
+export const DEFAULT_ORDER = ["parakeet-v3", "gigaam-v3", "gigaam-v3-ctc"];
+/** Every model, in the CLI's order; those it did not name follow in the default order. */
+export const orderedModels = (order = DEFAULT_ORDER) => [...new Set([...order, ...DEFAULT_ORDER])].map(speechModel);
+export const findModel = (id) => MODELS.find((model) => model.id === id);
+export const speechModel = (id) => {
+    const model = findModel(id);
+    if (!model) {
+        throw new CliError("validation_error", `no speech model "${id}" — one of ${MODELS.map((known) => known.id).join(", ")}`);
+    }
+    return model;
+};
+//# sourceMappingURL=models.js.map

package/dist/speech/models.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"models.js","sourceRoot":"","sources":["../../src/speech/models.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,QAAQ,EAAE,MAAM,mBAAmB,CAAA;AA6B5C,MAAM,WAAW,GAAG,CAAC,UAAkB,EAAE,MAAc,EAAE,EAAE,CAAC,CAAC,IAAY,EAAE,EAAE,CAC3E,qCAAqC,UAAU,YAAY,MAAM,IAAI,IAAI,EAAE,CAAA;AAE7E,MAAM,MAAM,GAAG,WAAW,CACxB,iEAAiE,EACjE,0CAA0C,CAC3C,CAAA;AACD,MAAM,SAAS,GAAG,WAAW,CAC3B,0DAA0D,EAC1D,0CAA0C,CAC3C,CAAA;AACD,MAAM,QAAQ,GAAG,WAAW,CAAC,4CAA4C,EAAE,0CAA0C,CAAC,CAAA;AAEtH,MAAM,CAAC,MAAM,MAAM,GAAkB;IACnC;QACE,EAAE,EAAE,WAAW;QACf,KAAK,EAAE,6BAA6B;QACpC,SAAS,EAAE,6CAA6C;QACxD,UAAU,EAAE,EAAE;QACd,KAAK,EAAE;YACL;gBACE,IAAI,EAAE,mBAAmB;gBACzB,GAAG,EAAE,MAAM,CAAC,mBAAmB,CAAC;gBAChC,MAAM,EAAE,kEAAkE;gBAC1E,KAAK,EAAE,WAAW;aACnB;YACD;gBACE,IAAI,EAAE,cAAc;gBACpB,GAAG,EAAE,MAAM,CAAC,cAAc,CAAC;gBAC3B,MAAM,EAAE,kEAAkE;gBAC1E,KAAK,EAAE,SAAS;aACjB;YACD;gBACE,IAAI,EAAE,aAAa;gBACnB,GAAG,EAAE,MAAM,CAAC,aAAa,CAAC;gBAC1B,MAAM,EAAE,kEAAkE;gBAC1E,KAAK,EAAE,SAAS;aACjB;YACD;gBACE,IAAI,EAAE,YAAY;gBAClB,GAAG,EAAE,MAAM,CAAC,YAAY,CAAC;gBACzB,MAAM,EAAE,kEAAkE;gBAC1E,KAAK,EAAE,MAAM;aACd;SACF;QACD,MAAM,EAAE,CAAC,IAAI,EAAE,EAAE,CAAC,CAAC;YACjB,UAAU,EAAE,EAAE,OAAO,EAAE,IAAI,CAAC,mBAAmB,CAAC,EAAE,OAAO,EAAE,IAAI,CAAC,cAAc,CAAC,EAAE,MAAM,EAAE,IAAI,CAAC,aAAa,CAAC,EAAE;YAC9G,MAAM,EAAE,IAAI,CAAC,YAAY,CAAC;YAC1B,SAAS,EAAE,iBAAiB;SAC7B,CAAC;KACH;IACD;QACE,EAAE,EAAE,eAAe;QACnB,KAAK,EAAE,mDAAmD;QAC1D,SAAS,EAAE,yDAAyD;QACpE,UAAU,EAAE,EAAE;QACd,KAAK,EAAE;YACL;gBACE,IAAI,EAAE,iBAAiB;gBACvB,GAAG,EAAE,SAAS,CAAC,iBAAiB,CAAC;gBACjC,MAAM,EAAE,kEAAkE;gBAC1E,KAAK,EAAE,WAAW;aACnB;YACD;gBACE,IAAI,EAAE,YAAY;gBAClB,GAAG,EAAE,SAAS,CAAC,YAAY,CAAC;gBAC5B,MAAM,EAAE,kEAAkE;gBAC1E,KAAK,EAAE,KAAK;aACb;SACF;QACD,MAAM,EAAE,CAAC,IAAI,EAAE,EAAE,CAAC,CAAC,EAAE,OAAO,EAAE,EAAE,KAAK,EAAE,IAAI,CAAC,iBAAiB,CAAC,EAAE,EAAE,MAAM,EAAE,IAAI,CAAC,YAAY,CAAC,EAAE,CAAC;KAChG;IACD;QACE,EAAE,EAAE,aAAa;QACjB,KAAK,EAAE,6BAA6B;QACpC,0FAA0F;QAC1F,SAAS,EAAE,gFAAgF;QAC3F,UAAU,EAAE,EAAE;QACd,KAAK,EAAE;YACL;gBACE,IAAI,EAAE,mBAAmB;gBACzB,GAAG,EAAE,QAAQ,CAAC,mBAAmB,CAAC;gBAClC,MAAM,EAAE,kEAAkE;gBAC1E,KAAK,EAAE,WAAW;aACnB;YACD;gBACE,IAAI,EAAE,mBAAmB;gBACzB,GAAG,EAAE,QAAQ,CAAC,mBAAmB,CAAC;gBAClC,MAAM,EAAE,kEAAkE;gBAC1E,KAAK,EAAE,UAAU;aAClB;YACD;gBACE,IAAI,EAAE,kBAAkB;gBACxB,GAAG,EAAE,QAAQ,CAAC,kBAAkB,CAAC;gBACjC,MAAM,EAAE,kEAAkE;gBAC1E,KAAK,EAAE,SAAS;aACjB;YACD;gBACE,IAAI,EAAE,YAAY;gBAClB,GAAG,EAAE,QAAQ,CAAC,YAAY,CAAC;gBAC3B,MAAM,EAAE,kEAAkE;gBAC1E,KAAK,EAAE,MAAM;aACd;SACF;QACD,MAAM,EAAE,CAAC,IAAI,EAAE,EAAE,CAAC,CAAC;YACjB,UAAU,EAAE;gBACV,OAAO,EAAE,IAAI,CAAC,mBAAmB,CAAC;gBAClC,OAAO,EAAE,IAAI,CAAC,mBAAmB,CAAC;gBAClC,MAAM,EAAE,IAAI,CAAC,kBAAkB,CAAC;aACjC;YACD,MAAM,EAAE,IAAI,CAAC,YAAY,CAAC;YAC1B,SAAS,EAAE,iBAAiB;SAC7B,CAAC;KACH;CACF,CAAA;AAED,0GAA0G;AAC1G,MAAM,CAAC,MAAM,GAAG,GAAc;IAC5B,IAAI,EAAE,iBAAiB;IACvB,GAAG,EAAE,oFAAoF;IACzF,MAAM,EAAE,kEAAkE;IAC1E,KAAK,EAAE,OAAO;CACf,CAAA;AAED;;;GAGG;AACH,MAAM,CAAC,MAAM,aAAa,GAAG,CAAC,aAAa,EAAE,WAAW,EAAE,eAAe,CAAC,CAAA;AAE1E,0FAA0F;AAC1F,MAAM,CAAC,MAAM,aAAa,GAAG,CAAC,KAAK,GAAsB,aAAa,EAAiB,EAAE,CACvF,CAAC,GAAG,IAAI,GAAG,CAAC,CAAC,GAAG,KAAK,EAAE,GAAG,aAAa,CAAC,CAAC,CAAC,CAAC,GAAG,CAAC,WAAW,CAAC,CAAA;AAE7D,MAAM,CAAC,MAAM,SAAS,GAAG,CAAC,EAAU,EAA2B,EAAE,CAAC,MAAM,CAAC,IAAI,CAAC,CAAC,KAAK,EAAE,EAAE,CAAC,KAAK,CAAC,EAAE,KAAK,EAAE,CAAC,CAAA;AAEzG,MAAM,CAAC,MAAM,WAAW,GAAG,CAAC,EAAU,EAAe,EAAE;IACrD,MAAM,KAAK,GAAG,SAAS,CAAC,EAAE,CAAC,CAAA;IAC3B,IAAI,CAAC,KAAK,EAAE,CAAC;QACX,MAAM,IAAI,QAAQ,CAChB,kBAAkB,EAClB,oBAAoB,EAAE,cAAc,MAAM,CAAC,GAAG,CAAC,CAAC,KAAK,EAAE,EAAE,CAAC,KAAK,CAAC,EAAE,CAAC,CAAC,IAAI,CAAC,IAAI,CAAC,EAAE,CACjF,CAAA;IACH,CAAC;IACD,OAAO,KAAK,CAAA;AACd,CAAC,CAAA"}

package/dist/speech/ogg-limit.d.ts

@@ -0,0 +1,3 @@
+/** Validate supported packet framing before handing bytes to the decoder. */
+export declare const checkOggDuration: (bytes: Uint8Array) => void;
+//# sourceMappingURL=ogg-limit.d.ts.map

package/dist/speech/ogg-limit.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"ogg-limit.d.ts","sourceRoot":"","sources":["../../src/speech/ogg-limit.ts"],"names":[],"mappings":"AAKA,6EAA6E;AAC7E,eAAO,MAAM,gBAAgB,UAAW,UAAU,KAAG,IAwDpD,CAAA"}

package/dist/speech/ogg-limit.js

@@ -0,0 +1,69 @@
+import { CliError } from "@wirecat/cli-base";
+const invalid = () => new CliError("validation_error", "local transcription needs a complete mono or stereo Ogg Opus recording");
+/** Validate supported packet framing before handing bytes to the decoder. */
+export const checkOggDuration = (bytes) => {
+    const view = new DataView(bytes.buffer, bytes.byteOffset, bytes.byteLength);
+    let offset = 0;
+    let serial;
+    let packet = 0;
+    let length = 0;
+    const prefix = [];
+    while (offset < bytes.length) {
+        if (offset + 27 > bytes.length || view.getUint32(offset, false) !== 0x4f676753 || bytes[offset + 4] !== 0)
+            throw invalid();
+        const stream = view.getUint32(offset + 14, true);
+        if (serial !== undefined && serial !== stream)
+            throw invalid();
+        serial = stream;
+        const segments = bytes[offset + 26] ?? 0;
+        let data = offset + 27 + segments;
+        if (data > bytes.length || Boolean((bytes[offset + 5] ?? 0) & 1) !== length > 0)
+            throw invalid();
+        for (let index = 0; index < segments; index++) {
+            const size = bytes[offset + 27 + index] ?? 0;
+            if (data + size > bytes.length)
+                throw invalid();
+            const take = Math.min(size, 19 - prefix.length);
+            for (let i = 0; i < take; i++)
+                prefix.push(bytes[data + i] ?? 0);
+            length += size;
+            data += size;
+            if (size === 255)
+                continue;
+            if (packet === 0) {
+                if (length < 19 ||
+                    String.fromCharCode(...prefix.slice(0, 8)) !== "OpusHead" ||
+                    ![1, 2].includes(prefix[9] ?? 0) ||
+                    prefix[18] !== 0)
+                    throw invalid();
+            }
+            else if (packet === 1) {
+                if (length < 8 || String.fromCharCode(...prefix.slice(0, 8)) !== "OpusTags")
+                    throw invalid();
+            }
+            else {
+                if (length < 1)
+                    throw invalid();
+                const toc = prefix[0] ?? 0;
+                const config = toc >> 3;
+                const frame = config >= 16
+                    ? 120 << (config & 3)
+                    : config >= 12
+                        ? 480 << (config & 1)
+                        : (config & 3) === 3
+                            ? 2880
+                            : 480 << (config & 3);
+                const count = (toc & 3) === 0 ? 1 : (toc & 3) === 3 ? (length >= 2 ? (prefix[1] ?? 0) & 63 : 0) : 2;
+                if (count < 1 || frame * count > 5760)
+                    throw invalid();
+            }
+            packet++;
+            length = 0;
+            prefix.length = 0;
+        }
+        offset = data;
+    }
+    if (length !== 0 || packet < 3)
+        throw invalid();
+};
+//# sourceMappingURL=ogg-limit.js.map

package/dist/speech/ogg-limit.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"ogg-limit.js","sourceRoot":"","sources":["../../src/speech/ogg-limit.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,QAAQ,EAAE,MAAM,mBAAmB,CAAA;AAE5C,MAAM,OAAO,GAAG,GAAG,EAAE,CACnB,IAAI,QAAQ,CAAC,kBAAkB,EAAE,wEAAwE,CAAC,CAAA;AAE5G,6EAA6E;AAC7E,MAAM,CAAC,MAAM,gBAAgB,GAAG,CAAC,KAAiB,EAAQ,EAAE;IAC1D,MAAM,IAAI,GAAG,IAAI,QAAQ,CAAC,KAAK,CAAC,MAAM,EAAE,KAAK,CAAC,UAAU,EAAE,KAAK,CAAC,UAAU,CAAC,CAAA;IAC3E,IAAI,MAAM,GAAG,CAAC,CAAA;IACd,IAAI,MAA0B,CAAA;IAC9B,IAAI,MAAM,GAAG,CAAC,CAAA;IACd,IAAI,MAAM,GAAG,CAAC,CAAA;IACd,MAAM,MAAM,GAAa,EAAE,CAAA;IAC3B,OAAO,MAAM,GAAG,KAAK,CAAC,MAAM,EAAE,CAAC;QAC7B,IAAI,MAAM,GAAG,EAAE,GAAG,KAAK,CAAC,MAAM,IAAI,IAAI,CAAC,SAAS,CAAC,MAAM,EAAE,KAAK,CAAC,KAAK,UAAU,IAAI,KAAK,CAAC,MAAM,GAAG,CAAC,CAAC,KAAK,CAAC;YACvG,MAAM,OAAO,EAAE,CAAA;QACjB,MAAM,MAAM,GAAG,IAAI,CAAC,SAAS,CAAC,MAAM,GAAG,EAAE,EAAE,IAAI,CAAC,CAAA;QAChD,IAAI,MAAM,KAAK,SAAS,IAAI,MAAM,KAAK,MAAM;YAAE,MAAM,OAAO,EAAE,CAAA;QAC9D,MAAM,GAAG,MAAM,CAAA;QACf,MAAM,QAAQ,GAAG,KAAK,CAAC,MAAM,GAAG,EAAE,CAAC,IAAI,CAAC,CAAA;QACxC,IAAI,IAAI,GAAG,MAAM,GAAG,EAAE,GAAG,QAAQ,CAAA;QACjC,IAAI,IAAI,GAAG,KAAK,CAAC,MAAM,IAAI,OAAO,CAAC,CAAC,KAAK,CAAC,MAAM,GAAG,CAAC,CAAC,IAAI,CAAC,CAAC,GAAG,CAAC,CAAC,KAAK,MAAM,GAAG,CAAC;YAAE,MAAM,OAAO,EAAE,CAAA;QAChG,KAAK,IAAI,KAAK,GAAG,CAAC,EAAE,KAAK,GAAG,QAAQ,EAAE,KAAK,EAAE,EAAE,CAAC;YAC9C,MAAM,IAAI,GAAG,KAAK,CAAC,MAAM,GAAG,EAAE,GAAG,KAAK,CAAC,IAAI,CAAC,CAAA;YAC5C,IAAI,IAAI,GAAG,IAAI,GAAG,KAAK,CAAC,MAAM;gBAAE,MAAM,OAAO,EAAE,CAAA;YAC/C,MAAM,IAAI,GAAG,IAAI,CAAC,GAAG,CAAC,IAAI,EAAE,EAAE,GAAG,MAAM,CAAC,MAAM,CAAC,CAAA;YAC/C,KAAK,IAAI,CAAC,GAAG,CAAC,EAAE,CAAC,GAAG,IAAI,EAAE,CAAC,EAAE;gBAAE,MAAM,CAAC,IAAI,CAAC,KAAK,CAAC,IAAI,GAAG,CAAC,CAAC,IAAI,CAAC,CAAC,CAAA;YAChE,MAAM,IAAI,IAAI,CAAA;YACd,IAAI,IAAI,IAAI,CAAA;YACZ,IAAI,IAAI,KAAK,GAAG;gBAAE,SAAQ;YAC1B,IAAI,MAAM,KAAK,CAAC,EAAE,CAAC;gBACjB,IACE,MAAM,GAAG,EAAE;oBACX,MAAM,CAAC,YAAY,CAAC,GAAG,MAAM,CAAC,KAAK,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,KAAK,UAAU;oBACzD,CAAC,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,QAAQ,CAAC,MAAM,CAAC,CAAC,CAAC,IAAI,CAAC,CAAC;oBAChC,MAAM,CAAC,EAAE,CAAC,KAAK,CAAC;oBAEhB,MAAM,OAAO,EAAE,CAAA;YACnB,CAAC;iBAAM,IAAI,MAAM,KAAK,CAAC,EAAE,CAAC;gBACxB,IAAI,MAAM,GAAG,CAAC,IAAI,MAAM,CAAC,YAAY,CAAC,GAAG,MAAM,CAAC,KAAK,CAAC,CAAC,EAAE,CAAC,CAAC,CAAC,KAAK,UAAU;oBAAE,MAAM,OAAO,EAAE,CAAA;YAC9F,CAAC;iBAAM,CAAC;gBACN,IAAI,MAAM,GAAG,CAAC;oBAAE,MAAM,OAAO,EAAE,CAAA;gBAC/B,MAAM,GAAG,GAAG,MAAM,CAAC,CAAC,CAAC,IAAI,CAAC,CAAA;gBAC1B,MAAM,MAAM,GAAG,GAAG,IAAI,CAAC,CAAA;gBACvB,MAAM,KAAK,GACT,MAAM,IAAI,EAAE;oBACV,CAAC,CAAC,GAAG,IAAI,CAAC,MAAM,GAAG,CAAC,CAAC;oBACrB,CAAC,CAAC,MAAM,IAAI,EAAE;wBACZ,CAAC,CAAC,GAAG,IAAI,CAAC,MAAM,GAAG,CAAC,CAAC;wBACrB,CAAC,CAAC,CAAC,MAAM,GAAG,CAAC,CAAC,KAAK,CAAC;4BAClB,CAAC,CAAC,IAAI;4BACN,CAAC,CAAC,GAAG,IAAI,CAAC,MAAM,GAAG,CAAC,CAAC,CAAA;gBAC7B,MAAM,KAAK,GAAG,CAAC,GAAG,GAAG,CAAC,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,GAAG,GAAG,CAAC,CAAC,KAAK,CAAC,CAAC,CAAC,CAAC,CAAC,MAAM,IAAI,CAAC,CAAC,CAAC,CAAC,CAAC,MAAM,CAAC,CAAC,CAAC,IAAI,CAAC,CAAC,GAAG,EAAE,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAC,CAAA;gBACnG,IAAI,KAAK,GAAG,CAAC,IAAI,KAAK,GAAG,KAAK,GAAG,IAAI;oBAAE,MAAM,OAAO,EAAE,CAAA;YACxD,CAAC;YACD,MAAM,EAAE,CAAA;YACR,MAAM,GAAG,CAAC,CAAA;YACV,MAAM,CAAC,MAAM,GAAG,CAAC,CAAA;QACnB,CAAC;QACD,MAAM,GAAG,IAAI,CAAA;IACf,CAAC;IACD,IAAI,MAAM,KAAK,CAAC,IAAI,MAAM,GAAG,CAAC;QAAE,MAAM,OAAO,EAAE,CAAA;AACjD,CAAC,CAAA"}

package/dist/speech/recognize.d.ts

@@ -0,0 +1,66 @@
+import type { SpeechModel } from "./models.js";
+/** What transcription needs from a speech engine — small enough for a test to stand in for it. */
+export interface Recognizer {
+    recognize(pcm: Float32Array): string;
+    free(): void;
+}
+export declare const SAMPLE_RATE = 16000;
+/**
+ * Voice messages are Ogg Opus at 48 kHz (measured on a MAX one), and the models want 16 kHz. Averaging each three samples is ours rather than the engine's resampler, which announces
+ * itself on stderr on every call.
+ */
+export declare const toModelRate: (samples: Float32Array, rate: number) => Float32Array;
+export declare const decodeOgg: (bytes: Uint8Array) => Promise<{
+    samples: Float32Array;
+    rate: number;
+}>;
+interface SherpaStream {
+    acceptWaveform(rate: number, samples: Float32Array): void;
+    free(): void;
+}
+interface Sherpa {
+    createOfflineRecognizer(config: object): {
+        createStream(): SherpaStream;
+        decode(stream: SherpaStream): void;
+        getResult(stream: SherpaStream): {
+            text: string;
+        };
+        free(): void;
+    };
+    createVad(config: object): {
+        acceptWaveform(samples: Float32Array): void;
+        isEmpty(): boolean;
+        front(): {
+            samples: Float32Array;
+        };
+        pop(): void;
+        flush(): void;
+        reset(): void;
+        free(): void;
+    };
+}
+/** Where the voice detector cuts. Tested on a recording with a quiet stretch in the middle (`speech.test.ts`). */
+export declare const detector: (vadModel: string) => {
+    sileroVad: {
+        model: string;
+        threshold: number;
+        minSilenceDuration: number;
+        minSpeechDuration: number;
+        maxSpeechDuration: number;
+        windowSize: number;
+    };
+    sampleRate: number;
+    numThreads: number;
+    provider: string;
+    debug: number;
+    bufferSizeInSeconds: number;
+};
+/**
+ * The engine's WebAssembly writes its logs (a growing buffer says "Overflow! … No data loss!") with
+ * `printf`, which its file system sends straight to file descriptor 1 and so into `--json` output.
+ * `sherpa-onnx`'s own entry keeps that file system to itself; built here, descriptor 1 is pointed at 2.
+ */
+export declare const loadSherpa: () => Sherpa;
+export declare const openRecognizer: (model: SpeechModel, path: (name: string) => string, vadModel: string) => Recognizer;
+export {};
+//# sourceMappingURL=recognize.d.ts.map

package/dist/speech/recognize.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"recognize.d.ts","sourceRoot":"","sources":["../../src/speech/recognize.ts"],"names":[],"mappings":"AACA,OAAO,KAAK,EAAE,WAAW,EAAE,MAAM,aAAa,CAAA;AAG9C,kGAAkG;AAClG,MAAM,WAAW,UAAU;IACzB,SAAS,CAAC,GAAG,EAAE,YAAY,GAAG,MAAM,CAAA;IACpC,IAAI,IAAI,IAAI,CAAA;CACb;AAED,eAAO,MAAM,WAAW,QAAS,CAAA;AAQjC;;;GAGG;AACH,eAAO,MAAM,WAAW,YAAa,YAAY,QAAQ,MAAM,KAAG,YAYjE,CAAA;AAED,eAAO,MAAM,SAAS,UAAiB,UAAU,KAAG,OAAO,CAAC;IAAE,OAAO,EAAE,YAAY,CAAC;IAAC,IAAI,EAAE,MAAM,CAAA;CAAE,CAalG,CAAA;AAED,UAAU,YAAY;IACpB,cAAc,CAAC,IAAI,EAAE,MAAM,EAAE,OAAO,EAAE,YAAY,GAAG,IAAI,CAAA;IACzD,IAAI,IAAI,IAAI,CAAA;CACb;AACD,UAAU,MAAM;IACd,uBAAuB,CAAC,MAAM,EAAE,MAAM,GAAG;QACvC,YAAY,IAAI,YAAY,CAAA;QAC5B,MAAM,CAAC,MAAM,EAAE,YAAY,GAAG,IAAI,CAAA;QAClC,SAAS,CAAC,MAAM,EAAE,YAAY,GAAG;YAAE,IAAI,EAAE,MAAM,CAAA;SAAE,CAAA;QACjD,IAAI,IAAI,IAAI,CAAA;KACb,CAAA;IACD,SAAS,CAAC,MAAM,EAAE,MAAM,GAAG;QACzB,cAAc,CAAC,OAAO,EAAE,YAAY,GAAG,IAAI,CAAA;QAC3C,OAAO,IAAI,OAAO,CAAA;QAClB,KAAK,IAAI;YAAE,OAAO,EAAE,YAAY,CAAA;SAAE,CAAA;QAClC,GAAG,IAAI,IAAI,CAAA;QACX,KAAK,IAAI,IAAI,CAAA;QACb,KAAK,IAAI,IAAI,CAAA;QACb,IAAI,IAAI,IAAI,CAAA;KACb,CAAA;CACF;AAID,kHAAkH;AAClH,eAAO,MAAM,QAAQ,aAAc,MAAM;;;;;;;;;;;;;;CAgBvC,CAAA;AAUF;;;;GAIG;AACH,eAAO,MAAM,UAAU,QAAO,MAc7B,CAAA;AAED,eAAO,MAAM,cAAc,UAAW,WAAW,QAAQ,CAAC,IAAI,EAAE,MAAM,KAAK,MAAM,YAAY,MAAM,KAAG,UA4CrG,CAAA"}

package/dist/speech/recognize.js

@@ -0,0 +1,135 @@
+import { createRequire } from "node:module";
+import { checkOggDuration } from "./ogg-limit.js";
+export const SAMPLE_RATE = 16_000;
+/**
+ * Half a second of silence ahead of the speech. Without it the voice detector clipped a recording
+ * that starts talking at once: "Ничьих" came back as "чьих" (measured 2026-09-24).
+ */
+const LEAD = SAMPLE_RATE / 2;
+/**
+ * Voice messages are Ogg Opus at 48 kHz (measured on a MAX one), and the models want 16 kHz. Averaging each three samples is ours rather than the engine's resampler, which announces
+ * itself on stderr on every call.
+ */
+export const toModelRate = (samples, rate) => {
+    const step = rate / SAMPLE_RATE;
+    const length = Math.floor(samples.length / step);
+    const pcm = new Float32Array(LEAD + length);
+    for (let i = 0; i < length; i++) {
+        const from = Math.floor(i * step);
+        const to = Math.max(from + 1, Math.floor((i + 1) * step));
+        let sum = 0;
+        for (let j = from; j < to; j++)
+            sum += samples[j] ?? 0;
+        pcm[LEAD + i] = sum / (to - from);
+    }
+    return pcm;
+};
+export const decodeOgg = async (bytes) => {
+    checkOggDuration(bytes);
+    const { OggOpusDecoder } = await import("ogg-opus-decoder");
+    const decoder = new OggOpusDecoder();
+    await decoder.ready;
+    try {
+        const { channelData, sampleRate, samplesDecoded } = await decoder.decodeFile(bytes);
+        const samples = channelData[0];
+        if (!samples || samplesDecoded === 0)
+            throw new Error("no Ogg Opus audio in it");
+        return { samples, rate: sampleRate };
+    }
+    finally {
+        decoder.free();
+    }
+};
+const WINDOW = 512;
+/** Where the voice detector cuts. Tested on a recording with a quiet stretch in the middle (`speech.test.ts`). */
+export const detector = (vadModel) => ({
+    sileroVad: {
+        model: vadModel,
+        // At the default 0.5 a quietly spoken stretch in the middle of a voice message never counted as
+        // speech and was lost; 0.3 keeps it for both models and changes nothing on clear speech (126508, 2026-09-30).
+        threshold: 0.3,
+        minSilenceDuration: 0.5,
+        minSpeechDuration: 0.25,
+        maxSpeechDuration: 20,
+        windowSize: WINDOW,
+    },
+    sampleRate: SAMPLE_RATE,
+    numThreads: 1,
+    provider: "cpu",
+    debug: 0,
+    bufferSizeInSeconds: 30,
+});
+/**
+ * The model, and a voice detector that cuts the audio at pauses into pieces of at most 20 seconds.
+ * The cutting is not optional: five minutes in one piece crash the WebAssembly build (measured).
+ *
+ * `sherpa-onnx` is loaded here, on first use, so no other command pays for its 15 MB of WebAssembly.
+ */
+let loaded;
+/**
+ * The engine's WebAssembly writes its logs (a growing buffer says "Overflow! … No data loss!") with
+ * `printf`, which its file system sends straight to file descriptor 1 and so into `--json` output.
+ * `sherpa-onnx`'s own entry keeps that file system to itself; built here, descriptor 1 is pointed at 2.
+ */
+export const loadSherpa = () => {
+    if (loaded)
+        return loaded;
+    const load = createRequire(import.meta.url);
+    const wasm = {};
+    load("sherpa-onnx/sherpa-onnx-wasm-nodejs.js")(wasm);
+    const stdout = wasm.FS?.getStream(1);
+    if (stdout)
+        stdout.nfd = 2;
+    const asr = load("sherpa-onnx/sherpa-onnx-asr.js");
+    const vad = load("sherpa-onnx/sherpa-onnx-vad.js");
+    loaded = {
+        createOfflineRecognizer: (config) => new asr.OfflineRecognizer(config, wasm),
+        createVad: (config) => vad.createVad(wasm, config),
+    };
+    return loaded;
+};
+export const openRecognizer = (model, path, vadModel) => {
+    const sherpa = loadSherpa();
+    const engine = sherpa.createOfflineRecognizer({
+        featConfig: { sampleRate: SAMPLE_RATE, featureDim: model.featureDim },
+        modelConfig: { ...model.config(path), numThreads: 1, provider: "cpu", debug: 0 },
+        decodingMethod: "greedy_search",
+    });
+    const vad = sherpa.createVad(detector(vadModel));
+    const piece = (samples) => {
+        const stream = engine.createStream();
+        try {
+            stream.acceptWaveform(SAMPLE_RATE, samples);
+            engine.decode(stream);
+            return engine.getResult(stream).text.trim();
+        }
+        finally {
+            stream.free();
+        }
+    };
+    return {
+        recognize: (pcm) => {
+            // One recognizer can hear several recordings in a row; nothing of the last may leak into this one.
+            vad.reset();
+            const parts = [];
+            const drain = () => {
+                while (!vad.isEmpty()) {
+                    parts.push(piece(vad.front().samples));
+                    vad.pop();
+                }
+            };
+            for (let i = 0; i + WINDOW <= pcm.length; i += WINDOW) {
+                vad.acceptWaveform(pcm.subarray(i, i + WINDOW));
+                drain();
+            }
+            vad.flush();
+            drain();
+            return parts.filter(Boolean).join(" ");
+        },
+        free: () => {
+            vad.free();
+            engine.free();
+        },
+    };
+};
+//# sourceMappingURL=recognize.js.map

package/dist/speech/recognize.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"recognize.js","sourceRoot":"","sources":["../../src/speech/recognize.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,aAAa,EAAE,MAAM,aAAa,CAAA;AAE3C,OAAO,EAAE,gBAAgB,EAAE,MAAM,gBAAgB,CAAA;AAQjD,MAAM,CAAC,MAAM,WAAW,GAAG,MAAM,CAAA;AAEjC;;;GAGG;AACH,MAAM,IAAI,GAAG,WAAW,GAAG,CAAC,CAAA;AAE5B;;;GAGG;AACH,MAAM,CAAC,MAAM,WAAW,GAAG,CAAC,OAAqB,EAAE,IAAY,EAAgB,EAAE;IAC/E,MAAM,IAAI,GAAG,IAAI,GAAG,WAAW,CAAA;IAC/B,MAAM,MAAM,GAAG,IAAI,CAAC,KAAK,CAAC,OAAO,CAAC,MAAM,GAAG,IAAI,CAAC,CAAA;IAChD,MAAM,GAAG,GAAG,IAAI,YAAY,CAAC,IAAI,GAAG,MAAM,CAAC,CAAA;IAC3C,KAAK,IAAI,CAAC,GAAG,CAAC,EAAE,CAAC,GAAG,MAAM,EAAE,CAAC,EAAE,EAAE,CAAC;QAChC,MAAM,IAAI,GAAG,IAAI,CAAC,KAAK,CAAC,CAAC,GAAG,IAAI,CAAC,CAAA;QACjC,MAAM,EAAE,GAAG,IAAI,CAAC,GAAG,CAAC,IAAI,GAAG,CAAC,EAAE,IAAI,CAAC,KAAK,CAAC,CAAC,CAAC,GAAG,CAAC,CAAC,GAAG,IAAI,CAAC,CAAC,CAAA;QACzD,IAAI,GAAG,GAAG,CAAC,CAAA;QACX,KAAK,IAAI,CAAC,GAAG,IAAI,EAAE,CAAC,GAAG,EAAE,EAAE,CAAC,EAAE;YAAE,GAAG,IAAI,OAAO,CAAC,CAAC,CAAC,IAAI,CAAC,CAAA;QACtD,GAAG,CAAC,IAAI,GAAG,CAAC,CAAC,GAAG,GAAG,GAAG,CAAC,EAAE,GAAG,IAAI,CAAC,CAAA;IACnC,CAAC;IACD,OAAO,GAAG,CAAA;AACZ,CAAC,CAAA;AAED,MAAM,CAAC,MAAM,SAAS,GAAG,KAAK,EAAE,KAAiB,EAAoD,EAAE;IACrG,gBAAgB,CAAC,KAAK,CAAC,CAAA;IACvB,MAAM,EAAE,cAAc,EAAE,GAAG,MAAM,MAAM,CAAC,kBAAkB,CAAC,CAAA;IAC3D,MAAM,OAAO,GAAG,IAAI,cAAc,EAAE,CAAA;IACpC,MAAM,OAAO,CAAC,KAAK,CAAA;IACnB,IAAI,CAAC;QACH,MAAM,EAAE,WAAW,EAAE,UAAU,EAAE,cAAc,EAAE,GAAG,MAAM,OAAO,CAAC,UAAU,CAAC,KAAK,CAAC,CAAA;QACnF,MAAM,OAAO,GAAG,WAAW,CAAC,CAAC,CAAC,CAAA;QAC9B,IAAI,CAAC,OAAO,IAAI,cAAc,KAAK,CAAC;YAAE,MAAM,IAAI,KAAK,CAAC,yBAAyB,CAAC,CAAA;QAChF,OAAO,EAAE,OAAO,EAAE,IAAI,EAAE,UAAU,EAAE,CAAA;IACtC,CAAC;YAAS,CAAC;QACT,OAAO,CAAC,IAAI,EAAE,CAAA;IAChB,CAAC;AACH,CAAC,CAAA;AAwBD,MAAM,MAAM,GAAG,GAAG,CAAA;AAElB,kHAAkH;AAClH,MAAM,CAAC,MAAM,QAAQ,GAAG,CAAC,QAAgB,EAAE,EAAE,CAAC,CAAC;IAC7C,SAAS,EAAE;QACT,KAAK,EAAE,QAAQ;QACf,gGAAgG;QAChG,8GAA8G;QAC9G,SAAS,EAAE,GAAG;QACd,kBAAkB,EAAE,GAAG;QACvB,iBAAiB,EAAE,IAAI;QACvB,iBAAiB,EAAE,EAAE;QACrB,UAAU,EAAE,MAAM;KACnB;IACD,UAAU,EAAE,WAAW;IACvB,UAAU,EAAE,CAAC;IACb,QAAQ,EAAE,KAAK;IACf,KAAK,EAAE,CAAC;IACR,mBAAmB,EAAE,EAAE;CACxB,CAAC,CAAA;AAEF;;;;;GAKG;AACH,IAAI,MAA0B,CAAA;AAE9B;;;;GAIG;AACH,MAAM,CAAC,MAAM,UAAU,GAAG,GAAW,EAAE;IACrC,IAAI,MAAM;QAAE,OAAO,MAAM,CAAA;IACzB,MAAM,IAAI,GAAG,aAAa,CAAC,OAAO,IAAI,CAAC,GAAG,CAAC,CAAA;IAC3C,MAAM,IAAI,GAA+D,EAAE,CAAA;IAC3E,IAAI,CAAC,wCAAwC,CAAC,CAAC,IAAI,CAAC,CAAA;IACpD,MAAM,MAAM,GAAG,IAAI,CAAC,EAAE,EAAE,SAAS,CAAC,CAAC,CAAC,CAAA;IACpC,IAAI,MAAM;QAAE,MAAM,CAAC,GAAG,GAAG,CAAC,CAAA;IAC1B,MAAM,GAAG,GAAG,IAAI,CAAC,gCAAgC,CAAC,CAAA;IAClD,MAAM,GAAG,GAAG,IAAI,CAAC,gCAAgC,CAAC,CAAA;IAClD,MAAM,GAAG;QACP,uBAAuB,EAAE,CAAC,MAAM,EAAE,EAAE,CAAC,IAAI,GAAG,CAAC,iBAAiB,CAAC,MAAM,EAAE,IAAI,CAAC;QAC5E,SAAS,EAAE,CAAC,MAAM,EAAE,EAAE,CAAC,GAAG,CAAC,SAAS,CAAC,IAAI,EAAE,MAAM,CAAC;KACnD,CAAA;IACD,OAAO,MAAM,CAAA;AACf,CAAC,CAAA;AAED,MAAM,CAAC,MAAM,cAAc,GAAG,CAAC,KAAkB,EAAE,IAA8B,EAAE,QAAgB,EAAc,EAAE;IACjH,MAAM,MAAM,GAAG,UAAU,EAAE,CAAA;IAC3B,MAAM,MAAM,GAAG,MAAM,CAAC,uBAAuB,CAAC;QAC5C,UAAU,EAAE,EAAE,UAAU,EAAE,WAAW,EAAE,UAAU,EAAE,KAAK,CAAC,UAAU,EAAE;QACrE,WAAW,EAAE,EAAE,GAAG,KAAK,CAAC,MAAM,CAAC,IAAI,CAAC,EAAE,UAAU,EAAE,CAAC,EAAE,QAAQ,EAAE,KAAK,EAAE,KAAK,EAAE,CAAC,EAAE;QAChF,cAAc,EAAE,eAAe;KAChC,CAAC,CAAA;IACF,MAAM,GAAG,GAAG,MAAM,CAAC,SAAS,CAAC,QAAQ,CAAC,QAAQ,CAAC,CAAC,CAAA;IAEhD,MAAM,KAAK,GAAG,CAAC,OAAqB,EAAU,EAAE;QAC9C,MAAM,MAAM,GAAG,MAAM,CAAC,YAAY,EAAE,CAAA;QACpC,IAAI,CAAC;YACH,MAAM,CAAC,cAAc,CAAC,WAAW,EAAE,OAAO,CAAC,CAAA;YAC3C,MAAM,CAAC,MAAM,CAAC,MAAM,CAAC,CAAA;YACrB,OAAO,MAAM,CAAC,SAAS,CAAC,MAAM,CAAC,CAAC,IAAI,CAAC,IAAI,EAAE,CAAA;QAC7C,CAAC;gBAAS,CAAC;YACT,MAAM,CAAC,IAAI,EAAE,CAAA;QACf,CAAC;IACH,CAAC,CAAA;IAED,OAAO;QACL,SAAS,EAAE,CAAC,GAAG,EAAE,EAAE;YACjB,mGAAmG;YACnG,GAAG,CAAC,KAAK,EAAE,CAAA;YACX,MAAM,KAAK,GAAa,EAAE,CAAA;YAC1B,MAAM,KAAK,GAAG,GAAG,EAAE;gBACjB,OAAO,CAAC,GAAG,CAAC,OAAO,EAAE,EAAE,CAAC;oBACtB,KAAK,CAAC,IAAI,CAAC,KAAK,CAAC,GAAG,CAAC,KAAK,EAAE,CAAC,OAAO,CAAC,CAAC,CAAA;oBACtC,GAAG,CAAC,GAAG,EAAE,CAAA;gBACX,CAAC;YACH,CAAC,CAAA;YACD,KAAK,IAAI,CAAC,GAAG,CAAC,EAAE,CAAC,GAAG,MAAM,IAAI,GAAG,CAAC,MAAM,EAAE,CAAC,IAAI,MAAM,EAAE,CAAC;gBACtD,GAAG,CAAC,cAAc,CAAC,GAAG,CAAC,QAAQ,CAAC,CAAC,EAAE,CAAC,GAAG,MAAM,CAAC,CAAC,CAAA;gBAC/C,KAAK,EAAE,CAAA;YACT,CAAC;YACD,GAAG,CAAC,KAAK,EAAE,CAAA;YACX,KAAK,EAAE,CAAA;YACP,OAAO,KAAK,CAAC,MAAM,CAAC,OAAO,CAAC,CAAC,IAAI,CAAC,GAAG,CAAC,CAAA;QACxC,CAAC;QACD,IAAI,EAAE,GAAG,EAAE;YACT,GAAG,CAAC,IAAI,EAAE,CAAA;YACV,MAAM,CAAC,IAAI,EAAE,CAAA;QACf,CAAC;KACF,CAAA;AACH,CAAC,CAAA"}

package/dist/speech/transcribe.d.ts

@@ -0,0 +1,40 @@
+import { CliError } from "@wirecat/cli-base";
+import type { Messenger } from "../cli/messenger/context.js";
+import { type MessengerAdapter, type Transcript } from "../cli/messenger/port.js";
+import { type Settings } from "../cli/settings.js";
+import { type SpeechModel } from "./models.js";
+import type { Recognizer } from "./recognize.js";
+/** `messenger`: its own recognition (Telegram Premium). `local`: a model on this machine. `auto`: the first, then the second. */
+export declare const TRANSCRIBE_WITH: readonly ["auto", "messenger", "local"];
+export type TranscribeWith = (typeof TRANSCRIBE_WITH)[number];
+export interface Heard extends Transcript {
+    messageId: string;
+    /** The messenger's provider, or `local`. */
+    via: string;
+    /** The local model, when it was one. */
+    model?: string;
+}
+export type OpenRecognizer = (model: SpeechModel, path: (name: string) => string, vadModel: string) => Recognizer;
+export interface Choice {
+    with: TranscribeWith;
+    model: SpeechModel;
+    /** Where the models are, `modelsDirectory` for the command's environment. */
+    directory: string;
+    /** The local recognizer, where a test hands one in. */
+    open?: OpenRecognizer;
+}
+/** The flag, then the profile's `transcribeWith` and `speechModel`, then `auto` and the CLI's first model. */
+export declare const choose: (messenger: Messenger, settings: Pick<Settings, "configured" | "shared">, flags: {
+    local?: boolean;
+    model?: string;
+}, env?: NodeJS.ProcessEnv) => Choice;
+/** Never a download: a model is hundreds of MB, and whether to fetch one is the owner's call. */
+export declare const notDownloaded: (messenger: Messenger, model: SpeechModel) => CliError;
+/**
+ * Over an open connection: the messenger's own transcript, or the recording's bytes for a local model.
+ * The model runs after the connection closes — a minute of speech should not hold a socket open.
+ */
+export declare const hearOnline: (messenger: Messenger, connection: MessengerAdapter, chat: string, messageId: string, choice: Choice) => Promise<Heard | Uint8Array>;
+/** The local model, on the bytes `hearOnline` fetched. */
+export declare const hearLocally: (bytes: Uint8Array, messageId: string, { model, directory, open }: Choice) => Promise<Heard>;
+//# sourceMappingURL=transcribe.d.ts.map

package/dist/speech/transcribe.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"transcribe.d.ts","sourceRoot":"","sources":["../../src/speech/transcribe.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,QAAQ,EAAc,MAAM,mBAAmB,CAAA;AACxD,OAAO,KAAK,EAAE,SAAS,EAAE,MAAM,6BAA6B,CAAA;AAC5D,OAAO,EAAc,KAAK,gBAAgB,EAAmB,KAAK,UAAU,EAAE,MAAM,0BAA0B,CAAA;AAC9G,OAAO,EAAY,KAAK,QAAQ,EAAE,MAAM,oBAAoB,CAAA;AAE5D,OAAO,EAAiB,KAAK,WAAW,EAAe,MAAM,aAAa,CAAA;AAC1E,OAAO,KAAK,EAAE,UAAU,EAAE,MAAM,gBAAgB,CAAA;AAEhD,iIAAiI;AACjI,eAAO,MAAM,eAAe,YAAI,MAAM,EAAE,WAAW,EAAE,OAAO,CAAU,CAAA;AACtE,MAAM,MAAM,cAAc,GAAG,CAAC,OAAO,eAAe,CAAC,CAAC,MAAM,CAAC,CAAA;AAE7D,MAAM,WAAW,KAAM,SAAQ,UAAU;IACvC,SAAS,EAAE,MAAM,CAAA;IACjB,4CAA4C;IAC5C,GAAG,EAAE,MAAM,CAAA;IACX,wCAAwC;IACxC,KAAK,CAAC,EAAE,MAAM,CAAA;CACf;AAKD,MAAM,MAAM,cAAc,GAAG,CAAC,KAAK,EAAE,WAAW,EAAE,IAAI,EAAE,CAAC,IAAI,EAAE,MAAM,KAAK,MAAM,EAAE,QAAQ,EAAE,MAAM,KAAK,UAAU,CAAA;AAEjH,MAAM,WAAW,MAAM;IACrB,IAAI,EAAE,cAAc,CAAA;IACpB,KAAK,EAAE,WAAW,CAAA;IAClB,6EAA6E;IAC7E,SAAS,EAAE,MAAM,CAAA;IACjB,uDAAuD;IACvD,IAAI,CAAC,EAAE,cAAc,CAAA;CACtB;AAED,8GAA8G;AAC9G,eAAO,MAAM,MAAM,cACN,SAAS,YACV,IAAI,CAAC,QAAQ,EAAE,YAAY,GAAG,QAAQ,CAAC,SAC1C;IAAE,KAAK,CAAC,EAAE,OAAO,CAAC;IAAC,KAAK,CAAC,EAAE,MAAM,CAAA;CAAE,QACrC,MAAM,CAAC,UAAU,KACrB,MAUF,CAAA;AAED,iGAAiG;AACjG,eAAO,MAAM,aAAa,cAAe,SAAS,SAAS,WAAW,KAAG,QAKtE,CAAA;AAEH;;;GAGG;AACH,eAAO,MAAM,UAAU,cACV,SAAS,cACR,gBAAgB,QACtB,MAAM,aACD,MAAM,UACT,MAAM,KACb,OAAO,CAAC,KAAK,GAAG,UAAU,CAkB5B,CAAA;AAED,0DAA0D;AAC1D,eAAO,MAAM,WAAW,UACf,UAAU,aACN,MAAM,8BACW,MAAM,KACjC,OAAO,CAAC,KAAK,CAef,CAAA"}

package/dist/speech/transcribe.js

@@ -0,0 +1,80 @@
+import { CliError, isCliError } from "@wirecat/cli-base";
+import { capability } from "../cli/messenger/port.js";
+import { fromFile } from "../cli/settings.js";
+import { installedBytes, isInstalled, megabytes, modelPath, modelsDirectory, vadPath } from "./install.js";
+import { orderedModels, speechModel } from "./models.js";
+/** `messenger`: its own recognition (Telegram Premium). `local`: a model on this machine. `auto`: the first, then the second. */
+export const TRANSCRIBE_WITH = ["auto", "messenger", "local"];
+/** A voice message is read into memory whole; minutes of Opus are a few megabytes. */
+const LARGEST_VOICE = 32 * 1024 * 1024;
+/** The flag, then the profile's `transcribeWith` and `speechModel`, then `auto` and the CLI's first model. */
+export const choose = (messenger, settings, flags, env = process.env) => {
+    const model = flags.model ?? fromFile(settings, "speechModel", undefined).value;
+    return {
+        directory: modelsDirectory(env),
+        with: flags.local || flags.model !== undefined
+            ? "local"
+            : fromFile(settings, "transcribeWith", "auto").value,
+        model: model === undefined ? orderedModels(messenger.speechModels)[0] : speechModel(model),
+    };
+};
+/** Never a download: a model is hundreds of MB, and whether to fetch one is the owner's call. */
+export const notDownloaded = (messenger, model) => new CliError("not_found", `the speech model ${model.id} is not on this machine — the owner downloads it once with ` +
+    `\`${messenger.app.command} models audio download ${model.id}\` (${megabytes(installedBytes(model))})`);
+/**
+ * Over an open connection: the messenger's own transcript, or the recording's bytes for a local model.
+ * The model runs after the connection closes — a minute of speech should not hold a socket open.
+ */
+export const hearOnline = async (messenger, connection, chat, messageId, choice) => {
+    if (choice.with !== "local") {
+        const canAsk = typeof connection.transcribe === "function";
+        if (choice.with === "messenger" || canAsk) {
+            try {
+                const heard = await capability(connection, "transcribe", "transcribe voice messages")(chat, messageId);
+                return { messageId, ...heard, via: messenger.provider };
+            }
+            catch (error) {
+                // `auto` falls back only when the messenger refuses this account; any other failure is real.
+                if (choice.with === "messenger" || !(isCliError(error) && error.code === "permission_error"))
+                    throw error;
+            }
+        }
+    }
+    if (!isInstalled(choice.model, choice.directory))
+        throw notDownloaded(messenger, choice.model);
+    const { files } = await capability(connection, "download", "download attachments")(chat, messageId);
+    const voice = files.find((file) => file.kind === "voice" || file.mime === "audio/ogg");
+    if (!voice)
+        throw new CliError("validation_error", "that message is not a voice message");
+    return readAll(voice);
+};
+/** The local model, on the bytes `hearOnline` fetched. */
+export const hearLocally = async (bytes, messageId, { model, directory, open }) => {
+    const { decodeOgg, openRecognizer, toModelRate } = await import("./recognize.js");
+    const { samples, rate } = await decodeOgg(bytes);
+    const recognizer = (open ?? openRecognizer)(model, modelPath(directory, model), vadPath(directory));
+    try {
+        return {
+            messageId,
+            text: recognizer.recognize(toModelRate(samples, rate)),
+            pending: false,
+            via: "local",
+            model: model.id,
+        };
+    }
+    finally {
+        recognizer.free();
+    }
+};
+const readAll = async (file) => {
+    const chunks = [];
+    let total = 0;
+    for await (const chunk of file.bytes()) {
+        total += chunk.length;
+        if (total > LARGEST_VOICE)
+            throw new CliError("validation_error", "the voice message is larger than 32 MB");
+        chunks.push(chunk);
+    }
+    return new Uint8Array(Buffer.concat(chunks));
+};
+//# sourceMappingURL=transcribe.js.map

package/dist/speech/transcribe.js.map

@@ -0,0 +1 @@
+{"version":3,"file":"transcribe.js","sourceRoot":"","sources":["../../src/speech/transcribe.ts"],"names":[],"mappings":"AAAA,OAAO,EAAE,QAAQ,EAAE,UAAU,EAAE,MAAM,mBAAmB,CAAA;AAExD,OAAO,EAAE,UAAU,EAA2D,MAAM,0BAA0B,CAAA;AAC9G,OAAO,EAAE,QAAQ,EAAiB,MAAM,oBAAoB,CAAA;AAC5D,OAAO,EAAE,cAAc,EAAE,WAAW,EAAE,SAAS,EAAE,SAAS,EAAE,eAAe,EAAE,OAAO,EAAE,MAAM,cAAc,CAAA;AAC1G,OAAO,EAAE,aAAa,EAAoB,WAAW,EAAE,MAAM,aAAa,CAAA;AAG1E,iIAAiI;AACjI,MAAM,CAAC,MAAM,eAAe,GAAG,CAAC,MAAM,EAAE,WAAW,EAAE,OAAO,CAAU,CAAA;AAWtE,sFAAsF;AACtF,MAAM,aAAa,GAAG,EAAE,GAAG,IAAI,GAAG,IAAI,CAAA;AAatC,8GAA8G;AAC9G,MAAM,CAAC,MAAM,MAAM,GAAG,CACpB,SAAoB,EACpB,QAAiD,EACjD,KAA0C,EAC1C,GAAG,GAAsB,OAAO,CAAC,GAAG,EAC5B,EAAE;IACV,MAAM,KAAK,GAAG,KAAK,CAAC,KAAK,IAAI,QAAQ,CAAqB,QAAQ,EAAE,aAAa,EAAE,SAAS,CAAC,CAAC,KAAK,CAAA;IACnG,OAAO;QACL,SAAS,EAAE,eAAe,CAAC,GAAG,CAAC;QAC/B,IAAI,EACF,KAAK,CAAC,KAAK,IAAI,KAAK,CAAC,KAAK,KAAK,SAAS;YACtC,CAAC,CAAC,OAAO;YACT,CAAC,CAAC,QAAQ,CAAiB,QAAQ,EAAE,gBAAgB,EAAE,MAAM,CAAC,CAAC,KAAK;QACxE,KAAK,EAAE,KAAK,KAAK,SAAS,CAAC,CAAC,CAAE,aAAa,CAAC,SAAS,CAAC,YAAY,CAAC,CAAC,CAAC,CAAiB,CAAC,CAAC,CAAC,WAAW,CAAC,KAAK,CAAC;KAC5G,CAAA;AACH,CAAC,CAAA;AAED,iGAAiG;AACjG,MAAM,CAAC,MAAM,aAAa,GAAG,CAAC,SAAoB,EAAE,KAAkB,EAAY,EAAE,CAClF,IAAI,QAAQ,CACV,WAAW,EACX,oBAAoB,KAAK,CAAC,EAAE,6DAA6D;IACvF,KAAK,SAAS,CAAC,GAAG,CAAC,OAAO,0BAA0B,KAAK,CAAC,EAAE,OAAO,SAAS,CAAC,cAAc,CAAC,KAAK,CAAC,CAAC,GAAG,CACzG,CAAA;AAEH;;;GAGG;AACH,MAAM,CAAC,MAAM,UAAU,GAAG,KAAK,EAC7B,SAAoB,EACpB,UAA4B,EAC5B,IAAY,EACZ,SAAiB,EACjB,MAAc,EACe,EAAE;IAC/B,IAAI,MAAM,CAAC,IAAI,KAAK,OAAO,EAAE,CAAC;QAC5B,MAAM,MAAM,GAAG,OAAO,UAAU,CAAC,UAAU,KAAK,UAAU,CAAA;QAC1D,IAAI,MAAM,CAAC,IAAI,KAAK,WAAW,IAAI,MAAM,EAAE,CAAC;YAC1C,IAAI,CAAC;gBACH,MAAM,KAAK,GAAG,MAAM,UAAU,CAAC,UAAU,EAAE,YAAY,EAAE,2BAA2B,CAAC,CAAC,IAAI,EAAE,SAAS,CAAC,CAAA;gBACtG,OAAO,EAAE,SAAS,EAAE,GAAG,KAAK,EAAE,GAAG,EAAE,SAAS,CAAC,QAAQ,EAAE,CAAA;YACzD,CAAC;YAAC,OAAO,KAAK,EAAE,CAAC;gBACf,6FAA6F;gBAC7F,IAAI,MAAM,CAAC,IAAI,KAAK,WAAW,IAAI,CAAC,CAAC,UAAU,CAAC,KAAK,CAAC,IAAI,KAAK,CAAC,IAAI,KAAK,kBAAkB,CAAC;oBAAE,MAAM,KAAK,CAAA;YAC3G,CAAC;QACH,CAAC;IACH,CAAC;IACD,IAAI,CAAC,WAAW,CAAC,MAAM,CAAC,KAAK,EAAE,MAAM,CAAC,SAAS,CAAC;QAAE,MAAM,aAAa,CAAC,SAAS,EAAE,MAAM,CAAC,KAAK,CAAC,CAAA;IAC9F,MAAM,EAAE,KAAK,EAAE,GAAG,MAAM,UAAU,CAAC,UAAU,EAAE,UAAU,EAAE,sBAAsB,CAAC,CAAC,IAAI,EAAE,SAAS,CAAC,CAAA;IACnG,MAAM,KAAK,GAAG,KAAK,CAAC,IAAI,CAAC,CAAC,IAAI,EAAE,EAAE,CAAC,IAAI,CAAC,IAAI,KAAK,OAAO,IAAI,IAAI,CAAC,IAAI,KAAK,WAAW,CAAC,CAAA;IACtF,IAAI,CAAC,KAAK;QAAE,MAAM,IAAI,QAAQ,CAAC,kBAAkB,EAAE,qCAAqC,CAAC,CAAA;IACzF,OAAO,OAAO,CAAC,KAAK,CAAC,CAAA;AACvB,CAAC,CAAA;AAED,0DAA0D;AAC1D,MAAM,CAAC,MAAM,WAAW,GAAG,KAAK,EAC9B,KAAiB,EACjB,SAAiB,EACjB,EAAE,KAAK,EAAE,SAAS,EAAE,IAAI,EAAU,EAClB,EAAE;IAClB,MAAM,EAAE,SAAS,EAAE,cAAc,EAAE,WAAW,EAAE,GAAG,MAAM,MAAM,CAAC,gBAAgB,CAAC,CAAA;IACjF,MAAM,EAAE,OAAO,EAAE,IAAI,EAAE,GAAG,MAAM,SAAS,CAAC,KAAK,CAAC,CAAA;IAChD,MAAM,UAAU,GAAG,CAAC,IAAI,IAAI,cAAc,CAAC,CAAC,KAAK,EAAE,SAAS,CAAC,SAAS,EAAE,KAAK,CAAC,EAAE,OAAO,CAAC,SAAS,CAAC,CAAC,CAAA;IACnG,IAAI,CAAC;QACH,OAAO;YACL,SAAS;YACT,IAAI,EAAE,UAAU,CAAC,SAAS,CAAC,WAAW,CAAC,OAAO,EAAE,IAAI,CAAC,CAAC;YACtD,OAAO,EAAE,KAAK;YACd,GAAG,EAAE,OAAO;YACZ,KAAK,EAAE,KAAK,CAAC,EAAE;SAChB,CAAA;IACH,CAAC;YAAS,CAAC;QACT,UAAU,CAAC,IAAI,EAAE,CAAA;IACnB,CAAC;AACH,CAAC,CAAA;AAED,MAAM,OAAO,GAAG,KAAK,EAAE,IAAgB,EAAuB,EAAE;IAC9D,MAAM,MAAM,GAAiB,EAAE,CAAA;IAC/B,IAAI,KAAK,GAAG,CAAC,CAAA;IACb,IAAI,KAAK,EAAE,MAAM,KAAK,IAAI,IAAI,CAAC,KAAK,EAAE,EAAE,CAAC;QACvC,KAAK,IAAI,KAAK,CAAC,MAAM,CAAA;QACrB,IAAI,KAAK,GAAG,aAAa;YAAE,MAAM,IAAI,QAAQ,CAAC,kBAAkB,EAAE,wCAAwC,CAAC,CAAA;QAC3G,MAAM,CAAC,IAAI,CAAC,KAAK,CAAC,CAAA;IACpB,CAAC;IACD,OAAO,IAAI,UAAU,CAAC,MAAM,CAAC,MAAM,CAAC,MAAM,CAAC,CAAC,CAAA;AAC9C,CAAC,CAAA"}

package/dist/speech/wav.d.ts

@@ -0,0 +1,28 @@
+export interface WavAudioDescriptor {
+    id: string;
+    contentHash: string;
+    size: number;
+    durationMs: number;
+    format: "wav";
+    sampleRate: 16000 | 48000;
+    channels: 1 | 2;
+}
+export interface WavAudioOptions {
+    maxBytes: number;
+    maxDurationMs: number;
+    signal?: AbortSignal;
+}
+export interface WavAudioPlan {
+    readonly media: Readonly<WavAudioDescriptor>;
+}
+export interface WavPcmWindow {
+    startMs: number;
+    endMs: number;
+    samples: Float32Array;
+}
+export declare const preflightWavAudio: (path: string, input: WavAudioOptions) => Promise<WavAudioPlan>;
+export declare const openWavAudio: (plan: WavAudioPlan, input: WavAudioOptions) => Promise<{
+    windows(): AsyncIterable<WavPcmWindow>;
+    close(): Promise<void>;
+}>;
+//# sourceMappingURL=wav.d.ts.map

package/dist/speech/wav.d.ts.map

@@ -0,0 +1 @@
+{"version":3,"file":"wav.d.ts","sourceRoot":"","sources":["../../src/speech/wav.ts"],"names":[],"mappings":"AAMA,MAAM,WAAW,kBAAkB;IACjC,EAAE,EAAE,MAAM,CAAA;IACV,WAAW,EAAE,MAAM,CAAA;IACnB,IAAI,EAAE,MAAM,CAAA;IACZ,UAAU,EAAE,MAAM,CAAA;IAClB,MAAM,EAAE,KAAK,CAAA;IACb,UAAU,EAAE,KAAK,GAAG,KAAK,CAAA;IACzB,QAAQ,EAAE,CAAC,GAAG,CAAC,CAAA;CAChB;AACD,MAAM,WAAW,eAAe;IAC9B,QAAQ,EAAE,MAAM,CAAA;IAChB,aAAa,EAAE,MAAM,CAAA;IACrB,MAAM,CAAC,EAAE,WAAW,CAAA;CACrB;AACD,MAAM,WAAW,YAAY;IAC3B,QAAQ,CAAC,KAAK,EAAE,QAAQ,CAAC,kBAAkB,CAAC,CAAA;CAC7C;AACD,MAAM,WAAW,YAAY;IAC3B,OAAO,EAAE,MAAM,CAAA;IACf,KAAK,EAAE,MAAM,CAAA;IACb,OAAO,EAAE,YAAY,CAAA;CACtB;AAyHD,eAAO,MAAM,iBAAiB,SAAgB,MAAM,SAAS,eAAe,KAAG,OAAO,CAAC,YAAY,CA8BlG,CAAA;AACD,eAAO,MAAM,YAAY,SACjB,YAAY,SACX,eAAe,KACrB,OAAO,CAAC;IAAE,OAAO,IAAI,aAAa,CAAC,YAAY,CAAC,CAAC;IAAC,KAAK,IAAI,OAAO,CAAC,IAAI,CAAC,CAAA;CAAE,CA+D5E,CAAA"}