opencode-tbot 0.1.14 → 0.1.16

package/dist/plugin.js

@@ -1,4 +1,5 @@
 import { c as loadAppConfig, i as preparePluginConfiguration, o as OPENCODE_TBOT_VERSION } from "./assets/plugin-config-DA71_jD3.js";
+import { createRequire } from "node:module";
 import { mkdir, readFile, rename, stat, writeFile } from "node:fs/promises";
 import { basename, dirname, extname, isAbsolute, join } from "node:path";
 import { parse, printParseErrorCode } from "jsonc-parser";
@@ -6,6 +7,7 @@ import { z } from "zod";
 import { OpenRouter } from "@openrouter/sdk";
 import { createOpencodeClient } from "@opencode-ai/sdk/v2/client";
 import { randomUUID } from "node:crypto";
+import { spawn } from "node:child_process";
 import { run } from "@grammyjs/runner";
 import { Bot, InlineKeyboard } from "grammy";
 //#region src/infra/utils/redact.ts
@@ -1178,6 +1180,145 @@ var NOOP_FOREGROUND_SESSION_TRACKER = {
 	}
 };
 //#endregion
+//#region src/services/voice-transcription/audio-transcoder.ts
+var OPENROUTER_SUPPORTED_AUDIO_FORMATS = ["mp3", "wav"];
+var VoiceTranscodingFailedError = class extends Error {
+	data;
+	constructor(message) {
+		super(message);
+		this.name = "VoiceTranscodingFailedError";
+		this.data = { message };
+	}
+};
+var DEFAULT_TRANSCODE_TIMEOUT_MS = 15e3;
+var FfmpegAudioTranscoder = class {
+	ffmpegPath;
+	spawnProcess;
+	timeoutMs;
+	constructor(options) {
+		this.ffmpegPath = options.ffmpegPath?.trim() || null;
+		this.spawnProcess = options.spawnProcess ?? defaultSpawnProcess;
+		this.timeoutMs = options.timeoutMs ?? DEFAULT_TRANSCODE_TIMEOUT_MS;
+	}
+	async transcode(input) {
+		if (!this.ffmpegPath) throw new VoiceTranscodingFailedError(buildTranscodingMessage(input.sourceFormat, input.targetFormat, "Bundled ffmpeg is unavailable."));
+		if (input.targetFormat !== "wav") throw new VoiceTranscodingFailedError(buildTranscodingMessage(input.sourceFormat, input.targetFormat, `Unsupported transcode target: ${input.targetFormat}.`));
+		return {
+			data: await runFfmpegTranscode({
+				data: toUint8Array$1(input.data),
+				ffmpegPath: this.ffmpegPath,
+				filename: input.filename,
+				sourceFormat: input.sourceFormat,
+				spawnProcess: this.spawnProcess,
+				timeoutMs: this.timeoutMs,
+				targetFormat: input.targetFormat
+			}),
+			filename: replaceExtension(input.filename, ".wav"),
+			format: "wav",
+			mimeType: "audio/wav"
+		};
+	}
+};
+async function runFfmpegTranscode(input) {
+	return await new Promise((resolve, reject) => {
+		const child = input.spawnProcess(input.ffmpegPath, buildFfmpegArgs(input.targetFormat), {
+			stdio: [
+				"pipe",
+				"pipe",
+				"pipe"
+			],
+			windowsHide: true
+		});
+		const stdoutChunks = [];
+		const stderrChunks = [];
+		let settled = false;
+		let timedOut = false;
+		const timer = setTimeout(() => {
+			timedOut = true;
+			child.kill();
+		}, input.timeoutMs);
+		const cleanup = () => {
+			clearTimeout(timer);
+		};
+		const rejectOnce = (message) => {
+			if (settled) return;
+			settled = true;
+			cleanup();
+			reject(new VoiceTranscodingFailedError(buildTranscodingMessage(input.sourceFormat, input.targetFormat, message)));
+		};
+		const resolveOnce = (value) => {
+			if (settled) return;
+			settled = true;
+			cleanup();
+			resolve(value);
+		};
+		child.stdout.on("data", (chunk) => {
+			stdoutChunks.push(Buffer.isBuffer(chunk) ? chunk : Buffer.from(chunk));
+		});
+		child.stderr.on("data", (chunk) => {
+			stderrChunks.push(Buffer.isBuffer(chunk) ? chunk : Buffer.from(chunk));
+		});
+		child.once("error", (error) => {
+			rejectOnce(`Failed to start bundled ffmpeg: ${error.message}`);
+		});
+		child.once("close", (code, signal) => {
+			if (timedOut) {
+				rejectOnce(`Bundled ffmpeg timed out after ${input.timeoutMs} ms.`);
+				return;
+			}
+			if (code !== 0) {
+				rejectOnce(Buffer.concat(stderrChunks).toString("utf8").trim() || `Bundled ffmpeg exited with code ${code}${signal ? ` (${signal})` : ""}.`);
+				return;
+			}
+			const output = Buffer.concat(stdoutChunks);
+			if (output.length === 0) {
+				rejectOnce("Bundled ffmpeg returned empty audio output.");
+				return;
+			}
+			resolveOnce(new Uint8Array(output));
+		});
+		child.stdin.on("error", (error) => {
+			rejectOnce(`Failed to write audio data to bundled ffmpeg: ${error.message}`);
+		});
+		child.stdin.write(Buffer.from(input.data));
+		child.stdin.end();
+	});
+}
+function buildFfmpegArgs(targetFormat) {
+	if (targetFormat !== "wav") throw new Error(`Unsupported target format: ${targetFormat}`);
+	return [
+		"-hide_banner",
+		"-loglevel",
+		"error",
+		"-i",
+		"pipe:0",
+		"-f",
+		"wav",
+		"-acodec",
+		"pcm_s16le",
+		"-ac",
+		"1",
+		"-ar",
+		"16000",
+		"pipe:1"
+	];
+}
+function buildTranscodingMessage(sourceFormat, targetFormat, reason) {
+	return `Failed to transcode audio from ${sourceFormat} to ${targetFormat}. ${reason}`;
+}
+function replaceExtension(filename, nextExtension) {
+	const trimmedFilename = basename(filename).trim();
+	if (!trimmedFilename) return `telegram-voice${nextExtension}`;
+	const currentExtension = extname(trimmedFilename);
+	return currentExtension ? `${trimmedFilename.slice(0, -currentExtension.length)}${nextExtension}` : `${trimmedFilename}${nextExtension}`;
+}
+function toUint8Array$1(data) {
+	return data instanceof Uint8Array ? data : new Uint8Array(data);
+}
+function defaultSpawnProcess(command, args, options) {
+	return spawn(command, args, options);
+}
+//#endregion
 //#region src/services/voice-transcription/openrouter-voice.client.ts
 var VoiceTranscriptionNotConfiguredError = class extends Error {
 	data;
@@ -1215,11 +1356,16 @@ var DisabledVoiceTranscriptionClient = class {
 	}
 };
 var OpenRouterVoiceTranscriptionClient = class {
+	audioTranscoder;
 	model;
 	sdk;
 	timeoutMs;
 	transcriptionPrompt;
-	constructor(options, sdk) {
+	constructor(options, sdk, audioTranscoder = new FfmpegAudioTranscoder({
+		ffmpegPath: null,
+		timeoutMs: options.timeoutMs
+	})) {
+		this.audioTranscoder = audioTranscoder;
 		this.model = options.model;
 		this.sdk = sdk;
 		this.timeoutMs = options.timeoutMs;
@@ -1232,8 +1378,8 @@ var OpenRouterVoiceTranscriptionClient = class {
 		};
 	}
 	async transcribe(input) {
-		const format = resolveAudioFormat(input.filename, input.mimeType);
-		const audioData = toBase64(input.data);
+		const preparedAudio = await prepareAudioForOpenRouter(input, resolveAudioFormat(input.filename, input.mimeType), this.audioTranscoder);
+		const audioData = toBase64(preparedAudio.data);
 		const prompt = buildTranscriptionPrompt(this.transcriptionPrompt);
 		let response;
 		try {
@@ -1247,7 +1393,7 @@ var OpenRouterVoiceTranscriptionClient = class {
 						type: "input_audio",
 						inputAudio: {
 							data: audioData,
-							format
+							format: preparedAudio.format
 						}
 					}]
 				}],
@@ -1257,13 +1403,29 @@ var OpenRouterVoiceTranscriptionClient = class {
 			} }, { timeoutMs: this.timeoutMs });
 		} catch (error) {
 			throw new VoiceTranscriptionFailedError(buildTranscriptionErrorMessage(error, {
-				format,
+				format: preparedAudio.format,
 				model: this.model
 			}));
 		}
 		return { text: extractTranscript(response) };
 	}
 };
+async function prepareAudioForOpenRouter(input, sourceFormat, audioTranscoder) {
+	if (isOpenRouterSupportedAudioFormat(sourceFormat)) return {
+		data: toUint8Array(input.data),
+		format: sourceFormat
+	};
+	const transcoded = await audioTranscoder.transcode({
+		data: input.data,
+		filename: input.filename,
+		sourceFormat,
+		targetFormat: "wav"
+	});
+	return {
+		data: transcoded.data,
+		format: transcoded.format
+	};
+}
 var MIME_TYPE_FORMAT_MAP = {
 	"audio/aac": "aac",
 	"audio/aiff": "aiff",
@@ -1301,9 +1463,15 @@ function resolveAudioFormat(filename, mimeType) {
 	return "ogg";
 }
 function toBase64(data) {
-	const bytes = data instanceof Uint8Array ? data : new Uint8Array(data);
+	const bytes = toUint8Array(data);
 	return Buffer.from(bytes).toString("base64");
 }
+function toUint8Array(data) {
+	return data instanceof Uint8Array ? data : new Uint8Array(data);
+}
+function isOpenRouterSupportedAudioFormat(format) {
+	return OPENROUTER_SUPPORTED_AUDIO_FORMATS.includes(format);
+}
 function buildTranscriptionPrompt(transcriptionPrompt) {
 	const basePrompt = [
 		"Transcribe the provided audio verbatim.",
@@ -2174,6 +2342,7 @@ function resolveExtension(mimeType) {
 }
 //#endregion
 //#region src/app/container.ts
+var require = createRequire(import.meta.url);
 function createAppContainer(config, client) {
 	const logger = createOpenCodeAppLogger(client, { level: config.logLevel });
 	return createContainer(config, createOpenCodeClientFromSdkClient(client), logger);
@@ -2250,8 +2419,19 @@ function createVoiceTranscriptionClient(config) {
 	}, new OpenRouter({
 		apiKey: config.apiKey,
 		timeoutMs: config.timeoutMs
+	}), new FfmpegAudioTranscoder({
+		ffmpegPath: loadBundledFfmpegPath(),
+		timeoutMs: config.timeoutMs
 	})) : new DisabledVoiceTranscriptionClient();
 }
+function loadBundledFfmpegPath() {
+	try {
+		const ffmpegInstaller = require("@ffmpeg-installer/ffmpeg");
+		return typeof ffmpegInstaller.path === "string" && ffmpegInstaller.path.trim().length > 0 ? ffmpegInstaller.path : null;
+	} catch {
+		return null;
+	}
+}
 //#endregion
 //#region src/app/bootstrap.ts
 function bootstrapPluginApp(client, configSource = {}, options = {}) {
@@ -2484,6 +2664,7 @@ var EN_BOT_COPY = {
 		structuredOutput: "Structured output validation failed.",
 		voiceNotConfigured: "Voice transcription is not configured.",
 		voiceDownload: "Failed to download the Telegram voice file.",
+		voiceTranscoding: "Voice audio preprocessing failed.",
 		voiceTranscription: "Voice transcription failed.",
 		voiceEmpty: "Voice transcription returned empty text.",
 		voiceUnsupported: "Voice message file is too large or unsupported.",
@@ -2693,6 +2874,7 @@ var ZH_CN_BOT_COPY = {
 		structuredOutput: "结构化输出校验失败。",
 		voiceNotConfigured: "未配置语音转写服务。",
 		voiceDownload: "下载 Telegram 语音文件失败。",
+		voiceTranscoding: "语音转码失败。",
 		voiceTranscription: "语音转写失败。",
 		voiceEmpty: "语音转写结果为空。",
 		voiceUnsupported: "语音文件过大或不受支持。",
@@ -3049,6 +3231,10 @@ function normalizeError(error, copy) {
 		message: copy.errors.voiceDownload,
 		cause: extractMessage(error.data) ?? null
 	};
+	if (isNamedError(error, "VoiceTranscodingFailedError")) return {
+		message: copy.errors.voiceTranscoding,
+		cause: extractMessage(error.data) ?? null
+	};
 	if (isNamedError(error, "VoiceTranscriptionFailedError")) return {
 		message: copy.errors.voiceTranscription,
 		cause: extractMessage(error.data) ?? null
@@ -3306,9 +3492,9 @@ function splitStatusLines(text) {
 function formatHealthBadge(healthy, layout) {
 	return healthy ? "🟢" : layout.errorStatus;
 }
-function formatVoiceRecognitionBadge(status, layout) {
-	if (status.status === "configured") return status.model ? `\uD83D\uDFE2 ${layout.voiceRecognitionConfiguredLabel} (${status.model})` : `\uD83D\uDFE2 ${layout.voiceRecognitionConfiguredLabel}`;
-	return `\u26AA ${layout.voiceRecognitionNotConfiguredLabel}`;
+function formatVoiceRecognitionBadge(status, _layout) {
+	if (status.status === "configured") return status.model ? `\uD83D\uDFE2 (${status.model})` : "🟡";
+	return "⚪";
 }
 function formatLspStatusBadge(status) {
 	switch (status.status) {
@@ -3378,9 +3564,7 @@ function getStatusLayoutCopy(copy) {
 		rootLabel: "Root",
 		statusLabel: "Status",
 		tbotVersionLabel: "opencode-tbot Version",
-		voiceRecognitionConfiguredLabel: "configured",
 		voiceRecognitionLabel: "Voice Recognition",
-		voiceRecognitionNotConfiguredLabel: "not configured",
 		workspaceTitle: "📁 Workspace"
 	};
 	return {
@@ -3403,9 +3587,7 @@ function getStatusLayoutCopy(copy) {
 		rootLabel: "根目录",
 		statusLabel: "状态",
 		tbotVersionLabel: "opencode-tbot版本",
-		voiceRecognitionConfiguredLabel: "已配置",
 		voiceRecognitionLabel: "语音识别",
-		voiceRecognitionNotConfiguredLabel: "未配置",
 		workspaceTitle: "📁 工作区"
 	};
 }