fix(tts): bound generated speech downloads · openclaw/openclaw@4438be7

Recent Commits to openclaw:main

vincentkoc · 2026-05-29 · via Recent Commits to openclaw:main

Original file line number	Diff line number	Diff line change
`@@ -98,6 +98,29 @@ describe("gradium speech provider", () => {`
`98`	`98`	`expect(result.audioBuffer).toEqual(audioData);`
`99`	`99`	`});`
`100`	`100`
	`101`	`+it("applies the configured media byte cap to synthesized audio", async () => {`
	`102`	`+vi.stubGlobal(`
	`103`	`+"fetch",`
	`104`	`+vi.fn().mockResolvedValue(new Response(new Uint8Array(2048), { status: 200 })),`
	`105`	`+);`
	`106`	`+`
	`107`	`+await expect(`
	`108`	`+provider.synthesize({`
	`109`	`+text: "OpenClaw test",`
	`110`	`+cfg: {`
	`111`	`+agents: {`
	`112`	`+defaults: {`
	`113`	`+mediaMaxMb: 0.001,`
	`114`	`+},`
	`115`	`+},`
	`116`	`+} as never,`
	`117`	`+providerConfig: { apiKey: "gsk_test123" },`
	`118`	`+target: "audio-file",`
	`119`	`+timeoutMs: 30_000,`
	`120`	`+}),`
	`121`	`+).rejects.toThrow("Gradium TTS audio response exceeds");`
	`122`	`+});`
	`123`	`+`
`101`	`124`	`it("uses ulaw_8000 for telephony synthesis", async () => {`
`102`	`125`	`const audioData = Buffer.from("ulaw-audio-data");`
`103`	`126`	`const fetchMock = vi.fn().mockResolvedValue(new Response(audioData, { status: 200 }));`

Original file line number	Diff line number	Diff line change
`@@ -8,6 +8,8 @@ import { asObject, trimToUndefined } from "openclaw/plugin-sdk/speech";`
`8`	`8`	`import { DEFAULT_GRADIUM_VOICE_ID, GRADIUM_VOICES, normalizeGradiumBaseUrl } from "./shared.js";`
`9`	`9`	`import { gradiumTTS } from "./tts.js";`
`10`	`10`
	`11`	`+const DEFAULT_GENERATED_AUDIO_MAX_BYTES = 16 * 1024 * 1024;`
	`12`	`+`
`11`	`13`	`type GradiumProviderConfig = {`
`12`	`14`	`apiKey?: string;`
`13`	`15`	`baseUrl: string;`
`@@ -36,6 +38,16 @@ function readGradiumProviderConfig(config: SpeechProviderConfig): GradiumProvide`
`36`	`38`	`};`
`37`	`39`	`}`
`38`	`40`
	`41`	`+function resolveGeneratedAudioMaxBytes(req: {`
	`42`	`+cfg: { agents?: { defaults?: { mediaMaxMb?: number } } };`
	`43`	`+}): number {`
	`44`	`+const configured = req.cfg.agents?.defaults?.mediaMaxMb;`
	`45`	`+if (typeof configured === "number" && Number.isFinite(configured) && configured > 0) {`
	`46`	`+return Math.floor(configured * 1024 * 1024);`
	`47`	`+}`
	`48`	`+return DEFAULT_GENERATED_AUDIO_MAX_BYTES;`
	`49`	`+}`
	`50`	`+`
`39`	`51`	`function parseDirectiveToken(ctx: SpeechDirectiveTokenParseContext): {`
`40`	`52`	`handled: boolean;`
`41`	`53`	`overrides?: Record<string, unknown>;`
`@@ -86,6 +98,7 @@ export function buildGradiumSpeechProvider(): SpeechProviderPlugin {`
`86`	`98`	`voiceId: trimToUndefined(overrides.voiceId) ?? config.voiceId,`
`87`	`99`	`outputFormat,`
`88`	`100`	`timeoutMs: req.timeoutMs,`
	`101`	`+maxBytes: resolveGeneratedAudioMaxBytes(req),`
`89`	`102`	`});`
`90`	`103`	`return {`
`91`	`104`	`audioBuffer,`
`@@ -110,6 +123,7 @@ export function buildGradiumSpeechProvider(): SpeechProviderPlugin {`
`110`	`123`	`voiceId: trimToUndefined(overrides.voiceId) ?? config.voiceId,`
`111`	`124`	`outputFormat,`
`112`	`125`	`timeoutMs: req.timeoutMs,`
	`126`	`+maxBytes: resolveGeneratedAudioMaxBytes(req),`
`113`	`127`	`});`
`114`	`128`	`return { audioBuffer, outputFormat, sampleRate };`
`115`	`129`	`},`

Original file line number	Diff line number	Diff line change
`@@ -28,6 +28,14 @@ describe("gradium tts diagnostics", () => {`
`28`	`28`	`};`
`29`	`29`	`}`
`30`	`30`
	`31`	`+function createStreamingAudioResponse(params: {`
	`32`	`+chunkCount: number;`
	`33`	`+chunkSize: number;`
	`34`	`+byte: number;`
	`35`	`+}): { response: Response; getReadCount: () => number } {`
	`36`	`+return createStreamingErrorResponse({ ...params, status: 200 });`
	`37`	`+}`
	`38`	`+`
`31`	`39`	`afterEach(() => {`
`32`	`40`	`vi.unstubAllGlobals();`
`33`	`41`	`vi.restoreAllMocks();`
`@@ -134,4 +142,27 @@ describe("gradium tts diagnostics", () => {`
`134`	`142`	`});`
`135`	`143`	`expect(result).toEqual(audioData);`
`136`	`144`	`});`
	`145`	`+`
	`146`	`+it("caps streamed audio responses instead of buffering oversized TTS output", async () => {`
	`147`	`+const streamed = createStreamingAudioResponse({`
	`148`	`+chunkCount: 20,`
	`149`	`+chunkSize: 1024,`
	`150`	`+byte: 121,`
	`151`	`+});`
	`152`	`+vi.stubGlobal("fetch", vi.fn().mockResolvedValue(streamed.response));`
	`153`	`+`
	`154`	`+await expect(`
	`155`	`+gradiumTTS({`
	`156`	`+text: "hello",`
	`157`	`+apiKey: "test-key",`
	`158`	`+baseUrl: "https://api.gradium.ai",`
	`159`	`+voiceId: "YTpq7expH9539ERJ",`
	`160`	`+outputFormat: "wav",`
	`161`	`+timeoutMs: 5_000,`
	`162`	`+maxBytes: 2048,`
	`163`	`+}),`
	`164`	`+).rejects.toThrow("Gradium TTS audio response exceeds 2048 bytes");`
	`165`	`+`
	`166`	`+expect(streamed.getReadCount()).toBeLessThan(20);`
	`167`	`+});`
`137`	`168`	`});`

Original file line number	Diff line number	Diff line change
`@@ -1,16 +1,28 @@`
`1`	`1`	`import { assertOkOrThrowProviderError } from "openclaw/plugin-sdk/provider-http";`
	`2`	`+import { readResponseWithLimit } from "openclaw/plugin-sdk/response-limit-runtime";`
`2`	`3`	`import { fetchWithSsrFGuard } from "openclaw/plugin-sdk/ssrf-runtime";`
`3`	`4`	`import { normalizeGradiumBaseUrl } from "./shared.js";`
`4`	`5`
	`6`	`+const DEFAULT_TTS_MAX_BYTES = 16 * 1024 * 1024;`
	`7`	`+`
`5`	`8`	`export async function gradiumTTS(params: {`
`6`	`9`	`text: string;`
`7`	`10`	`apiKey: string;`
`8`	`11`	`baseUrl: string;`
`9`	`12`	`voiceId: string;`
`10`	`13`	`outputFormat: "wav" \| "opus" \| "ulaw_8000" \| "pcm" \| "pcm_24000" \| "alaw_8000";`
`11`	`14`	`timeoutMs: number;`
	`15`	`+maxBytes?: number;`
`12`	`16`	`}): Promise<Buffer> {`
`13`		`-const { text, apiKey, baseUrl, voiceId, outputFormat, timeoutMs } = params;`
	`17`	`+const {`
	`18`	`+ text,`
	`19`	`+ apiKey,`
	`20`	`+ baseUrl,`
	`21`	`+ voiceId,`
	`22`	`+ outputFormat,`
	`23`	`+ timeoutMs,`
	`24`	`+ maxBytes = DEFAULT_TTS_MAX_BYTES,`
	`25`	`+} = params;`
`14`	`26`	`const normalizedBaseUrl = normalizeGradiumBaseUrl(baseUrl);`
`15`	`27`	const url = `${normalizedBaseUrl}/api/post/speech/tts`;
`16`	`28`	`const hostname = new URL(normalizedBaseUrl).hostname;`
`@@ -39,7 +51,10 @@ export async function gradiumTTS(params: {`
`39`	`51`	`try {`
`40`	`52`	`await assertOkOrThrowProviderError(response, "Gradium API error");`
`41`	`53`
`42`		`-return Buffer.from(await response.arrayBuffer());`
	`54`	`+return await readResponseWithLimit(response, maxBytes, {`
	`55`	`+onOverflow: ({ maxBytes }) =>`
	`56`	+new Error(`Gradium TTS audio response exceeds ${maxBytes} bytes`),
	`57`	`+});`
`43`	`58`	`} finally {`
`44`	`59`	`await release();`
`45`	`60`	`}`

Original file line number	Diff line number	Diff line change
`@@ -291,6 +291,33 @@ describe("buildOpenAISpeechProvider", () => {`
`291`	`291`	`expect(result.voiceCompatible).toBe(false);`
`292`	`292`	`});`
`293`	`293`
	`294`	`+it("applies the configured media byte cap to synthesized audio", async () => {`
	`295`	`+const provider = buildOpenAISpeechProvider();`
	`296`	`+globalThis.fetch = vi.fn(`
	`297`	`+async () => new Response(new Uint8Array(2048), { status: 200 }),`
	`298`	`+) as unknown as typeof fetch;`
	`299`	`+`
	`300`	`+await expect(`
	`301`	`+provider.synthesize({`
	`302`	`+text: "hello",`
	`303`	`+cfg: {`
	`304`	`+agents: {`
	`305`	`+defaults: {`
	`306`	`+mediaMaxMb: 0.001,`
	`307`	`+},`
	`308`	`+},`
	`309`	`+} as never,`
	`310`	`+providerConfig: {`
	`311`	`+apiKey: "sk-test",`
	`312`	`+model: "gpt-4o-mini-tts",`
	`313`	`+voice: "alloy",`
	`314`	`+},`
	`315`	`+target: "audio-file",`
	`316`	`+timeoutMs: 1_000,`
	`317`	`+}),`
	`318`	`+).rejects.toThrow("OpenAI TTS audio response exceeds");`
	`319`	`+});`
	`320`	`+`
`294`	`321`	`it("applies provider overrides to telephony synthesis", async () => {`
`295`	`322`	`const provider = buildOpenAISpeechProvider();`
`296`	`323`	`const fetchMock = vi.fn(async (_url: string, init?: RequestInit) => {`

Original file line number	Diff line number	Diff line change
`@@ -26,6 +26,7 @@ import {`
`26`	`26`	`} from "./tts.js";`
`27`	`27`
`28`	`28`	`const OPENAI_SPEECH_RESPONSE_FORMATS = ["mp3", "opus", "wav"] as const;`
	`29`	`+const DEFAULT_GENERATED_AUDIO_MAX_BYTES = 16 * 1024 * 1024;`
`29`	`30`
`30`	`31`	`type OpenAiSpeechResponseFormat = (typeof OPENAI_SPEECH_RESPONSE_FORMATS)[number];`
`31`	`32`
`@@ -174,6 +175,16 @@ function readOpenAIOverrides(`
`174`	`175`	`};`
`175`	`176`	`}`
`176`	`177`
	`178`	`+function resolveGeneratedAudioMaxBytes(req: {`
	`179`	`+cfg: { agents?: { defaults?: { mediaMaxMb?: number } } };`
	`180`	`+}): number {`
	`181`	`+const configured = req.cfg.agents?.defaults?.mediaMaxMb;`
	`182`	`+if (typeof configured === "number" && Number.isFinite(configured) && configured > 0) {`
	`183`	`+return Math.floor(configured * 1024 * 1024);`
	`184`	`+}`
	`185`	`+return DEFAULT_GENERATED_AUDIO_MAX_BYTES;`
	`186`	`+}`
	`187`	`+`
`177`	`188`	`function renderOpenAITtsPersonaInstructions(req: {`
`178`	`189`	`label?: string;`
`179`	`190`	`prompt?: {`
`@@ -328,6 +339,7 @@ export function buildOpenAISpeechProvider(): SpeechProviderPlugin {`
`328`	`339`	`responseFormat,`
`329`	`340`	`extraBody: config.extraBody,`
`330`	`341`	`timeoutMs: req.timeoutMs,`
	`342`	`+maxBytes: resolveGeneratedAudioMaxBytes(req),`
`331`	`343`	`});`
`332`	`344`	`return {`
`333`	`345`	`audioBuffer,`
`@@ -356,6 +368,7 @@ export function buildOpenAISpeechProvider(): SpeechProviderPlugin {`
`356`	`368`	`responseFormat: outputFormat,`
`357`	`369`	`extraBody: config.extraBody,`
`358`	`370`	`timeoutMs: req.timeoutMs,`
	`371`	`+maxBytes: resolveGeneratedAudioMaxBytes(req),`
`359`	`372`	`});`
`360`	`373`	`return { audioBuffer, outputFormat, sampleRate };`
`361`	`374`	`},`

Original file line number	Diff line number	Diff line change
`@@ -322,6 +322,32 @@ describe("openai tts", () => {`
`322`	`322`	`).rejects.toThrow("OpenAI TTS API error (503): temporary upstream outage");`
`323`	`323`	`});`
`324`	`324`
	`325`	`+it("caps streamed audio responses instead of buffering oversized TTS output", async () => {`
	`326`	`+const streamed = createStreamingErrorResponse({`
	`327`	`+status: 200,`
	`328`	`+chunkCount: 20,`
	`329`	`+chunkSize: 1024,`
	`330`	`+byte: 121,`
	`331`	`+});`
	`332`	`+const fetchMock = vi.fn(async () => streamed.response);`
	`333`	`+globalThis.fetch = fetchMock as unknown as typeof fetch;`
	`334`	`+`
	`335`	`+await expect(`
	`336`	`+openaiTTS({`
	`337`	`+text: "hello",`
	`338`	`+apiKey: "test-key",`
	`339`	`+baseUrl: "https://api.openai.com/v1",`
	`340`	`+model: "gpt-4o-mini-tts",`
	`341`	`+voice: "alloy",`
	`342`	`+responseFormat: "mp3",`
	`343`	`+timeoutMs: 5_000,`
	`344`	`+maxBytes: 2048,`
	`345`	`+}),`
	`346`	`+).rejects.toThrow("OpenAI TTS audio response exceeds 2048 bytes");`
	`347`	`+`
	`348`	`+expect(streamed.getReadCount()).toBeLessThan(20);`
	`349`	`+});`
	`350`	`+`
`325`	`351`	`it("caps streamed non-JSON error reads instead of consuming full response bodies", async () => {`
`326`	`352`	`const streamed = createStreamingErrorResponse({`
`327`	`353`	`status: 503,`

Original file line number	Diff line number	Diff line change
`@@ -6,12 +6,14 @@ import {`
`6`	`6`	`captureHttpExchange,`
`7`	`7`	`isDebugProxyGlobalFetchPatchInstalled,`
`8`	`8`	`} from "openclaw/plugin-sdk/proxy-capture";`
	`9`	`+import { readResponseWithLimit } from "openclaw/plugin-sdk/response-limit-runtime";`
`9`	`10`	`import {`
`10`	`11`	`fetchWithSsrFGuard,`
`11`	`12`	`ssrfPolicyFromHttpBaseUrlAllowedHostname,`
`12`	`13`	`} from "openclaw/plugin-sdk/ssrf-runtime";`
`13`	`14`
`14`	`15`	`export const DEFAULT_OPENAI_BASE_URL = "https://api.openai.com/v1";`
	`16`	`+const DEFAULT_TTS_MAX_BYTES = 16 * 1024 * 1024;`
`15`	`17`
`16`	`18`	`export const OPENAI_TTS_MODELS = ["gpt-4o-mini-tts", "tts-1", "tts-1-hd"] as const;`
`17`	`19`
`@@ -100,6 +102,7 @@ export async function openaiTTS(params: {`
`100`	`102`	`responseFormat: "mp3" \| "opus" \| "pcm" \| "wav";`
`101`	`103`	`extraBody?: Record<string, unknown>;`
`102`	`104`	`timeoutMs: number;`
	`105`	`+maxBytes?: number;`
`103`	`106`	`}): Promise<Buffer> {`
`104`	`107`	`const {`
`105`	`108`	`text,`
`@@ -112,6 +115,7 @@ export async function openaiTTS(params: {`
`112`	`115`	`responseFormat,`
`113`	`116`	`extraBody,`
`114`	`117`	`timeoutMs,`
	`118`	`+ maxBytes = DEFAULT_TTS_MAX_BYTES,`
`115`	`119`	`} = params;`
`116`	`120`	`const effectiveInstructions = resolveOpenAITtsInstructions(model, instructions, baseUrl);`
`117`	`121`
`@@ -177,7 +181,10 @@ export async function openaiTTS(params: {`
`177`	`181`
`178`	`182`	`await assertOkOrThrowProviderError(response, "OpenAI TTS API error");`
`179`	`183`
`180`		`-return Buffer.from(await response.arrayBuffer());`
	`184`	`+return await readResponseWithLimit(response, maxBytes, {`
	`185`	`+onOverflow: ({ maxBytes }) =>`
	`186`	+new Error(`OpenAI TTS audio response exceeds ${maxBytes} bytes`),
	`187`	`+});`
`181`	`188`	`} finally {`
`182`	`189`	`await release();`
`183`	`190`	`}`

Original file line number	Diff line number	Diff line change
`@@ -37,6 +37,7 @@ function requireLastTtsCall(): {`
`37`	`37`	`language?: string;`
`38`	`38`	`speed?: number;`
`39`	`39`	`responseFormat?: string;`
	`40`	`+maxBytes?: number;`
`40`	`41`	`} {`
`41`	`42`	`const params = (xaiTTSMock.mock.calls as unknown as Array<[unknown]>).at(-1)?.[0] as`
`42`	`43`	`\| {`
`@@ -47,6 +48,7 @@ function requireLastTtsCall(): {`
`47`	`48`	`language?: string;`
`48`	`49`	`speed?: number;`
`49`	`50`	`responseFormat?: string;`
	`51`	`+maxBytes?: number;`
`50`	`52`	`}`
`51`	`53`	`\| undefined;`
`52`	`54`	`if (!params) {`
`@@ -68,7 +70,13 @@ describe("xai speech provider", () => {`
`68`	`70`	`const provider = buildXaiSpeechProvider();`
`69`	`71`	`const result = await provider.synthesize({`
`70`	`72`	`text: "hello",`
`71`		`-cfg: {},`
	`73`	`+cfg: {`
	`74`	`+agents: {`
	`75`	`+defaults: {`
	`76`	`+mediaMaxMb: 2,`
	`77`	`+},`
	`78`	`+},`
	`79`	`+},`
`72`	`80`	`providerConfig: {`
`73`	`81`	`apiKey: "xai-key",`
`74`	`82`	`voiceId: "eve",`
`@@ -87,6 +95,7 @@ describe("xai speech provider", () => {`
`87`	`95`	`expect(tts.baseUrl).toBe("https://api.x.ai/v1");`
`88`	`96`	`expect(tts.voiceId).toBe("eve");`
`89`	`97`	`expect(tts.responseFormat).toBe("mp3");`
	`98`	`+expect(tts.maxBytes).toBe(2 * 1024 * 1024);`
`90`	`99`	`});`
`91`	`100`
`92`	`101`	`it("honors configured response formats", async () => {`

Original file line number	Diff line number	Diff line change
`@@ -26,6 +26,7 @@ import {`
`26`	`26`	`} from "./tts.js";`
`27`	`27`
`28`	`28`	`const XAI_SPEECH_RESPONSE_FORMATS = ["mp3", "wav", "pcm", "mulaw", "alaw"] as const;`
	`29`	`+const DEFAULT_GENERATED_AUDIO_MAX_BYTES = 16 * 1024 * 1024;`
`29`	`30`
`30`	`31`	`type XaiSpeechResponseFormat = (typeof XAI_SPEECH_RESPONSE_FORMATS)[number];`
`31`	`32`
`@@ -130,6 +131,16 @@ function readXaiOverrides(overrides: SpeechProviderOverrides \| undefined): XaiTt`
`130`	`131`	`};`
`131`	`132`	`}`
`132`	`133`
	`134`	`+function resolveGeneratedAudioMaxBytes(req: {`
	`135`	`+cfg: { agents?: { defaults?: { mediaMaxMb?: number } } };`
	`136`	`+}): number {`
	`137`	`+const configured = req.cfg.agents?.defaults?.mediaMaxMb;`
	`138`	`+if (typeof configured === "number" && Number.isFinite(configured) && configured > 0) {`
	`139`	`+return Math.floor(configured * 1024 * 1024);`
	`140`	`+}`
	`141`	`+return DEFAULT_GENERATED_AUDIO_MAX_BYTES;`
	`142`	`+}`
	`143`	`+`
`133`	`144`	`function parseDirectiveToken(ctx: SpeechDirectiveTokenParseContext): {`
`134`	`145`	`handled: boolean;`
`135`	`146`	`overrides?: SpeechProviderOverrides;`
`@@ -231,6 +242,7 @@ export function buildXaiSpeechProvider(): SpeechProviderPlugin {`
`231`	`242`	`speed: overrides.speed ?? config.speed,`
`232`	`243`	`responseFormat,`
`233`	`244`	`timeoutMs: req.timeoutMs,`
	`245`	`+maxBytes: resolveGeneratedAudioMaxBytes(req),`
`234`	`246`	`});`
`235`	`247`	`return {`
`236`	`248`	`audioBuffer,`
`@@ -254,6 +266,7 @@ export function buildXaiSpeechProvider(): SpeechProviderPlugin {`
`254`	`266`	`speed: overrides.speed ?? config.speed,`
`255`	`267`	`responseFormat: outputFormat,`
`256`	`268`	`timeoutMs: req.timeoutMs,`
	`269`	`+maxBytes: resolveGeneratedAudioMaxBytes(req),`
`257`	`270`	`});`
`258`	`271`	`return { audioBuffer, outputFormat, sampleRate };`
`259`	`272`	`},`

此内容由惯性聚合(RSS阅读器)自动聚合整理，仅供阅读参考。原文来自 — 版权归原作者所有。

推荐订阅源

Recent Commits to openclaw:main