HeyPuter/puter · error · HttpError

bad_request

bad_request

Error message

TTS provider not configured: ${providerName}. Available: ${Object.keys(this.#providers).join(', ')}

What it means

After #resolveProvider returns a canonical provider name, TTSDriver looks it up in #providers (the providers actually instantiated from config). If the name is valid but no provider instance exists — because its credentials were never supplied or its constructor threw during boot — it throws HTTP 400 (legacyCode bad_request), listing the providers that ARE configured. Distinct from 425: here the name normalized fine; it just isn't configured.

Source

Thrown at src/backend/drivers/ai-tts/TTSDriver.ts:163

    }

    /**
     * Synthesize speech from text, routed to the provider named by `provider`
     * (or the default when none is given).
     */
    async synthesize(
        args: ISynthesizeArgs,
    ): Promise<DriverStreamResult | { url: string; content_type: string }> {
        const actor = Context.get('actor');
        if (!actor)
            throw new HttpError(401, 'Authentication required', {
                legacyCode: 'unauthorized',
            });

        const providerName = this.#resolveProvider(args);
        const provider = this.#providers[providerName];
        if (!provider) {
            throw new HttpError(
                400,
                `TTS provider not configured: ${providerName}. Available: ${Object.keys(this.#providers).join(', ')}`,
                { legacyCode: 'bad_request' },
            );
        }

        return provider.synthesize(
            this.#providerArgs(providerName, args),
        ) as Promise<
            DriverStreamResult | { url: string; content_type: string }
        >;
    }

    // -- Provider routing --------------------------------------------

    /**
     * Decide which provider handles a call. An explicit `provider` wins, then
     * an `engine` that names a provider (a long-standing shorthand), then the

View on GitHub (pinned to 908ec23eda)

Solutions

1. Supply credentials for the requested provider in config.providers (e.g. aws-polly needs aws.access_key and aws.secret_key).
2. Call the driver's list() method first to see which providers are actually configured, then route to one of them.
3. Omit the provider argument to let #defaultProvider pick a configured one.
4. Check boot logs for '[TTSDriver] Failed to init ... provider' warnings indicating a construction failure.

Example fix

// before
await driver.synthesize({ text: 'hi', provider: 'aws-polly' }); // not configured
// after — let the default (or a configured provider) handle it
await driver.synthesize({ text: 'hi' });

// Ask the driver which providers are actually configured, then pick one.
const configured = await driver.list(); // returns Object.keys(#providers)
if (!configured.includes(targetProvider)) {
  // fall back to a configured provider or surface 'unavailable'
}
await driver.synthesize({ text: 'hi', provider: configured[0] });

try {
  await driver.synthesize({ text: 'hi', provider: name });
} catch (e) {
  if (e?.fields?.legacyCode === 'bad_request' && /not configured/.test(e.message)) {
    const available = await driver.list();
    // retry with an available provider or notify the user
  } else throw e;
}

When it happens

Trigger: Requesting provider 'aws-polly' when no access_key/secret_key were given; 'elevenlabs' with no apiKey; 'gemini' with no key; or any provider whose constructor threw during #registerProviders (logged as a console.warn on boot).

Common situations: Default provider is aws-polly but the deployment only configured openai — any call that resolves to aws-polly fails; a key was typo'd so construction logged a warning and the provider stayed unregistered; requesting a provider by alias that maps to an unconfigured canonical name.

Related errors

• internal_error
• internal_error
• internal_error
• internal_error
• internal_error

AI-assisted analysis of HeyPuter/puter@908ec23eda (2026-08-12). Data as JSON: /api/errors/e6f6734c1931003a. Report an issue: GitHub.