Voice-to-text dictation app with local (Nvidia Parakeet/Whisper) and cloud models (BYOK). Privacy-first and available cross-platform.
JavaScript aianthropiccross-platformgeminigroqlinuxmacosnvidiaopen-sourceopenaiparakeetspeech-to-texttranscribewhisperwindows
const test = require("node:test");
const assert = require("node:assert/strict");
const load = () => import("../../src/services/ai/inferenceProviders/openai.ts");
const providerContext = {
getApiKey: async () => "test-key",
getSystemPrompt: () => "Clean the transcript",
getCustomDictionary: () => [],
getPreferredLanguage: () => "en",
getUiLanguage: () => "en",
callChatCompletionsApi: async () => {
throw new Error("Unexpected chat completions delegation");
},
calculateMaxTokens: () => 4096,
};
function jsonResponse(body, status = 200) {
return new Response(JSON.stringify(body), {
status,
headers: { "Content-Type": "application/json", "x-request-id": "req_classify_1" },
});
}
test("a single-candidate custom /responses endpoint's 429 insufficient_quota survives as PROVIDER_QUOTA_EXHAUSTED with its requestId", async (t) => {
const originalFetch = globalThis.fetch;
t.after(() => {
globalThis.fetch = originalFetch;
});
// Ending the base in "/responses" makes getEndpointCandidates return a
// single candidate, so detectServerType's own "already a full endpoint"
// check skips the GET /models probe entirely — one POST is all that fires.
globalThis.fetch = async () =>
jsonResponse(
{
error: {
message: "You exceeded your current quota, please check your plan and billing details.",
type: "insufficient_quota",
},
},
429
);
const { openaiProvider } = await load();
await assert.rejects(
openaiProvider.call({
text: "raw transcript",
model: "cleanup-model",
agentName: null,
config: {
provider: "custom",
baseUrl: "https://single-candidate.example/v1/responses",
customApiKey: "test-key",
},
ctx: providerContext,
}),
(error) => {
// Before the fix, the already-classified error was rebuilt from its own
// English message, which doesn't contain the quota signal text and so
// reclassified a 429 as a generic rate limit — dropping the requestId.
assert.equal(error.code, "PROVIDER_QUOTA_EXHAUSTED");
assert.equal(error.messageKey, "providerErrors.selfHosted.quotaExhausted");
assert.equal(error.technicalDetails.requestId, "req_classify_1");
return true;
}
);
});
test("OpenRouter is classified by its own name, never as self-hosted", async (t) => {
const originalFetch = globalThis.fetch;
t.after(() => {
globalThis.fetch = originalFetch;
});
globalThis.fetch = async () =>
jsonResponse({ error: { message: "The model `nope/nope` does not exist" } }, 404);
const { openaiProvider } = await load();
await assert.rejects(
openaiProvider.call({
text: "raw transcript",
model: "nope/nope",
agentName: null,
config: { provider: "openrouter" },
ctx: providerContext,
}),
(error) => {
assert.equal(error.code, "PROVIDER_MODEL_NOT_FOUND");
assert.equal(error.technicalDetails.provider, "OpenRouter");
assert.doesNotMatch(error.messageKey, /selfHosted/);
assert.match(error.message, /^OpenRouter /);
return true;
}
);
});
test("a custom endpoint's error is classified self-hosted", async (t) => {
const originalFetch = globalThis.fetch;
t.after(() => {
globalThis.fetch = originalFetch;
});
globalThis.fetch = async (_input, init = {}) => {
const method = init.method || "GET";
if (method === "GET") return jsonResponse({}, 404); // model discovery probe
return jsonResponse({ error: { message: "Incorrect API key provided" } }, 401);
};
const { openaiProvider } = await load();
await assert.rejects(
openaiProvider.call({
text: "raw transcript",
model: "cleanup-model",
agentName: null,
config: {
provider: "custom",
baseUrl: "https://my-own-server.example/v1",
customApiKey: "test-key",
},
ctx: providerContext,
}),
(error) => {
assert.equal(error.code, "PROVIDER_AUTH_FAILED");
assert.equal(error.messageKey, "providerErrors.selfHosted.authFailed");
assert.equal(error.technicalDetails.provider, "Your server");
assert.match(error.message, /^Your server /);
return true;
}
);
});