Open-source AI coworkers that each get a computer of their own: a browser, files and tools, with every action decided before it happens and recorded after. Bring any AG-UI agent.
import { expect, test } from "bun:test";
import { MAX_DICTATION_BYTES } from "@openbot/shared/dictation";
import type { MiddlewareHandler } from "hono";
import type { AppVariables } from "../src/auth/guards";
import { createDictationRoutes } from "../src/dictation/routes";
const user: MiddlewareHandler<{ Variables: AppVariables }> = async (
context,
next,
) => {
context.set("actor", {
id: "user-1",
email: "[email protected]",
role: "user",
});
await next();
};
function recording(type = "audio/webm", size = 8) {
const body = new FormData();
const extension = type.includes("webm") ? "webm" : "html";
body.set(
"file",
new File([new Uint8Array(size)], `recording.${extension}`, { type }),
);
return { method: "POST", headers: { "x-openbot-dictation": "1" }, body };
}
test("requires authentication before processing audio", async () => {
const app = createDictationRoutes(
(context) => context.json({ error: "Authentication required." }, 401),
{
transcribe: async () => {
throw new Error("Must not run");
},
},
);
expect((await app.request("/transcriptions", recording())).status).toBe(401);
});
test("disabled service and cross-site forms cannot transcribe", async () => {
const app = createDictationRoutes(user, undefined);
expect((await app.request("/transcriptions", recording())).status).toBe(503);
expect(
(await app.request("/transcriptions", { ...recording(), headers: {} }))
.status,
).toBe(403);
});
test("returns text without storing audio, and normalizes codec MIME parameters", async () => {
const app = createDictationRoutes(user, {
transcribe: async (file) => {
expect(file.name).toBe("recording.webm");
expect(file.type).toBe("audio/webm");
return "Hello world";
},
});
const result = await app.request(
"/transcriptions",
recording("audio/webm;codecs=opus"),
);
expect(result.status).toBe(200);
expect(result.headers.get("cache-control")).toBe("no-store");
expect(await result.json()).toEqual({ text: "Hello world" });
});
test("rejects empty, unsupported, malformed and oversized recordings", async () => {
const app = createDictationRoutes(user, {
transcribe: async () => {
throw new Error("Must not run");
},
});
expect(
(await app.request("/transcriptions", recording("audio/webm", 0))).status,
).toBe(400);
expect(
(await app.request("/transcriptions", recording("text/html"))).status,
).toBe(415);
expect(
(
await app.request(
"/transcriptions",
recording("audio/webm", MAX_DICTATION_BYTES + 1),
)
).status,
).toBe(413);
expect(
(
await app.request(
"/transcriptions",
recording("audio/webm", MAX_DICTATION_BYTES + 100_000),
)
).status,
).toBe(413);
expect(
(
await app.request("/transcriptions", {
method: "POST",
headers: { "x-openbot-dictation": "1" },
body: "broken",
})
).status,
).toBe(400);
});
test("silence and upstream failures are actionable and do not leak secrets", async () => {
const silence = createDictationRoutes(user, { transcribe: async () => "" });
expect((await silence.request("/transcriptions", recording())).status).toBe(
422,
);
const failure = createDictationRoutes(user, {
transcribe: async () => {
throw new Error("secret-key");
},
});
const response = await failure.request("/transcriptions", recording());
expect(response.status).toBe(502);
expect(await response.text()).not.toContain("secret-key");
});
test("limits simultaneous requests and releases the slot after failure", async () => {
const gate = Promise.withResolvers<string>();
const started = Promise.withResolvers<void>();
const app = createDictationRoutes(user, {
transcribe: async () => {
started.resolve();
return gate.promise;
},
});
const first = app.request("/transcriptions", recording());
await started.promise;
expect((await app.request("/transcriptions", recording())).status).toBe(429);
gate.reject(new Error("network unavailable"));
expect((await first).status).toBe(502);
expect((await app.request("/transcriptions", recording())).status).toBe(502);
});