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 { describe, expect, test } from "bun:test";
import type { MiddlewareHandler } from "hono";
import { Hono } from "hono";
import type { AuditReader } from "../src/audit";
import type { AppVariables } from "../src/auth/guards";
import type { PolicyStore } from "../src/computer/policy-store";
import { createComputerRoutes } from "../src/computer/routes";
/**
* The dry-run answers from history, so its window is the answer.
*
* The route used to read `typeof limit === "number" ? limit : 200` and clamp, which meant
* `"abc"`, `null` and `true` silently became 200, `Infinity` silently became 500, fractions
* were silently truncated, and `NaN` became `NaN` and travelled into `auditReader.list`.
* A what-if answered from the wrong slice of history is worse than no answer.
*/
const ADMIN = { id: "u1", email: "[email protected]", role: "admin" } as const;
const POLICY = { mode: "enforce", deny: [], allow: [] };
function app(seen: { limit?: number }[]) {
const asActor: MiddlewareHandler<{ Variables: AppVariables }> = async (
context,
next,
) => {
context.set("actor", { ...ADMIN });
await next();
};
const auditReader = {
list: async (query: { limit?: number }) => {
seen.push({
...(query.limit !== undefined ? { limit: query.limit } : {}),
});
return { events: [], nextCursor: undefined };
},
} as unknown as AuditReader;
const routes = createComputerRoutes(
{} as never,
{} as PolicyStore,
asActor,
async () => false,
undefined,
auditReader,
);
return new Hono<{ Variables: AppVariables }>().route(
"/api/computers",
routes,
);
}
const post = (body: unknown) => ({
method: "POST",
headers: { "content-type": "application/json" },
body: JSON.stringify(body),
});
describe("POST /api/computers/policy-dry-run limit", () => {
test("an absent limit still replays the default window", async () => {
const seen: { limit?: number }[] = [];
const response = await app(seen).request(
"http://t/api/computers/policy-dry-run",
post({ policy: POLICY }),
);
expect(response.status).toBe(200);
expect(seen).toEqual([{ limit: 200 }]);
});
test.each([[1], [200], [500]])(
"a limit of %s reaches the reader",
async (limit) => {
const seen: { limit?: number }[] = [];
const response = await app(seen).request(
"http://t/api/computers/policy-dry-run",
post({ policy: POLICY, limit }),
);
expect(response.status).toBe(200);
expect(seen).toEqual([{ limit }]);
},
);
test.each([
["a string", "200"],
["null", null],
["true", true],
["an object", {}],
["an array", [200]],
["NaN", Number.NaN],
["Infinity", Number.POSITIVE_INFINITY],
["zero", 0],
["a negative", -5],
["above the ceiling", 501],
["a fraction", 200.5],
["an empty string", ""],
])("rejects %s with 400 and never reads history", async (_name, limit) => {
const seen: { limit?: number }[] = [];
const response = await app(seen).request(
"http://t/api/computers/policy-dry-run",
post({ policy: POLICY, limit }),
);
expect(response.status).toBe(400);
await expect(response.json()).resolves.toEqual({
error: "limit must be a whole number between 1 and 500.",
});
expect(seen).toEqual([]);
});
test("a malformed policy still answers 400 before the limit is read", async () => {
const seen: { limit?: number }[] = [];
const response = await app(seen).request(
"http://t/api/computers/policy-dry-run",
post({ policy: { mode: "nope" }, limit: "abc" }),
);
expect(response.status).toBe(400);
expect(seen).toEqual([]);
});
});