Files
9router/tests/unit/cursor-agent-exec-request.test.js
T
Amir Seify c933eefc27 fix(cursor): stop AgentService empty turns and silent tool hangs
Cursor-hosted models (cu/composer-2.5, cu/cursor-grok-*, cu/default) returned
HTTP 200 with an empty turn, or hung, whenever a client sent tools.

- Fold system prompts into the current user message. custom_system_prompt
  (RunRequest field 8) makes AgentService return an empty turn.
- Send ModelDetails (field 3); thinking variants (Composer, Grok, *-thinking)
  return an empty turn when only requested_model (field 9) is set.
- Route tool-call history and declared tool schemas through AgentService:
  encode OpenAI tools into mcp_tools (field 4), decode McpArgs and emit real
  tool_calls with finish_reason tool_calls.
- Map Composer  thinking / Grok thinking_delta (field 4) into visible content
  instead of dropping the answer with the unsigned reasoning.
- Ack request_context without echoing MCP tools (double-advertise stalls the
  HTTP/2 stream) and ack kv_server_message so the run proceeds.
- Reject IDE builtin execs instead of failing the turn, so the model can
  continue with MCP tools or a text answer.
- Add google.protobuf.Value / MCP encoders and a FIXED64 branch to
  encodeField in cursorProtobuf.js.

RTK now compresses the source-format body before translation for cursor only:
its translator rewrites role:tool into user XML, so the post-translate pass
missed those tool results. Every other provider keeps the post-translate pass
unchanged.
2026-09-21 19:49:48 +07:00

168 lines
5.9 KiB
JavaScript

import { describe, it, expect } from "vitest";
import { CursorExecutor } from "../../open-sse/executors/cursor.js";
import { encodeField, wrapConnectRPCFrame } from "../../open-sse/utils/cursorProtobuf.js";
const LEN = 2;
// agent.v1.AgentServerMessage.exec_request (field 2) carrying one ExecServerMessage variant.
function execRequestFrame(execField) {
const execServerMessage = Buffer.from(encodeField(execField, LEN, new Uint8Array()));
return Buffer.from(wrapConnectRPCFrame(encodeField(2, LEN, execServerMessage)));
}
// agent.v1.AgentServerMessage.interaction_update (field 1) → text delta.
function textFrame(text) {
const textPart = Buffer.from(encodeField(1, LEN, text));
const update = Buffer.from(encodeField(1, LEN, textPart));
return Buffer.from(wrapConnectRPCFrame(encodeField(1, LEN, update)));
}
// InteractionUpdate.thinking_delta (field 4) + turn_ended (field 14).
function thinkingFrame(text) {
const thinkingPart = Buffer.from(encodeField(1, LEN, text));
const update = Buffer.from(encodeField(4, LEN, thinkingPart));
return Buffer.from(wrapConnectRPCFrame(encodeField(1, LEN, update)));
}
function turnEndedFrame() {
const update = Buffer.from(encodeField(14, LEN, new Uint8Array()));
return Buffer.from(wrapConnectRPCFrame(encodeField(1, LEN, update)));
}
function stubAgentSession(executor, frames) {
const written = [];
const queue = [...frames];
executor.openAgentHttp2Stream = () => ({
responseHeaders: Promise.resolve({ ":status": 200 }),
write: (frame) => written.push(Buffer.from(frame)),
end() {},
close() {},
async read() {
if (!queue.length) return { value: undefined, done: true };
return { value: queue.shift(), done: false };
},
});
return written;
}
const credentials = {
accessToken: "test-token",
providerSpecificData: { machineId: "a".repeat(64) },
};
function parseSSE(text) {
return text
.split("\n\n")
.filter((chunk) => chunk.startsWith("data: "))
.map((chunk) => chunk.slice("data: ".length))
.filter((data) => data !== "[DONE]")
.map((data) => JSON.parse(data));
}
async function runAgent({ frames, stream, model = "gpt-5.2", tools }) {
const executor = new CursorExecutor();
const written = stubAgentSession(executor, frames);
const result = await executor.executeAgent({
model,
body: { messages: [{ role: "user", content: "hi" }], ...(tools ? { tools } : {}) },
stream,
credentials,
});
return { result, written };
}
describe("CursorExecutor AgentService exec_request handling", () => {
it("acknowledges a request-context exec request without ending the turn", async () => {
const { result, written } = await runAgent({
frames: [execRequestFrame(10), textFrame("hello")],
stream: true,
});
expect(written.length).toBe(2); // run frame + request-context reply
const events = parseSSE(await result.response.text());
const content = events.map((e) => e.choices?.[0]?.delta?.content || "").join("");
expect(content).toBe("hello");
});
it("does not echo client tools on the request_context ack", async () => {
const { written, result } = await runAgent({
tools: [{ function: { name: "read_file", parameters: { type: "object" } } }],
frames: [execRequestFrame(10), textFrame("hello")],
stream: true,
});
expect(written.length).toBe(2);
expect(written[1].toString("utf8")).not.toContain("read_file");
const content = parseSSE(await result.response.text())
.map((e) => e.choices?.[0]?.delta?.content || "")
.join("");
expect(content).toBe("hello");
});
it("does not render an unsupported exec request as assistant content", async () => {
const { result, written } = await runAgent({
frames: [textFrame("partial answer"), execRequestFrame(2), textFrame(" more")],
stream: true,
});
const body = await result.response.text();
expect(body).not.toContain("unsupported IDE tool");
const events = parseSSE(body);
const content = events.map((e) => e.choices?.[0]?.delta?.content || "").join("");
expect(content).toBe("partial answer more");
expect(events.some((e) => e.error)).toBe(false);
expect(written.length).toBe(2); // run frame + IDE rejection
});
it("still emits later text after rejecting an IDE exec in the same read", async () => {
const { result } = await runAgent({
frames: [Buffer.concat([execRequestFrame(2), textFrame("late")])],
stream: true,
});
const body = await result.response.text();
expect(body).not.toContain("unsupported IDE tool");
expect(body).toContain("late");
});
it("returns a non-200 error body for an unsupported exec request when not streaming", async () => {
const { result } = await runAgent({
frames: [execRequestFrame(11)],
stream: false,
});
expect(result.response.status).not.toBe(200);
const payload = await result.response.json();
expect(payload.error.message).toContain("unsupported IDE tool");
});
it("streams Composer visible content from thinking_delta after </think>", async () => {
const { result } = await runAgent({
model: "composer-2.5",
frames: [
thinkingFrame("private reasoning that must not leak</think>OK"),
turnEndedFrame(),
],
stream: true,
});
const events = parseSSE(await result.response.text());
const content = events.map((e) => e.choices?.[0]?.delta?.content || "").join("");
expect(content).toBe("OK");
expect(JSON.stringify(events)).not.toContain("private reasoning");
});
it("flushes Grok thinking as visible content when the turn has no text_delta", async () => {
const { result } = await runAgent({
model: "grok-4.5",
frames: [thinkingFrame("hello from grok"), turnEndedFrame()],
stream: true,
});
const events = parseSSE(await result.response.text());
const content = events.map((e) => e.choices?.[0]?.delta?.content || "").join("");
expect(content).toBe("hello from grok");
});
});