Files
Sybil-2/server/tests/chat-tools-streaming.test.ts

283 lines
8.7 KiB
TypeScript
Raw Permalink Normal View History

2026-05-02 23:09:39 -07:00
import assert from "node:assert/strict";
import test from "node:test";
import {
2026-05-04 21:52:39 -07:00
runPlainChatCompletionsStream,
runToolAwareChatCompletions,
2026-05-02 23:09:39 -07:00
runToolAwareChatCompletionsStream,
runToolAwareOpenAIChatStream,
type ToolAwareStreamingEvent,
} from "../src/llm/chat-tools.js";
async function* streamFrom(events: any[]) {
for (const event of events) {
await Promise.resolve();
yield event;
}
}
async function collectEvents(iterable: AsyncIterable<ToolAwareStreamingEvent>) {
const events: ToolAwareStreamingEvent[] = [];
for await (const event of iterable) {
events.push(event);
}
return events;
}
test("OpenAI Responses stream emits text deltas as they arrive", async () => {
const outputMessage = {
id: "msg_1",
type: "message",
role: "assistant",
status: "completed",
content: [{ type: "output_text", text: "Hello" }],
};
const client = {
responses: {
create: async () =>
streamFrom([
{ type: "response.output_item.added", item: { ...outputMessage, content: [] }, output_index: 0 },
{ type: "response.output_text.delta", delta: "Hel", output_index: 0, content_index: 0 },
{ type: "response.output_text.delta", delta: "lo", output_index: 0, content_index: 0 },
{ type: "response.output_item.done", item: outputMessage, output_index: 0 },
{
type: "response.completed",
response: {
status: "completed",
output_text: "Hello",
output: [outputMessage],
usage: { input_tokens: 2, output_tokens: 1, total_tokens: 3 },
},
},
]),
},
};
const events = await collectEvents(
runToolAwareOpenAIChatStream({
client: client as any,
model: "gpt-test",
messages: [{ role: "user", content: "Say hello" }],
})
);
assert.deepEqual(
events.map((event) => event.type),
["delta", "delta", "done"]
);
assert.deepEqual(
events.filter((event) => event.type === "delta").map((event) => event.text),
["Hel", "lo"]
);
assert.equal(events.at(-1)?.type === "done" ? events.at(-1)?.result.text : null, "Hello");
});
test("OpenAI-compatible Chat Completions stream emits text deltas as they arrive", async () => {
const client = {
chat: {
completions: {
create: async () =>
streamFrom([
{ choices: [{ delta: { role: "assistant" } }] },
{ choices: [{ delta: { content: "Hel" } }] },
{ choices: [{ delta: { content: "lo" } }] },
{
choices: [{ delta: {}, finish_reason: "stop" }],
usage: { prompt_tokens: 2, completion_tokens: 1, total_tokens: 3 },
},
]),
},
},
};
const events = await collectEvents(
runToolAwareChatCompletionsStream({
client: client as any,
model: "grok-test",
messages: [{ role: "user", content: "Say hello" }],
})
);
assert.deepEqual(
events.map((event) => event.type),
["delta", "delta", "done"]
);
assert.deepEqual(
events.filter((event) => event.type === "delta").map((event) => event.text),
["Hel", "lo"]
);
assert.equal(events.at(-1)?.type === "done" ? events.at(-1)?.result.text : null, "Hello");
});
2026-05-04 21:52:39 -07:00
test("plain Chat Completions stream does not send Sybil-managed tools", async () => {
let requestBody: any = null;
const client = {
chat: {
completions: {
create: async (body: any) => {
requestBody = body;
return streamFrom([
{ choices: [{ delta: { content: "Hi" } }] },
{ choices: [{ delta: {}, finish_reason: "stop" }] },
]);
},
},
},
};
const events = await collectEvents(
runPlainChatCompletionsStream({
client: client as any,
model: "hermes-agent",
messages: [{ role: "user", content: "Say hi" }],
})
);
assert.equal(requestBody.model, "hermes-agent");
assert.equal(requestBody.stream, true);
assert.equal("tools" in requestBody, false);
assert.deepEqual(
events.map((event) => event.type),
["delta", "done"]
);
assert.equal(events.at(-1)?.type === "done" ? events.at(-1)?.result.text : null, "Hi");
});
2026-06-05 22:20:56 -07:00
test("fetch_url sends browser-like navigation headers", async () => {
const originalFetch = globalThis.fetch;
const fetchCalls: Array<{ input: RequestInfo | URL; init?: RequestInit }> = [];
globalThis.fetch = (async (input: RequestInfo | URL, init?: RequestInit) => {
fetchCalls.push({ input, init });
return new Response("<!doctype html><title>CPI</title><main>Consumer price index</main>", {
status: 200,
headers: { "content-type": "text/html; charset=utf-8" },
});
}) as typeof fetch;
try {
let requestCount = 0;
const client = {
chat: {
completions: {
create: async () => {
requestCount += 1;
if (requestCount === 1) {
return {
choices: [
{
message: {
tool_calls: [
{
id: "call_1",
type: "function",
function: {
name: "fetch_url",
arguments: JSON.stringify({ url: "https://www.bls.gov/news.release/pdf/cpi.pdf" }),
},
},
],
},
},
],
};
}
return {
choices: [{ message: { content: "Fetched" } }],
};
},
},
},
};
const result = await runToolAwareChatCompletions({
client: client as any,
model: "grok-test",
messages: [{ role: "user", content: "Fetch CPI PDF" }],
});
assert.equal(result.text, "Fetched");
assert.equal(fetchCalls.length, 1);
assert.equal(String(fetchCalls[0]?.input), "https://www.bls.gov/news.release/pdf/cpi.pdf");
assert.deepEqual(fetchCalls[0]?.init?.headers, {
"User-Agent":
"Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/130.0.0.0 Safari/537.36",
Accept: "text/html,application/xhtml+xml,application/xml;q=0.9,application/pdf;q=0.9,*/*;q=0.8",
"Accept-Language": "en-US,en;q=0.9",
"Upgrade-Insecure-Requests": "1",
"Sec-Fetch-Dest": "document",
"Sec-Fetch-Mode": "navigate",
"Sec-Fetch-Site": "none",
"Sec-Fetch-User": "?1",
});
assert.equal(result.toolEvents[0]?.status, "completed");
} finally {
globalThis.fetch = originalFetch;
}
});
2026-06-05 22:20:56 -07:00
test("OpenAI-compatible Chat Completions stream emits initiated and terminal tool call updates", async () => {
let requestCount = 0;
const client = {
chat: {
completions: {
create: async () => {
requestCount += 1;
if (requestCount === 1) {
return streamFrom([
{
choices: [
{
delta: {
tool_calls: [
{
index: 0,
id: "call_1",
function: {
name: "unknown_tool",
arguments: "{\"query\":\"current weather\"}",
},
},
],
},
finish_reason: "tool_calls",
},
],
},
]);
}
return streamFrom([
{ choices: [{ delta: { content: "Done" } }] },
{ choices: [{ delta: {}, finish_reason: "stop" }] },
]);
},
},
},
};
const events = await collectEvents(
runToolAwareChatCompletionsStream({
client: client as any,
model: "grok-test",
messages: [{ role: "user", content: "Use a tool" }],
})
);
assert.deepEqual(
events.map((event) => event.type),
["tool_call", "tool_call", "delta", "done"]
);
const toolEvents = events.flatMap((event) => (event.type === "tool_call" ? [event.event] : []));
assert.equal(toolEvents[0]?.toolCallId, "call_1");
assert.equal(toolEvents[0]?.status, "initiated");
assert.equal(toolEvents[0]?.completedAt, undefined);
assert.equal(toolEvents[0]?.durationMs, undefined);
assert.equal(toolEvents[1]?.toolCallId, "call_1");
assert.equal(toolEvents[1]?.status, "failed");
assert.match(toolEvents[1]?.error ?? "", /Unknown tool: unknown_tool/);
assert.equal(typeof toolEvents[1]?.completedAt, "string");
assert.equal(typeof toolEvents[1]?.durationMs, "number");
assert.equal(events.at(-1)?.type === "done" ? events.at(-1)?.result.text : null, "Done");
});