File size: 9,312 Bytes
92a1d83
 
 
 
 
 
 
ff49634
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
92a1d83
 
 
 
 
 
 
 
 
 
 
 
ff49634
 
92a1d83
 
 
ff49634
 
92a1d83
 
 
ff49634
92a1d83
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
ff49634
 
 
92a1d83
 
 
 
 
 
 
 
 
 
 
ff49634
92a1d83
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
ff49634
92a1d83
 
ff49634
92a1d83
 
 
 
 
 
ff49634
92a1d83
 
 
 
 
 
 
 
ff49634
92a1d83
 
 
 
 
 
 
 
 
 
 
 
 
 
ff49634
92a1d83
 
 
 
 
 
 
 
 
ff49634
92a1d83
 
 
 
 
 
 
 
 
 
 
 
ff49634
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
92a1d83
ff49634
92a1d83
 
 
 
ff49634
 
92a1d83
 
 
 
 
 
 
 
 
ff49634
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
92a1d83
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
import { describe, expect, it, beforeAll, afterAll } from "bun:test";
import { OpenAICompatibleClient } from "../client";

const BASE_URL = "https://api.openai.com/v1";
const API_KEY = "sk-test-key-12345";
const MODEL = "gpt-4";

type FetchFn = typeof globalThis.fetch;

/** Bun/TS types fetch as an object with static helpers (e.g. preconnect), not just a function. */
function asFetchMock(fn: (...args: Parameters<FetchFn>) => ReturnType<FetchFn>): FetchFn {
  return fn as FetchFn;
}

function setMockFetch(fn: (...args: Parameters<FetchFn>) => ReturnType<FetchFn>): void {
  globalThis.fetch = asFetchMock(fn);
}

function parseRequestBody(opts?: RequestInit): Record<string, unknown> {
  if (opts?.body == null) {
    throw new Error("Expected request body in mock fetch");
  }
  return JSON.parse(String(opts.body));
}

function makeMockStream(chunks: string[]): ReadableStream {
  const encoder = new TextEncoder();
  return new ReadableStream({
    async start(controller) {
      for (const chunk of chunks) {
        controller.enqueue(encoder.encode(chunk));
      }
      controller.close();
    },
  });
}

function makeFetchMock(streamChunks: string[], status = 200): FetchFn {
  return asFetchMock(async () =>
    new Response(makeMockStream(streamChunks), {
      status,
      headers: { "content-type": "text/event-stream" },
    }),
  );
}

describe("OpenAICompatibleClient", () => {
  let originalFetch: FetchFn;

  beforeAll(() => {
    originalFetch = globalThis.fetch;
  });

  afterAll(() => {
    globalThis.fetch = originalFetch;
  });

  it("sends correct request and parses SSE response", async () => {
    globalThis.fetch = makeFetchMock([
      `data: ${JSON.stringify({ choices: [{ delta: { content: "Hello" } }] })}\n`,
      `data: ${JSON.stringify({ choices: [{ delta: { content: " World" } }] })}\n`,
      `data: ${JSON.stringify({ choices: [{ delta: {} }], usage: { total_tokens: 10 } })}\n`,
      "data: [DONE]\n",
    ]);

    const client = new OpenAICompatibleClient(BASE_URL, API_KEY);
    const result = await client.chatCompletion({
      model: MODEL,
      messages: [{ role: "user", content: "Test" }],
    });

    expect(result.content).toBe("Hello World");
  });

  it("calls onToken callback for each token", async () => {
    globalThis.fetch = makeFetchMock([
      `data: ${JSON.stringify({ choices: [{ delta: { content: "A" } }] })}\n`,
      `data: ${JSON.stringify({ choices: [{ delta: { content: "B" } }] })}\n`,
      "data: [DONE]\n",
    ]);

    const tokens: string[] = [];
    const client = new OpenAICompatibleClient(BASE_URL, API_KEY);
    await client.chatCompletion(
      { model: MODEL, messages: [{ role: "user", content: "Test" }] },
      { onToken: (t) => tokens.push(t) },
    );

    expect(tokens).toEqual(["A", "B"]);
  });

  it("returns usage from the last chunk", async () => {
    globalThis.fetch = makeFetchMock([
      `data: ${JSON.stringify({ choices: [{ delta: { content: "Hi" } }] })}\n`,
      `data: ${JSON.stringify({ choices: [{ delta: {} }], usage: { total_tokens: 5 } })}\n`,
      "data: [DONE]\n",
    ]);

    const client = new OpenAICompatibleClient(BASE_URL, API_KEY);
    const result = await client.chatCompletion({
      model: MODEL,
      messages: [{ role: "user", content: "Test" }],
    });

    expect(result.usage?.total_tokens).toBe(5);
  });

  it("throws on non-OK response", async () => {
    setMockFetch(async () =>
      new Response("Bad Request", { status: 400, headers: { "content-type": "text/plain" } }),
    );

    const client = new OpenAICompatibleClient(BASE_URL, API_KEY);
    expect(
      client.chatCompletion({
        model: MODEL,
        messages: [{ role: "user", content: "Test" }],
      }),
    ).rejects.toThrow("400");
  });

  it("throws on empty response body", async () => {
    setMockFetch(async () => new Response(null, { status: 200 }));

    const client = new OpenAICompatibleClient(BASE_URL, API_KEY);
    expect(
      client.chatCompletion({
        model: MODEL,
        messages: [{ role: "user", content: "Test" }],
      }),
    ).rejects.toThrow("Empty response body");
  });

  it("throws on invalid base URL", async () => {
    const client = new OpenAICompatibleClient("not-a-url", API_KEY);
    expect(
      client.chatCompletion({
        model: MODEL,
        messages: [{ role: "user", content: "Test" }],
      }),
    ).rejects.toThrow("Invalid base URL");
  });

  it("throws on metadata host (169.254.169.254)", async () => {
    const client = new OpenAICompatibleClient("https://169.254.169.254/v1", API_KEY);
    expect(
      client.chatCompletion({
        model: MODEL,
        messages: [{ role: "user", content: "Test" }],
      }),
    ).rejects.toThrow("metadata/private network");
  });

  it("retries without response_format on 400 with response_format error", async () => {
    let callCount = 0;
    setMockFetch(async (_input, opts) => {
      callCount++;
      if (callCount === 1) {
        const body = parseRequestBody(opts);
        expect(body.response_format).toBeDefined();
        return new Response("response_format is not supported", {
          status: 400,
          headers: { "content-type": "text/plain" },
        });
      }
      const body = parseRequestBody(opts);
      expect(body.response_format).toBeUndefined();
      return new Response(makeMockStream([
        `data: ${JSON.stringify({ choices: [{ delta: { content: "retried" } }] })}\n`,
        "data: [DONE]\n",
      ]), {
        status: 200,
        headers: { "content-type": "text/event-stream" },
      });
    });

    const client = new OpenAICompatibleClient(BASE_URL, API_KEY);
    const result = await client.chatCompletion({
      model: MODEL,
      messages: [{ role: "user", content: "Test" }],
      response_format: { type: "json_object" },
    });

    expect(callCount).toBe(2);
    expect(result.content).toBe("retried");
  });

  it("retries with more tokens on truncated response", async () => {
    let callCount = 0;
    setMockFetch(async (_input, _opts) => {
      callCount++;
      const responseContent = callCount === 1
        ? `data: ${JSON.stringify({ choices: [{ delta: { content: '{"incomplete":' } }] })}\n` + "data: [DONE]\n"
        : `data: ${JSON.stringify({ choices: [{ delta: { content: '{"complete": true}' } }] })}\n` + "data: [DONE]\n";

      return new Response(makeMockStream([responseContent]), {
        status: 200,
        headers: { "content-type": "text/event-stream" },
      });
    });

    const client = new OpenAICompatibleClient(BASE_URL, API_KEY);
    const result = await client.chatCompletion({
      model: MODEL,
      messages: [{ role: "user", content: "Test" }],
      max_tokens: 100,
    });

    expect(callCount).toBe(2);
    expect(result.content).toBe('{"complete": true}');
  });

  it("strips reasoning blocks and returns JSON from GLM-style streams", async () => {
    const json = '{"questions":[{"format":"mcq"}]}';
    globalThis.fetch = makeFetchMock([
      `data: ${JSON.stringify({ choices: [{ delta: { content: "<think>\nPlanning questions...\n</think>" } }] })}\n`,
      `data: ${JSON.stringify({ choices: [{ delta: { content: json } }] })}\n`,
      "data: [DONE]\n",
    ]);

    const client = new OpenAICompatibleClient(BASE_URL, API_KEY);
    const result = await client.chatCompletion({
      model: MODEL,
      messages: [{ role: "user", content: "Test" }],
    });

    expect(result.content).toBe(json);
  });

  it("ignores reasoning_content delta and keeps final content", async () => {
    globalThis.fetch = makeFetchMock([
      `data: ${JSON.stringify({ choices: [{ delta: { reasoning_content: "internal reasoning only" } }] })}\n`,
      `data: ${JSON.stringify({ choices: [{ delta: { content: '{"ok":true}' } }] })}\n`,
      "data: [DONE]\n",
    ]);

    const client = new OpenAICompatibleClient(BASE_URL, API_KEY);
    const result = await client.chatCompletion({
      model: MODEL,
      messages: [{ role: "user", content: "Test" }],
    });

    expect(result.content).toBe('{"ok":true}');
  });

  it("throws on HTML error page in stream", async () => {
    setMockFetch(async () =>
      new Response(makeMockStream([
        `data: ${JSON.stringify({ choices: [{ delta: { content: "<!DOCTYPE html><html><body>502 Bad Gateway</body></html>" } }] })}\n`,
        "data: [DONE]\n",
      ]), {
        status: 200,
        headers: { "content-type": "text/event-stream" },
      }),
    );

    const client = new OpenAICompatibleClient(BASE_URL, API_KEY);
    expect(
      client.chatCompletion({
        model: MODEL,
        messages: [{ role: "user", content: "Test" }],
      }),
    ).rejects.toThrow("HTML");
  });

  it("falls back to non-stream JSON body when SSE yields no tokens", async () => {
    const body = JSON.stringify({
      choices: [{ message: { content: '{"questions":[]}' } }],
    });
    setMockFetch(async () =>
      new Response(body, {
        status: 200,
        headers: { "content-type": "application/json" },
      }),
    );

    const client = new OpenAICompatibleClient(BASE_URL, API_KEY);
    const result = await client.chatCompletion({
      model: MODEL,
      messages: [{ role: "user", content: "Test" }],
    });

    expect(result.content).toBe('{"questions":[]}');
  });
});