-
Notifications
You must be signed in to change notification settings - Fork 41
/
client.js
406 lines (385 loc) · 11.7 KB
/
client.js
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
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
const VERSION = '0.3.0';
const RETRY_STATUS_CODES = [429, 500, 502, 503, 504];
const ENDPOINT = 'https://api.mistral.ai';
// We can't use a top level await if eventually this is to be converted
// to typescript and compiled to commonjs, or similarly using babel.
const configuredFetch = Promise.resolve(
globalThis.fetch ?? import('node-fetch').then((m) => m.default),
);
/**
* MistralAPIError
* @return {MistralAPIError}
* @extends {Error}
*/
class MistralAPIError extends Error {
/**
* A simple error class for Mistral API errors
* @param {*} message
*/
constructor(message) {
super(message);
this.name = 'MistralAPIError';
}
}
/**
* @param {Array<AbortSignal|undefined>} signals to merge
* @return {AbortSignal} signal which will abort when any of signals abort
*/
function combineSignals(signals) {
const controller = new AbortController();
signals.forEach((signal) => {
if (!signal) {
return;
}
signal.addEventListener(
'abort',
() => {
controller.abort(signal.reason);
},
{once: true},
);
if (signal.aborted) {
controller.abort(signal.reason);
}
});
return controller.signal;
}
/**
* MistralClient
* @return {MistralClient}
*/
class MistralClient {
/**
* A simple and lightweight client for the Mistral API
* @param {*} apiKey can be set as an environment variable MISTRAL_API_KEY,
* or provided in this parameter
* @param {*} endpoint defaults to https://api.mistral.ai
* @param {*} maxRetries defaults to 5
* @param {*} timeout defaults to 120 seconds
*/
constructor(
apiKey = process.env.MISTRAL_API_KEY,
endpoint = ENDPOINT,
maxRetries = 5,
timeout = 120,
) {
this.endpoint = endpoint;
this.apiKey = apiKey;
this.maxRetries = maxRetries;
this.timeout = timeout;
if (this.endpoint.indexOf('inference.azure.com')) {
this.modelDefault = 'mistral';
}
}
/**
* @return {Promise}
* @private
* @param {...*} args - fetch args
* hook point for non-global fetch override
*/
async _fetch(...args) {
const fetchFunc = await configuredFetch;
return fetchFunc(...args);
}
/**
*
* @param {*} method
* @param {*} path
* @param {*} request
* @param {*} signal
* @return {Promise<*>}
*/
_request = async function(method, path, request, signal) {
const url = `${this.endpoint}/${path}`;
const options = {
method: method,
headers: {
'User-Agent': `mistral-client-js/${VERSION}`,
'Accept': request?.stream ? 'text/event-stream' : 'application/json',
'Content-Type': 'application/json',
'Authorization': `Bearer ${this.apiKey}`,
},
body: method !== 'get' ? JSON.stringify(request) : null,
signal: combineSignals([
AbortSignal.timeout(this.timeout * 1000),
signal,
]),
};
for (let attempts = 0; attempts < this.maxRetries; attempts++) {
try {
const response = await this._fetch(url, options);
if (response.ok) {
if (request?.stream) {
// When using node-fetch or test mocks, getReader is not defined
if (typeof response.body.getReader === 'undefined') {
return response.body;
} else {
const reader = response.body.getReader();
// Chrome does not support async iterators yet, so polyfill it
const asyncIterator = async function* () {
try {
while (true) {
// Read from the stream
const {done, value} = await reader.read();
// Exit if we're done
if (done) return;
// Else yield the chunk
yield value;
}
} finally {
reader.releaseLock();
}
};
return asyncIterator();
}
}
return await response.json();
} else if (RETRY_STATUS_CODES.includes(response.status)) {
console.debug(
`Retrying request on response status: ${response.status}`,
`Response: ${await response.text()}`,
`Attempt: ${attempts + 1}`,
);
// eslint-disable-next-line max-len
await new Promise((resolve) =>
setTimeout(resolve, Math.pow(2, attempts + 1) * 500),
);
} else {
throw new MistralAPIError(
`HTTP error! status: ${response.status} ` +
`Response: \n${await response.text()}`,
);
}
} catch (error) {
console.error(`Request failed: ${error.message}`);
if (error.name === 'MistralAPIError') {
throw error;
}
if (attempts === this.maxRetries - 1) throw error;
// eslint-disable-next-line max-len
await new Promise((resolve) =>
setTimeout(resolve, Math.pow(2, attempts + 1) * 500),
);
}
}
throw new Error('Max retries reached');
};
/**
* Creates a chat completion request
* @param {*} model
* @param {*} messages
* @param {*} tools
* @param {*} temperature
* @param {*} maxTokens
* @param {*} topP
* @param {*} randomSeed
* @param {*} stream
* @param {*} safeMode deprecated use safePrompt instead
* @param {*} safePrompt
* @param {*} toolChoice
* @param {*} responseFormat
* @return {Promise<Object>}
*/
_makeChatCompletionRequest = function(
model,
messages,
tools,
temperature,
maxTokens,
topP,
randomSeed,
stream,
safeMode,
safePrompt,
toolChoice,
responseFormat,
) {
// if modelDefault and model are undefined, throw an error
if (!model && !this.modelDefault) {
throw new MistralAPIError('You must provide a model name');
}
return {
model: model ?? this.modelDefault,
messages: messages,
tools: tools ?? undefined,
temperature: temperature ?? undefined,
max_tokens: maxTokens ?? undefined,
top_p: topP ?? undefined,
random_seed: randomSeed ?? undefined,
stream: stream ?? undefined,
safe_prompt: (safeMode || safePrompt) ?? undefined,
tool_choice: toolChoice ?? undefined,
response_format: responseFormat ?? undefined,
};
};
/**
* Returns a list of the available models
* @return {Promise<Object>}
*/
listModels = async function() {
const response = await this._request('get', 'v1/models');
return response;
};
/**
* A chat endpoint without streaming.
*
* @param {Object} data - The main chat configuration.
* @param {*} data.model - the name of the model to chat with,
* e.g. mistral-tiny
* @param {*} data.messages - an array of messages to chat with, e.g.
* [{role: 'user', content: 'What is the best
* French cheese?'}]
* @param {*} data.tools - a list of tools to use.
* @param {*} data.temperature - the temperature to use for sampling, e.g. 0.5
* @param {*} data.maxTokens - the maximum number of tokens to generate,
* e.g. 100
* @param {*} data.topP - the cumulative probability of tokens to generate,
* e.g. 0.9
* @param {*} data.randomSeed - the random seed to use for sampling, e.g. 42
* @param {*} data.safeMode - deprecated use safePrompt instead
* @param {*} data.safePrompt - whether to use safe mode, e.g. true
* @param {*} data.toolChoice - the tool to use, e.g. 'auto'
* @param {*} data.responseFormat - the format of the response,
* e.g. 'json_format'
* @param {Object} options - Additional operational options.
* @param {*} [options.signal] - optional AbortSignal instance to control
* request The signal will be combined with
* default timeout signal
* @return {Promise<Object>}
*/
chat = async function(
{
model,
messages,
tools,
temperature,
maxTokens,
topP,
randomSeed,
safeMode,
safePrompt,
toolChoice,
responseFormat,
},
{signal} = {},
) {
const request = this._makeChatCompletionRequest(
model,
messages,
tools,
temperature,
maxTokens,
topP,
randomSeed,
false,
safeMode,
safePrompt,
toolChoice,
responseFormat,
);
const response = await this._request(
'post',
'v1/chat/completions',
request,
signal,
);
return response;
};
/**
* A chat endpoint that streams responses.
*
* @param {Object} data - The main chat configuration.
* @param {*} data.model - the name of the model to chat with,
* e.g. mistral-tiny
* @param {*} data.messages - an array of messages to chat with, e.g.
* [{role: 'user', content: 'What is the best
* French cheese?'}]
* @param {*} data.tools - a list of tools to use.
* @param {*} data.temperature - the temperature to use for sampling, e.g. 0.5
* @param {*} data.maxTokens - the maximum number of tokens to generate,
* e.g. 100
* @param {*} data.topP - the cumulative probability of tokens to generate,
* e.g. 0.9
* @param {*} data.randomSeed - the random seed to use for sampling, e.g. 42
* @param {*} data.safeMode - deprecated use safePrompt instead
* @param {*} data.safePrompt - whether to use safe mode, e.g. true
* @param {*} data.toolChoice - the tool to use, e.g. 'auto'
* @param {*} data.responseFormat - the format of the response,
* e.g. 'json_format'
* @param {Object} options - Additional operational options.
* @param {*} [options.signal] - optional AbortSignal instance to control
* request The signal will be combined with
* default timeout signal
* @return {Promise<Object>}
*/
chatStream = async function* (
{
model,
messages,
tools,
temperature,
maxTokens,
topP,
randomSeed,
safeMode,
safePrompt,
toolChoice,
responseFormat,
},
{signal} = {},
) {
const request = this._makeChatCompletionRequest(
model,
messages,
tools,
temperature,
maxTokens,
topP,
randomSeed,
true,
safeMode,
safePrompt,
toolChoice,
responseFormat,
);
const response = await this._request(
'post',
'v1/chat/completions',
request,
signal,
);
let buffer = '';
const decoder = new TextDecoder();
for await (const chunk of response) {
buffer += decoder.decode(chunk, {stream: true});
let firstNewline;
while ((firstNewline = buffer.indexOf('\n')) !== -1) {
const chunkLine = buffer.substring(0, firstNewline);
buffer = buffer.substring(firstNewline + 1);
if (chunkLine.startsWith('data:')) {
const json = chunkLine.substring(6).trim();
if (json !== '[DONE]') {
yield JSON.parse(json);
}
}
}
}
};
/**
* An embeddings endpoint that returns embeddings for a single,
* or batch of inputs
* @param {*} model The embedding model to use, e.g. mistral-embed
* @param {*} input The input to embed,
* e.g. ['What is the best French cheese?']
* @return {Promise<Object>}
*/
embeddings = async function({model, input}) {
const request = {
model: model,
input: input,
};
const response = await this._request('post', 'v1/embeddings', request);
return response;
};
}
export default MistralClient;