-
Notifications
You must be signed in to change notification settings - Fork 2k
/
baidu_qianfan.ts
238 lines (199 loc) Β· 6.78 KB
/
baidu_qianfan.ts
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
import { Embeddings, type EmbeddingsParams } from "@langchain/core/embeddings";
import { chunkArray } from "@langchain/core/utils/chunk_array";
import { getEnvironmentVariable } from "@langchain/core/utils/env";
export interface BaiduQianfanEmbeddingsParams extends EmbeddingsParams {
/** Model name to use */
modelName: "embedding-v1" | "bge_large_zh" | "bge-large-en" | "tao-8k";
/**
* Timeout to use when making requests to BaiduQianfan.
*/
timeout?: number;
/**
* The maximum number of characters allowed for embedding in a single request varies by model:
* - Embedding-V1 model: up to 1000 characters
* - bge-large-zh model: up to 2000 characters
* - bge-large-en model: up to 2000 characters
* - tao-8k model: up to 28000 characters
*
* Note: These limits are model-specific and should be adhered to for optimal performance.
*/
batchSize?: number;
/**
* Whether to strip new lines from the input text.
*/
stripNewLines?: boolean;
}
interface EmbeddingCreateParams {
input: string[];
}
interface EmbeddingResponse {
data: { object: "embedding"; index: number; embedding: number[] }[];
usage: {
prompt_tokens: number;
total_tokens: number;
};
id: string;
}
interface EmbeddingErrorResponse {
error_code: number | string;
error_msg: string;
}
export class BaiduQianfanEmbeddings
extends Embeddings
implements BaiduQianfanEmbeddingsParams
{
modelName: BaiduQianfanEmbeddingsParams["modelName"] = "embedding-v1";
batchSize = 16;
stripNewLines = true;
baiduApiKey: string;
baiduSecretKey: string;
accessToken: string;
constructor(
fields?: Partial<BaiduQianfanEmbeddingsParams> & {
verbose?: boolean;
baiduApiKey?: string;
baiduSecretKey?: string;
}
) {
const fieldsWithDefaults = { maxConcurrency: 2, ...fields };
super(fieldsWithDefaults);
const baiduApiKey =
fieldsWithDefaults?.baiduApiKey ??
getEnvironmentVariable("BAIDU_API_KEY");
const baiduSecretKey =
fieldsWithDefaults?.baiduSecretKey ??
getEnvironmentVariable("BAIDU_SECRET_KEY");
if (!baiduApiKey) {
throw new Error("Baidu API key not found");
}
if (!baiduSecretKey) {
throw new Error("Baidu Secret key not found");
}
this.baiduApiKey = baiduApiKey;
this.baiduSecretKey = baiduSecretKey;
this.modelName = fieldsWithDefaults?.modelName ?? this.modelName;
if (this.modelName === "tao-8k") {
if (fieldsWithDefaults?.batchSize && fieldsWithDefaults.batchSize !== 1) {
throw new Error(
"tao-8k model supports only a batchSize of 1. Please adjust your batchSize accordingly"
);
}
this.batchSize = 1;
} else {
this.batchSize = fieldsWithDefaults?.batchSize ?? this.batchSize;
}
this.stripNewLines =
fieldsWithDefaults?.stripNewLines ?? this.stripNewLines;
}
/**
* Method to generate embeddings for an array of documents. Splits the
* documents into batches and makes requests to the BaiduQianFan API to generate
* embeddings.
* @param texts Array of documents to generate embeddings for.
* @returns Promise that resolves to a 2D array of embeddings for each document.
*/
async embedDocuments(texts: string[]): Promise<number[][]> {
const batches = chunkArray(
this.stripNewLines ? texts.map((t) => t.replace(/\n/g, " ")) : texts,
this.batchSize
);
const batchRequests = batches.map((batch) => {
const params = this.getParams(batch);
return this.embeddingWithRetry(params);
});
const batchResponses = await Promise.all(batchRequests);
const embeddings: number[][] = [];
for (let i = 0; i < batchResponses.length; i += 1) {
const batch = batches[i];
const batchResponse = batchResponses[i] || [];
for (let j = 0; j < batch.length; j += 1) {
embeddings.push(batchResponse[j]);
}
}
return embeddings;
}
/**
* Method to generate an embedding for a single document. Calls the
* embeddingWithRetry method with the document as the input.
* @param text Document to generate an embedding for.
* @returns Promise that resolves to an embedding for the document.
*/
async embedQuery(text: string): Promise<number[]> {
const params = this.getParams([
this.stripNewLines ? text.replace(/\n/g, " ") : text,
]);
const embeddings = (await this.embeddingWithRetry(params)) || [[]];
return embeddings[0];
}
/**
* Method to generate an embedding params.
* @param texts Array of documents to generate embeddings for.
* @returns an embedding params.
*/
private getParams(
texts: EmbeddingCreateParams["input"]
): EmbeddingCreateParams {
return {
input: texts,
};
}
/**
* Private method to make a request to the BaiduAI API to generate
* embeddings. Handles the retry logic and returns the response from the
* API.
* @param request Request to send to the BaiduAI API.
* @returns Promise that resolves to the response from the API.
*/
private async embeddingWithRetry(body: EmbeddingCreateParams) {
if (!this.accessToken) {
this.accessToken = await this.getAccessToken();
}
return fetch(
`https://aip.baidubce.com/rpc/2.0/ai_custom/v1/wenxinworkshop/embeddings/${this.modelName}?access_token=${this.accessToken}`,
{
method: "POST",
headers: {
"Content-Type": "application/json",
},
body: JSON.stringify(body),
}
).then(async (response) => {
const embeddingData: EmbeddingResponse | EmbeddingErrorResponse =
await response.json();
if ("error_code" in embeddingData && embeddingData.error_code) {
throw new Error(
`${embeddingData.error_code}: ${embeddingData.error_msg}`
);
}
return (embeddingData as EmbeddingResponse).data.map(
({ embedding }) => embedding
);
});
}
/**
* Method that retrieves the access token for making requests to the Baidu
* API.
* @returns The access token for making requests to the Baidu API.
*/
private async getAccessToken() {
const url = `https://aip.baidubce.com/oauth/2.0/token?grant_type=client_credentials&client_id=${this.baiduApiKey}&client_secret=${this.baiduSecretKey}`;
const response = await fetch(url, {
method: "POST",
headers: {
"Content-Type": "application/json",
Accept: "application/json",
},
});
if (!response.ok) {
const text = await response.text();
const error = new Error(
`Baidu get access token failed with status code ${response.status}, response: ${text}`
);
// eslint-disable-next-line @typescript-eslint/no-explicit-any
(error as any).response = response;
throw error;
}
const json = await response.json();
return json.access_token;
}
}