支持探索版和K1模型,修复图像解析能力

This commit is contained in:
Vinlic 2024-12-17 17:08:09 +08:00
parent 1dd7f89cbc
commit 1375e7c6c4
8 changed files with 2020 additions and 238 deletions

View File

@ -281,7 +281,14 @@ Authorization: Bearer [refresh_token]
请求数据:
```json
{
// model随意填写如果不希望输出检索过程模型名称请包含silent_search
// 模型名称
// kimi默认模型
// kimi-search联网检索模型
// kimi-research探索版模型
// kimi-k1K1模型
// kimi-math数学模型
// kimi-silent不输出检索过程模型
// search/research/k1/math/silent可自由组合使用
// 如果使用kimi+智能体model请填写智能体ID就是浏览器地址栏上尾部的一串英文+数字20个字符的ID
"model": "kimi",
// 目前多轮对话基于消息合并实现某些场景可能导致能力下降且受单轮最大Token数限制
@ -341,7 +348,15 @@ Authorization: Bearer [refresh_token]
请求数据:
```json
{
// 模型名称随意填写如果不希望输出检索过程模型名称请包含silent_search
// 模型名称
// kimi默认模型
// kimi-search联网检索模型
// kimi-research探索版模型
// kimi-k1K1模型
// kimi-math数学模型
// kimi-silent不输出检索过程模型
// search/research/k1/math/silent可自由组合使用
// 如果使用kimi+智能体model请填写智能体ID就是浏览器地址栏上尾部的一串英文+数字20个字符的ID
"model": "kimi",
"messages": [
{
@ -407,7 +422,15 @@ Authorization: Bearer [refresh_token]
请求数据:
```json
{
// 模型名称随意填写如果不希望输出检索过程模型名称请包含silent_search
// 模型名称
// kimi默认模型
// kimi-search联网检索模型
// kimi-research探索版模型
// kimi-k1K1模型
// kimi-math数学模型
// kimi-silent不输出检索过程模型
// search/research/k1/math/silent可自由组合使用
// 如果使用kimi+智能体model请填写智能体ID就是浏览器地址栏上尾部的一串英文+数字20个字符的ID
"model": "kimi",
"messages": [
{

View File

@ -261,7 +261,15 @@ Authorization: Bearer [refresh_token]
Request data:
```json
{
// Fill in the model name as you like. If you do not want to output the retrieval process model name, please include silent_search.
// Model name
// kimi: default model
// kimi-search: online search model
// kimi-research: exploration version model
// kimi-k1: K1 model
// kimi-math: math model
// kimi-silent: model without search process output
// search/research/k1/math/silent: can be freely combined
// If using kimi+agent, fill in the agent ID for model, which is the 20-character ID of letters and numbers at the end of the browser address bar
"model": "kimi",
"messages": [
{
@ -317,7 +325,15 @@ Authorization: Bearer [refresh_token]
Request data:
```json
{
// Fill in the model name as you like. If you do not want to output the retrieval process model name, please include silent_search.
// Model name
// kimi: default model
// kimi-search: online search model
// kimi-research: exploration version model
// kimi-k1: K1 model
// kimi-math: math model
// kimi-silent: model without search process output
// search/research/k1/math/silent: can be freely combined
// If using kimi+agent, fill in the agent ID for model, which is the 20-character ID of letters and numbers at the end of the browser address bar
"model": "kimi",
"messages": [
{
@ -383,7 +399,15 @@ Authorization: Bearer [refresh_token]
Request data:
```json
{
// Fill in the model name as you like. If you do not want to output the retrieval process model name, please include silent_search.
// Model name
// kimi: default model
// kimi-search: online search model
// kimi-research: exploration version model
// kimi-k1: K1 model
// kimi-math: math model
// kimi-silent: model without search process output
// search/research/k1/math/silent: can be freely combined
// If using kimi+agent, fill in the agent ID for model, which is the 20-character ID of letters and numbers at the end of the browser address bar
"model": "kimi",
"messages": [
{

View File

@ -1,6 +1,6 @@
{
"name": "kimi-free-api",
"version": "0.0.34",
"version": "0.0.36",
"description": "Kimi Free API Server",
"type": "module",
"main": "dist/index.js",
@ -13,8 +13,8 @@
"dist/"
],
"scripts": {
"dev": "tsup src/index.ts --format cjs,esm --sourcemap --dts --publicDir public --watch --onSuccess \"node dist/index.js\"",
"start": "node dist/index.js",
"dev": "tsup src/index.ts --format cjs,esm --sourcemap --dts --publicDir public --watch --onSuccess \"node --enable-source-maps --no-node-snapshot dist/index.js\"",
"start": "node --enable-source-maps --no-node-snapshot dist/index.js",
"build": "tsup src/index.ts --format cjs,esm --sourcemap --dts --clean --publicDir public"
},
"author": "Vinlic",

View File

@ -5,5 +5,6 @@ export default {
API_TOKEN_EXPIRES: [-2002, 'Token已失效'],
API_FILE_URL_INVALID: [-2003, '远程文件URL非法'],
API_FILE_EXECEEDS_SIZE: [-2004, '远程文件超出大小'],
API_CHAT_STREAM_PUSHING: [-2005, '已有对话流正在输出']
API_CHAT_STREAM_PUSHING: [-2005, '已有对话流正在输出'],
API_RESEARCH_EXCEEDS_LIMIT: [-2006, '探索版使用量已达到上限']
}

View File

@ -2,8 +2,9 @@ import { PassThrough } from "stream";
import path from 'path';
import _ from 'lodash';
import mime from 'mime';
import axios, { AxiosResponse } from 'axios';
import axios, { AxiosRequestConfig, AxiosResponse } from 'axios';
import type IStreamMessage from "../interfaces/IStreamMessage.ts";
import APIException from "@/lib/exceptions/APIException.ts";
import EX from "@/api/consts/exceptions.ts";
import { createParser } from 'eventsource-parser'
@ -12,27 +13,39 @@ import util from '@/lib/util.ts';
// 模型名称
const MODEL_NAME = 'kimi';
// 设备ID
const DEVICE_ID = Math.random() * 999999999999999999 + 7000000000000000000;
// SessionID
const SESSION_ID = Math.random() * 99999999999999999 + 1700000000000000000;
// access_token有效期
const ACCESS_TOKEN_EXPIRES = 300;
// 最大重试次数
const MAX_RETRY_COUNT = 3;
// 重试延迟
const RETRY_DELAY = 5000;
// 基础URL
const BASE_URL = 'https://kimi.moonshot.cn';
// 伪装headers
const FAKE_HEADERS = {
'Accept': '*/*',
'Accept-Encoding': 'gzip, deflate, br, zstd',
'Accept-Language': 'zh-CN,zh;q=0.9',
'Origin': 'https://kimi.moonshot.cn',
'Accept-Language': 'zh-CN,zh;q=0.9,en-US;q=0.8,en;q=0.7',
'Cache-Control': 'no-cache',
'Pragma': 'no-cache',
'Origin': BASE_URL,
'Cookie': util.generateCookie(),
'R-Timezone': 'Asia/Shanghai',
'Sec-Ch-Ua': '"Google Chrome";v="123", "Not:A-Brand";v="8", "Chromium";v="123"',
'Sec-Ch-Ua': '"Google Chrome";v="131", "Chromium";v="131", "Not_A Brand";v="24"',
'Sec-Ch-Ua-Mobile': '?0',
'Sec-Ch-Ua-Platform': '"Windows"',
'Sec-Fetch-Dest': 'empty',
'Sec-Fetch-Mode': 'cors',
'Sec-Fetch-Site': 'same-origin',
'User-Agent': 'Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/123.0.0.0 Safari/537.36'
'User-Agent': 'Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/131.0.0.0 Safari/537.36',
'Priority': 'u=1, i',
'X-Msh-Device-Id': `${DEVICE_ID}`,
'X-Msh-Platform': 'web',
'X-Msh-Session-Id': `${SESSION_ID}`
};
// 文件最大大小
const FILE_MAX_SIZE = 100 * 1024 * 1024;
@ -54,23 +67,10 @@ async function requestToken(refreshToken: string) {
accessTokenRequestQueueMap[refreshToken] = [];
logger.info(`Refresh token: ${refreshToken}`);
const result = await (async () => {
const result = await axios.get('https://kimi.moonshot.cn/api/auth/token/refresh', {
const result = await axios.get(`${BASE_URL}/api/auth/token/refresh`, {
headers: {
Accept: '*/*',
'Accept-Encoding': 'gzip, deflate, br, zstd',
'Accept-Language': 'zh-CN,zh;q=0.9',
Authorization: `Bearer ${refreshToken}`,
'Cache-Control': 'no-cache',
'Cookie': util.generateCookie(),
Pragma: 'no-cache',
Referer: 'https://kimi.moonshot.cn/',
'Sec-Ch-Ua': '"Google Chrome";v="123", "Not:A-Brand";v="8", "Chromium";v="123"',
'Sec-Ch-Ua-Mobile': '?0',
'Sec-Ch-Ua-Platform': '"Windows"',
'Sec-Fetch-Dest': 'empty',
'Sec-Fetch-Mode': 'cors',
'Sec-Fetch-Site': 'same-origin',
'User-Agent': 'Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/122.0.0.0 Safari/537.36'
...FAKE_HEADERS,
},
timeout: 15000,
validateStatus: () => true
@ -79,9 +79,18 @@ async function requestToken(refreshToken: string) {
access_token,
refresh_token
} = checkResult(result, refreshToken);
const { id: userId } = await getUserInfo(access_token, refreshToken);
const userResult = await axios.get(`${BASE_URL}/api/user`, {
headers: {
Authorization: `Bearer ${access_token}`,
...FAKE_HEADERS,
},
timeout: 15000,
validateStatus: () => true
});
if(!userResult.data.id)
throw new APIException(EX.API_REQUEST_FAILED, '获取用户信息失败');
return {
userId,
userId: userResult.data.id,
accessToken: access_token,
refreshToken: refresh_token,
refreshTime: util.unixTimestamp() + ACCESS_TOKEN_EXPIRES
@ -96,6 +105,7 @@ async function requestToken(refreshToken: string) {
return result;
})
.catch(err => {
logger.error(err);
if (accessTokenRequestQueueMap[refreshToken]) {
accessTokenRequestQueueMap[refreshToken].forEach(resolve => resolve(err));
delete accessTokenRequestQueueMap[refreshToken];
@ -128,20 +138,32 @@ async function acquireToken(refreshToken: string): Promise<any> {
}
/**
*
*
* @param refreshToken access_token的refresh_token
*
*/
async function getUserInfo(accessToken: string, refreshToken: string) {
const result = await axios.get('https://kimi.moonshot.cn/api/user', {
export async function request(
method: string,
uri: string,
refreshToken: string,
options: AxiosRequestConfig = {}
) {
const {
accessToken,
userId
} = await acquireToken(refreshToken);
logger.info(`url: ${uri}`);
const result = await axios({
method,
url: `${BASE_URL}${uri}`,
params: options.params,
data: options.data,
headers: {
Authorization: `Bearer ${accessToken}`,
Referer: 'https://kimi.moonshot.cn/',
'X-Msh-Platform': 'web',
'X-Traffic-Id': `7${util.generateRandomString({ length: 18, charset: 'numeric' })}`,
...FAKE_HEADERS
'X-Traffic-Id': userId,
...FAKE_HEADERS,
...(options.headers || {})
},
timeout: 15000,
timeout: options.timeout || 15000,
responseType: options.responseType,
validateStatus: () => true
});
return checkResult(result, refreshToken);
@ -156,28 +178,15 @@ async function getUserInfo(accessToken: string, refreshToken: string) {
*/
async function createConversation(model: string, name: string, refreshToken: string) {
const {
accessToken,
userId
} = await acquireToken(refreshToken);
const result = await axios.post('https://kimi.moonshot.cn/api/chat', {
born_from: '',
id: convId
} = await request('POST', '/api/chat', refreshToken, {
data: {
enter_method: 'new_chat',
is_example: false,
kimiplus_id: /^[0-9a-z]{20}$/.test(model) ? model : 'kimi',
name
}, {
headers: {
Authorization: `Bearer ${accessToken}`,
Referer: 'https://kimi.moonshot.cn/',
'X-Msh-Platform': 'web',
'X-Traffic-Id': userId,
...FAKE_HEADERS
},
timeout: 15000,
validateStatus: () => true
}
});
const {
id: convId
} = checkResult(result, refreshToken);
return convId;
}
@ -189,51 +198,77 @@ async function createConversation(model: string, name: string, refreshToken: str
* @param refreshToken access_token的refresh_token
*/
async function removeConversation(convId: string, refreshToken: string) {
const {
accessToken,
userId
} = await acquireToken(refreshToken);
const result = await axios.delete(`https://kimi.moonshot.cn/api/chat/${convId}`, {
headers: {
Authorization: `Bearer ${accessToken}`,
Referer: `https://kimi.moonshot.cn/chat/${convId}`,
'X-Msh-Platform': 'web',
'X-Traffic-Id': userId,
...FAKE_HEADERS
},
timeout: 15000,
validateStatus: () => true
});
checkResult(result, refreshToken);
return await request('DELETE', `/api/chat/${convId}`, refreshToken);
}
/**
* prompt片段提交
*
*
* @param query prompt
* @param refreshToken access_token的refresh_token
*/
async function promptSnippetSubmit(query: string, refreshToken: string) {
const {
accessToken,
userId
} = await acquireToken(refreshToken);
const result = await axios.post('https://kimi.moonshot.cn/api/prompt-snippet/instance', {
"offset": 0,
"size": 10,
"query": query.replace('user:', '').replace('assistant:', '')
}, {
headers: {
Authorization: `Bearer ${accessToken}`,
Referer: 'https://kimi.moonshot.cn/',
'X-Msh-Platform': 'web',
'X-Traffic-Id': userId,
...FAKE_HEADERS
},
timeout: 15000,
validateStatus: () => true
async function getSuggestion(query: string, refreshToken: string) {
return await request('POST', '/api/suggestion', refreshToken, {
data: {
offset: 0,
page_referer: 'chat',
query: query.replace('user:', '').replace('assistant:', ''),
scene: 'first_round',
size: 10
}
});
checkResult(result, refreshToken);
}
/**
* N2S
*
* N2S
*
* @param model
* @param messages gpt系列消息格式
* @param refs ID列表
* @param refreshToken access_token的refresh_token
* @param refConvId ID
*/
async function preN2s(model: string, messages: { role: string, content: string }[], refs: string[], refreshToken: string, refConvId?: string) {
const isSearchModel = model.indexOf('search') != -1;
return await request('POST', `/api/chat/${refConvId}/pre-n2s`, refreshToken, {
data: {
is_pro_search: false,
kimiplus_id: /^[0-9a-z]{20}$/.test(model) ? model : 'kimi',
messages,
refs,
use_search: isSearchModel
}
});
}
/**
* token计数
*
* @param query
* @param refreshToken access_token的refresh_token
* @param refConvId ID
*/
async function tokenSize(query: string, refs: string[], refreshToken: string, refConvId: string) {
return await request('POST', `/api/chat/${refConvId}/token_size`, refreshToken, {
data: {
content: query,
refs: []
}
});
}
/**
* 使
*
* @param refreshToken access_token的refresh_token
*/
async function getResearchUsage(refreshToken: string): Promise<{
remain,
total,
used
}> {
return await request('GET', '/api/chat/research/usage', refreshToken);
}
/**
@ -242,63 +277,111 @@ async function promptSnippetSubmit(query: string, refreshToken: string) {
* @param model
* @param messages gpt系列消息格式
* @param refreshToken access_token的refresh_token
* @param useSearch
* @param refConvId ID
* @param retryCount
*/
async function createCompletion(model = MODEL_NAME, messages: any[], refreshToken: string, useSearch = true, refConvId?: string, retryCount = 0) {
async function createCompletion(model = MODEL_NAME, messages: any[], refreshToken: string, refConvId?: string, retryCount = 0, segmentId?: string): Promise<IStreamMessage> {
return (async () => {
logger.info(messages);
// 创建会话
const convId = /[0-9a-zA-Z]{20}/.test(refConvId) ? refConvId : await createConversation(model, "未命名会话", refreshToken);
// 提取引用文件URL并上传kimi获得引用的文件ID列表
const refFileUrls = extractRefFileUrls(messages);
const refs = refFileUrls.length ? await Promise.all(refFileUrls.map(fileUrl => uploadFile(fileUrl, refreshToken))) : [];
const refResults = refFileUrls.length ? await Promise.all(refFileUrls.map(fileUrl => uploadFile(fileUrl, refreshToken, convId))) : [];
const refs = refResults.map(result => result.id);
const refsFile = refResults.map(result => ({
detail: result,
done: true,
file: {},
file_info: result,
id: result.id,
name: result.name,
parse_status: 'success',
size: result.size,
upload_progress: 100,
upload_status: 'success'
}));
// 伪装调用获取用户信息
fakeRequest(refreshToken)
.catch(err => logger.error(err));
// 创建会话
const convId = /[0-9a-zA-Z]{20}/.test(refConvId) ? refConvId : await createConversation(model, "未命名会话", refreshToken);
// 请求流
const {
accessToken,
userId
} = await acquireToken(refreshToken);
// 消息预处理
const sendMessages = messagesPrepare(messages, !!refConvId);
const result = await axios.post(`https://kimi.moonshot.cn/api/chat/${convId}/completion/stream`, {
kimiplus_id: /^[0-9a-z]{20}$/.test(model) ? model : 'kimi',
!segmentId && preN2s(model, sendMessages, refs, refreshToken, convId)
.catch(err => logger.error(err));
getSuggestion(sendMessages[0].content, refreshToken)
.catch(err => logger.error(err));
tokenSize(sendMessages[0].content, refs, refreshToken, convId)
.catch(err => logger.error(err));
const isMath = model.indexOf('math') != -1;
const isSearchModel = model.indexOf('search') != -1;
const isResearchModel = model.indexOf('research') != -1;
const isK1Model = model.indexOf('k1') != -1;
logger.info(`使用模型: ${model},是否联网检索: ${isSearchModel},是否探索版: ${isResearchModel}是否K1模型: ${isK1Model},是否数学模型: ${isMath}`);
if(segmentId)
logger.info(`继续请求segmentId: ${segmentId}`);
// 检查探索版使用量
if(isResearchModel) {
const {
total,
used
} = await getResearchUsage(refreshToken);
if(used >= total)
throw new APIException(EX.API_RESEARCH_EXCEEDS_LIMIT, `探索版使用量已达到上限`);
logger.info(`探索版当前额度: ${used}/${total}`);
}
const kimiplusId = isK1Model ? 'crm40ee9e5jvhsn7ptcg' : (/^[0-9a-z]{20}$/.test(model) ? model : 'kimi');
// 请求补全流
const stream = await request('POST', `/api/chat/${convId}/completion/stream`, refreshToken, {
data: segmentId ? {
segment_id: segmentId,
action: 'continue',
messages: [{ role: 'user', content: ' ' }],
kimiplus_id: kimiplusId,
extend: { sidebar: true }
} : {
kimiplus_id: kimiplusId,
messages: sendMessages,
refs,
is_pro_search: false,
use_search: useSearch
}, {
headers: {
Authorization: `Bearer ${accessToken}`,
Referer: `https://kimi.moonshot.cn/chat/${convId}`,
'Priority': 'u=1, i',
'X-Msh-Platform': 'web',
'X-Traffic-Id': userId,
...FAKE_HEADERS
refs_file: refsFile,
use_math: isMath,
use_research: isResearchModel,
use_search: isSearchModel,
extend: { sidebar: true }
},
headers: {
Referer: `https://kimi.moonshot.cn/chat/${convId}`
},
// 120秒超时
timeout: 120000,
validateStatus: () => true,
responseType: 'stream'
});
const streamStartTime = util.timestamp();
// 接收流为输出文本
const answer = await receiveStream(model, convId, result.data);
const answer = await receiveStream(model, convId, stream);
// 如果上次请求生成长度超限,则继续请求
if(answer.choices[0].finish_reason == 'length' && answer.segment_id) {
const continueAnswer = await createCompletion(model, [], refreshToken, convId, retryCount, answer.segment_id);
answer.choices[0].message.content += continueAnswer.choices[0].message.content;
}
logger.success(`Stream has completed transfer ${util.timestamp() - streamStartTime}ms`);
// 异步移除会话,如果消息不合规,此操作可能会抛出数据库错误异常,请忽略
// 如果引用会话将不会清除,因为我们不知道什么时候你会结束会话
!refConvId && removeConversation(convId, refreshToken)
.catch(err => console.error(err));
promptSnippetSubmit(sendMessages[0].content, refreshToken)
.catch(err => console.error(err));
return answer;
})()
@ -308,7 +391,7 @@ async function createCompletion(model = MODEL_NAME, messages: any[], refreshToke
logger.warn(`Try again after ${RETRY_DELAY / 1000}s...`);
return (async () => {
await new Promise(resolve => setTimeout(resolve, RETRY_DELAY));
return createCompletion(model, messages, refreshToken, useSearch, refConvId, retryCount + 1);
return createCompletion(model, messages, refreshToken, refConvId, retryCount + 1);
})();
}
throw err;
@ -321,60 +404,92 @@ async function createCompletion(model = MODEL_NAME, messages: any[], refreshToke
* @param model
* @param messages gpt系列消息格式
* @param refreshToken access_token的refresh_token
* @param useSearch
* @param refConvId ID
* @param retryCount
*/
async function createCompletionStream(model = MODEL_NAME, messages: any[], refreshToken: string, useSearch = true, refConvId?: string, retryCount = 0) {
async function createCompletionStream(model = MODEL_NAME, messages: any[], refreshToken: string, refConvId?: string, retryCount = 0) {
return (async () => {
logger.info(messages);
// 创建会话
const convId = /[0-9a-zA-Z]{20}/.test(refConvId) ? refConvId : await createConversation(model, "未命名会话", refreshToken);
// 提取引用文件URL并上传kimi获得引用的文件ID列表
const refFileUrls = extractRefFileUrls(messages);
const refs = refFileUrls.length ? await Promise.all(refFileUrls.map(fileUrl => uploadFile(fileUrl, refreshToken))) : [];
const refResults = refFileUrls.length ? await Promise.all(refFileUrls.map(fileUrl => uploadFile(fileUrl, refreshToken, convId))) : [];
const refs = refResults.map(result => result.id);
const refsFile = refResults.map(result => ({
detail: result,
done: true,
file: {},
file_info: result,
id: result.id,
name: result.name,
parse_status: 'success',
size: result.size,
upload_progress: 100,
upload_status: 'success'
}));
// 伪装调用获取用户信息
fakeRequest(refreshToken)
.catch(err => logger.error(err));
// 创建会话
const convId = /[0-9a-zA-Z]{20}/.test(refConvId) ? refConvId : await createConversation(model, "未命名会话", refreshToken);
// 请求流
const {
accessToken,
userId
} = await acquireToken(refreshToken);
const sendMessages = messagesPrepare(messages, !!refConvId);
const result = await axios.post(`https://kimi.moonshot.cn/api/chat/${convId}/completion/stream`, {
kimiplus_id: /^[0-9a-z]{20}$/.test(model) ? model : undefined,
preN2s(model, sendMessages, refs, refreshToken, convId)
.catch(err => logger.error(err));
getSuggestion(sendMessages[0].content, refreshToken)
.catch(err => logger.error(err));
tokenSize(sendMessages[0].content, refs, refreshToken, convId)
.catch(err => logger.error(err));
const isMath = model.indexOf('math') != -1;
const isSearchModel = model.indexOf('search') != -1;
const isResearchModel = model.indexOf('research') != -1;
const isK1Model = model.indexOf('k1') != -1;
logger.info(`使用模型: ${model},是否联网检索: ${isSearchModel},是否探索版: ${isResearchModel}是否K1模型: ${isK1Model},是否数学模型: ${isMath}`);
// 检查探索版使用量
if(isResearchModel) {
const {
total,
used
} = await getResearchUsage(refreshToken);
if(used >= total)
throw new APIException(EX.API_RESEARCH_EXCEEDS_LIMIT, `探索版使用量已达到上限`);
logger.info(`探索版当前额度: ${used}/${total}`);
}
const kimiplusId = isK1Model ? 'crm40ee9e5jvhsn7ptcg' : (/^[0-9a-z]{20}$/.test(model) ? model : 'kimi');
// 请求补全流
const stream = await request('POST', `/api/chat/${convId}/completion/stream`, refreshToken, {
data: {
kimiplus_id: kimiplusId,
messages: sendMessages,
refs,
use_search: useSearch
}, {
// 120秒超时
timeout: 120000,
headers: {
Authorization: `Bearer ${accessToken}`,
Referer: `https://kimi.moonshot.cn/chat/${convId}`,
'Priority': 'u=1, i',
'X-Msh-Platform': 'web',
'X-Traffic-Id': userId,
...FAKE_HEADERS
refs_file: refsFile,
use_math: isMath,
use_research: isResearchModel,
use_search: isSearchModel,
extend: { sidebar: true }
},
headers: {
Referer: `https://kimi.moonshot.cn/chat/${convId}`
},
validateStatus: () => true,
responseType: 'stream'
});
const streamStartTime = util.timestamp();
// 创建转换流将消息格式转换为gpt兼容格式
return createTransStream(model, convId, result.data, () => {
return createTransStream(model, convId, stream, () => {
logger.success(`Stream has completed transfer ${util.timestamp() - streamStartTime}ms`);
// 流传输结束后异步移除会话,如果消息不合规,此操作可能会抛出数据库错误异常,请忽略
// 如果引用会话将不会清除,因为我们不知道什么时候你会结束会话
!refConvId && removeConversation(convId, refreshToken)
.catch(err => console.error(err));
promptSnippetSubmit(sendMessages[0].content, refreshToken)
.catch(err => console.error(err));
});
})()
.catch(err => {
@ -383,7 +498,7 @@ async function createCompletionStream(model = MODEL_NAME, messages: any[], refre
logger.warn(`Try again after ${RETRY_DELAY / 1000}s...`);
return (async () => {
await new Promise(resolve => setTimeout(resolve, RETRY_DELAY));
return createCompletionStream(model, messages, refreshToken, useSearch, refConvId, retryCount + 1);
return createCompletionStream(model, messages, refreshToken, refConvId, retryCount + 1);
})();
}
throw err;
@ -398,33 +513,29 @@ async function createCompletionStream(model = MODEL_NAME, messages: any[], refre
* @param refreshToken access_token的refresh_token
*/
async function fakeRequest(refreshToken: string) {
const {
accessToken,
userId
} = await acquireToken(refreshToken);
const options = {
headers: {
Authorization: `Bearer ${accessToken}`,
Referer: `https://kimi.moonshot.cn/`,
'X-Msh-Platform': 'web',
'X-Traffic-Id': userId,
...FAKE_HEADERS
}
};
await [
() => axios.get('https://kimi.moonshot.cn/api/user', options),
() => axios.get('https://kimi.moonshot.cn/api/chat_1m/user/status', options),
() => axios.post('https://kimi.moonshot.cn/api/chat/list', {
() => request('GET', '/api/user', refreshToken),
() => request('POST', '/api/user/usage', refreshToken, {
data: {
usage: ['kimiv', 'math']
}
}),
() => request('GET', '/api/chat_1m/user/status', refreshToken),
() => request('GET', '/api/kimi_mv/user/status', refreshToken),
() => request('POST', '/api/kimiplus/history', refreshToken),
() => request('POST', '/api/kimiplus/search', refreshToken, {
data: {
offset: 0,
size: 20
}
}),
() => request('POST', '/api/chat/list', refreshToken, {
data: {
offset: 0,
size: 50
}, options),
() => axios.post('https://kimi.moonshot.cn/api/show_case/list', {
offset: 0,
size: 4,
enable_cache: true,
order: "asc"
}, options)
][Math.floor(Math.random() * 4)]();
}
}),
][Math.floor(Math.random() * 7)]();
}
/**
@ -536,20 +647,19 @@ function wrapUrlsToTags(content: string) {
* @param filename
* @param refreshToken access_token的refresh_token
*/
async function preSignUrl(filename: string, refreshToken: string) {
async function preSignUrl(action: string, filename: string, refreshToken: string) {
const {
accessToken,
userId
} = await acquireToken(refreshToken);
const result = await axios.post('https://kimi.moonshot.cn/api/pre-sign-url', {
action: 'file',
action,
name: filename
}, {
timeout: 15000,
headers: {
Authorization: `Bearer ${accessToken}`,
Referer: `https://kimi.moonshot.cn/`,
'X-Msh-Platform': 'web',
'X-Traffic-Id': userId,
...FAKE_HEADERS
},
@ -585,8 +695,9 @@ async function checkFileUrl(fileUrl: string) {
*
* @param fileUrl URL
* @param refreshToken access_token的refresh_token
* @param refConvId ID
*/
async function uploadFile(fileUrl: string, refreshToken: string) {
async function uploadFile(fileUrl: string, refreshToken: string, refConvId?: string) {
// 预检查远程文件URL可用性
await checkFileUrl(fileUrl);
@ -610,11 +721,14 @@ async function uploadFile(fileUrl: string, refreshToken: string) {
}));
}
const fileType = (mimeType || '').includes('image') ? 'image' : 'file';
// 获取预签名文件URL
const {
let {
url: uploadUrl,
object_name: objectName
} = await preSignUrl(filename, refreshToken);
object_name: objectName,
file_id: fileId
} = await preSignUrl(fileType, filename, refreshToken);
// 获取文件的MIME类型
mimeType = mimeType || mime.getType(filename);
@ -635,7 +749,6 @@ async function uploadFile(fileUrl: string, refreshToken: string) {
'Content-Type': mimeType,
Authorization: `Bearer ${accessToken}`,
Referer: `https://kimi.moonshot.cn/`,
'X-Msh-Platform': 'web',
'X-Traffic-Id': userId,
...FAKE_HEADERS
},
@ -643,30 +756,36 @@ async function uploadFile(fileUrl: string, refreshToken: string) {
});
checkResult(result, refreshToken);
let fileId, status, startTime = Date.now();
while (status != 'initialized') {
let status, startTime = Date.now();
let fileDetail;
while (status != 'initialized' && status != 'parsed') {
if (Date.now() - startTime > 30000)
throw new Error('文件等待处理超时');
// 获取文件上传结果
result = await axios.post('https://kimi.moonshot.cn/api/file', {
result = await axios.post('https://kimi.moonshot.cn/api/file', fileType == 'image' ? {
type: 'image',
file_id: fileId,
name: filename
} : {
type: 'file',
name: filename,
object_name: objectName,
timeout: 15000
file_id: '',
chat_id: refConvId
}, {
headers: {
Authorization: `Bearer ${accessToken}`,
Referer: `https://kimi.moonshot.cn/`,
'X-Msh-Platform': 'web',
'X-Traffic-Id': userId,
...FAKE_HEADERS
}
});
({ id: fileId, status } = checkResult(result, refreshToken));
fileDetail = checkResult(result, refreshToken);
({ id: fileId, status } = fileDetail);
}
startTime = Date.now();
let parseFinish = false;
let parseFinish = status == 'parsed';
while (!parseFinish) {
if (Date.now() - startTime > 30000)
throw new Error('文件等待处理超时');
@ -679,7 +798,6 @@ async function uploadFile(fileUrl: string, refreshToken: string) {
headers: {
Authorization: `Bearer ${accessToken}`,
Referer: `https://kimi.moonshot.cn/`,
'X-Msh-Platform': 'web',
'X-Traffic-Id': userId,
...FAKE_HEADERS
}
@ -689,7 +807,7 @@ async function uploadFile(fileUrl: string, refreshToken: string) {
});
}
return fileId;
return fileDetail;
}
/**
@ -722,8 +840,9 @@ function checkResult(result: AxiosResponse, refreshToken: string) {
* @param convId ID
* @param stream
*/
async function receiveStream(model: string, convId: string, stream: any) {
async function receiveStream(model: string, convId: string, stream: any): Promise<IStreamMessage> {
let webSearchCount = 0;
let temp = Buffer.from('');
return new Promise((resolve, reject) => {
// 消息初始化
const data = {
@ -734,10 +853,11 @@ async function receiveStream(model: string, convId: string, stream: any) {
{ index: 0, message: { role: 'assistant', content: '' }, finish_reason: 'stop' }
],
usage: { prompt_tokens: 1, completion_tokens: 1, total_tokens: 2 },
segment_id: '',
created: util.unixTimestamp()
};
let refContent = '';
const silentSearch = model.indexOf('silent_search') != -1;
const silentSearch = model.indexOf('silent') != -1;
const parser = createParser(event => {
try {
if (event.type !== "event") return;
@ -747,8 +867,16 @@ async function receiveStream(model: string, convId: string, stream: any) {
throw new Error(`Stream response invalid: ${event.data}`);
// 处理消息
if (result.event == 'cmpl' && result.text) {
const exceptCharIndex = result.text.indexOf("<22>");
data.choices[0].message.content += result.text.substring(0, exceptCharIndex == -1 ? result.text.length : exceptCharIndex);
data.choices[0].message.content += result.text;
}
// 处理请求ID
else if(result.event == 'req') {
data.segment_id = result.id;
}
// 处理超长文本
else if(result.event == 'length') {
logger.warn('此次生成达到max_tokens稍候将继续请求拼接完整响应');
data.choices[0].finish_reason = 'length';
}
// 处理结束或错误
else if (result.event == 'all_done' || result.event == 'error') {
@ -757,7 +885,7 @@ async function receiveStream(model: string, convId: string, stream: any) {
resolve(data);
}
// 处理联网搜索
else if (!silentSearch && result.event == 'search_plus' && result.msg && result.msg.type == 'get_res'){
else if (!silentSearch && result.event == 'search_plus' && result.msg && result.msg.type == 'get_res') {
webSearchCount += 1;
refContent += `【检索 ${webSearchCount}】 [${result.msg.title}](${result.msg.url})\n\n`;
}
@ -770,7 +898,20 @@ async function receiveStream(model: string, convId: string, stream: any) {
}
});
// 将流数据喂给SSE转换器
stream.on("data", buffer => parser.feed(buffer.toString()));
stream.on("data", buffer => {
// 检查buffer是否以完整UTF8字符结尾
if (buffer.toString().indexOf('<27>') != -1) {
// 如果不完整则累积buffer直到收到完整字符
temp = Buffer.concat([temp, buffer]);
return;
}
// 将之前累积的不完整buffer拼接
if (temp.length > 0) {
buffer = Buffer.concat([temp, buffer]);
temp = Buffer.from('');
}
parser.feed(buffer.toString());
});
stream.once("error", err => reject(err));
stream.once("close", () => resolve(data));
});
@ -793,7 +934,9 @@ function createTransStream(model: string, convId: string, stream: any, endCallba
const transStream = new PassThrough();
let webSearchCount = 0;
let searchFlag = false;
const silentSearch = model.indexOf('silent_search') != -1;
let lengthExceed = false;
let segmentId = '';
const silentSearch = model.indexOf('silent') != -1;
!transStream.closed && transStream.write(`data: ${JSON.stringify({
id: convId,
model,
@ -801,6 +944,7 @@ function createTransStream(model: string, convId: string, stream: any, endCallba
choices: [
{ index: 0, delta: { role: 'assistant', content: '' }, finish_reason: null }
],
segment_id: '',
created
})}\n\n`);
const parser = createParser(event => {
@ -821,12 +965,21 @@ function createTransStream(model: string, convId: string, stream: any, endCallba
choices: [
{ index: 0, delta: { content: (searchFlag ? '\n' : '') + chunk }, finish_reason: null }
],
segment_id: segmentId,
created
})}\n\n`;
if (searchFlag)
searchFlag = false;
!transStream.closed && transStream.write(data);
}
// 处理请求ID
else if(result.event == 'req') {
segmentId = result.id;
}
// 处理超长文本
else if (result.event == 'length') {
lengthExceed = true;
}
// 处理结束或错误
else if (result.event == 'all_done' || result.event == 'error') {
const data = `data: ${JSON.stringify({
@ -837,10 +990,11 @@ function createTransStream(model: string, convId: string, stream: any, endCallba
{
index: 0, delta: result.event == 'error' ? {
content: '\n[内容由于不合规被停止生成,我们换个话题吧]'
} : {}, finish_reason: 'stop'
} : {}, finish_reason: lengthExceed ? 'length' : 'stop'
}
],
usage: { prompt_tokens: 1, completion_tokens: 1, total_tokens: 2 },
segment_id: segmentId,
created
})}\n\n`;
!transStream.closed && transStream.write(data);
@ -863,6 +1017,7 @@ function createTransStream(model: string, convId: string, stream: any, endCallba
}, finish_reason: null
}
],
segment_id: segmentId,
created
})}\n\n`;
!transStream.closed && transStream.write(data);

View File

@ -0,0 +1,20 @@
export default interface IStreamMessage {
id: string;
model: string;
object: string;
choices: {
index: number;
message: {
role: string;
content: string;
};
finish_reason: string;
}[];
usage: {
prompt_tokens: number;
completion_tokens: number;
total_tokens: number;
};
segment_id?: string;
created: number;
}

View File

@ -20,15 +20,19 @@ export default {
const tokens = chat.tokenSplit(request.headers.authorization);
// 随机挑选一个refresh_token
const token = _.sample(tokens);
const { model, conversation_id: convId, messages, stream, use_search } = request.body;
let { model, conversation_id: convId, messages, stream, use_search } = request.body;
if(use_search)
model = 'kimi-search';
if (stream) {
const stream = await chat.createCompletionStream(model, messages, token, use_search, convId);
const stream = await chat.createCompletionStream(model, messages, token, convId);
return new Response(stream, {
type: "text/event-stream"
});
}
else
return await chat.createCompletion(model, messages, token, use_search, convId);
return await chat.createCompletion(model, messages, token, convId);
}
}

1555
yarn.lock Normal file

File diff suppressed because it is too large Load Diff