fix: merge conflict

This commit is contained in:
ikechan8370 2023-11-21 15:43:48 +08:00
commit 2b19302b87
14 changed files with 264 additions and 85 deletions

View file

@ -476,7 +476,7 @@ export default class SydneyAIClient {
context += chats
.map(chat => {
let sender = chat.sender || chat || {}
if (chat.raw_message.startsWith('建议的回复')) {
if (chat.raw_message?.startsWith('建议的回复')) {
// 建议的回复太容易污染设定导致对话太固定跑偏了
return ''
}

34
utils/chat.js Normal file
View file

@ -0,0 +1,34 @@
export async function getChatHistoryGroup (e, num) {
// if (e.adapter === 'shamrock') {
// return await e.group.getChatHistory(0, num, false)
// } else {
let latestChats = await e.group.getChatHistory(0, 1)
if (latestChats.length > 0) {
let latestChat = latestChats[0]
if (latestChat) {
let seq = latestChat.seq || latestChat.message_id
let chats = []
while (chats.length < num) {
let chatHistory = await e.group.getChatHistory(seq, 20)
chats.push(...chatHistory)
seq = chatHistory[0].seq || chatHistory[0].message_id
}
chats = chats.slice(0, num)
try {
let mm = await e.group.getMemberMap()
chats.forEach(chat => {
let sender = mm.get(chat.sender.user_id)
if (sender) {
chat.sender = sender
}
})
} catch (err) {
logger.warn(err)
}
// console.log(chats)
return chats
}
}
// }
return []
}

View file

@ -832,8 +832,10 @@ export function getMaxModelTokens (model = 'gpt-3.5-turbo') {
if (model.startsWith('gpt-3.5-turbo')) {
if (model.includes('16k')) {
return 16000
} else {
} else if (model.includes('0613') || model.includes('0314')) {
return 4000
} else {
return 16000
}
} else {
if (model.includes('32k')) {
@ -870,7 +872,7 @@ export async function generateAudio (e, pendingText, speakingEmotion, emotionDeg
if (!Config.ttsSpace && !Config.azureTTSKey && !Config.voicevoxSpace) return false
let wav
const speaker = getUserSpeaker(await getUserReplySetting(e))
let ignoreEncode = getUin(e) === 88888
let ignoreEncode = e.adapter === 'shamrock'
try {
if (Config.ttsMode === 'vits-uma-genshin-honkai' && Config.ttsSpace) {
if (Config.autoJapanese) {

View file

@ -102,7 +102,12 @@ export class SendAudioMessageTool extends AbstractTool {
return `audio generation failed, error: ${JSON.stringify(err)}`
}
if (sendable) {
let groupList = e.bot.gl || new Map()
let groupList
try {
groupList = await e.bot.getGroupList()
} catch (err) {
groupList = e.bot.gl
}
try {
if (groupList.get(target)) {
let group = await e.bot.pickGroup(target)

View file

@ -27,8 +27,12 @@ export class SendAvatarTool extends AbstractTool {
const target = isNaN(targetGroupIdOrQQNumber) || !targetGroupIdOrQQNumber
? defaultTarget
: parseInt(targetGroupIdOrQQNumber) === e.bot.uin ? defaultTarget : parseInt(targetGroupIdOrQQNumber)
let groupList = e.bot.gl || new Map()
let groupList
try {
groupList = await e.bot.getGroupList()
} catch (err) {
groupList = e.bot.gl
}
console.log('sendAvatar', target, pictures)
if (groupList.get(target)) {
let group = await e.bot.pickGroup(target)

View file

@ -24,7 +24,12 @@ export class SendDiceTool extends AbstractTool {
const target = isNaN(targetGroupIdOrQQNumber) || !targetGroupIdOrQQNumber
? defaultTarget
: parseInt(targetGroupIdOrQQNumber) === e.bot.uin ? defaultTarget : parseInt(targetGroupIdOrQQNumber)
let groupList = e.bot.gl || new Map()
let groupList
try {
groupList = await e.bot.getGroupList()
} catch (err) {
groupList = e.bot.gl
}
num = isNaN(num) || !num ? 1 : num > 5 ? 5 : num
if (groupList.get(target)) {
let group = await e.bot.pickGroup(target, true)

View file

@ -25,7 +25,12 @@ export class SendMessageToSpecificGroupOrUserTool extends AbstractTool {
? defaultTarget
: parseInt(targetGroupIdOrQQNumber) === e.bot.uin ? defaultTarget : parseInt(targetGroupIdOrQQNumber)
let groupList = e.bot.gl || new Map()
let groupList
try {
groupList = await e.bot.getGroupList()
} catch (err) {
groupList = e.bot.gl
}
try {
if (groupList.get(target)) {
let group = await e.bot.pickGroup(target)

View file

@ -32,7 +32,12 @@ export class SendPictureTool extends AbstractTool {
let pictures = urlOfPicture.trim().split(' ')
logger.mark('pictures to send: ', pictures)
pictures = pictures.map(img => segment.image(img))
let groupList = e.bot.gl || new Map()
let groupList
try {
groupList = await e.bot.getGroupList()
} catch (err) {
groupList = e.bot.gl
}
try {
if (groupList.get(target)) {
let group = await e.bot.pickGroup(target)

View file

@ -20,7 +20,12 @@ export class SendRPSTool extends AbstractTool {
const target = isNaN(targetGroupIdOrQQNumber) || !targetGroupIdOrQQNumber
? defaultTarget
: parseInt(targetGroupIdOrQQNumber) === e.bot.uin ? defaultTarget : parseInt(targetGroupIdOrQQNumber)
let groupList = await e.bot.getGroupList()
let groupList
try {
groupList = await e.bot.getGroupList()
} catch (err) {
groupList = e.bot.gl
}
if (groupList.get(target)) {
let group = await e.bot.pickGroup(target, true)
await group.sendMsg(segment.rps(num))

View file

@ -10,11 +10,11 @@ try {
}
export class Tokenizer {
async getHistory (groupId, date = new Date(), duration = 0) {
async getHistory (e, groupId, date = new Date(), duration = 0, userId) {
if (!groupId) {
throw new Error('no valid group id')
}
let group = Bot.pickGroup(groupId, true)
let group = e.bot.pickGroup(groupId, true)
let latestChat = await group.getChatHistory(0, 1)
let seq = latestChat[0].seq
let chats = latestChat
@ -41,14 +41,15 @@ export class Tokenizer {
let startOfSpecifiedDate = date.getTime()
// if duration > 0, go back to the specified number of hours
if (duration > 0) {
// duration should be in range [0, 24]
duration = Math.min(duration, 24)
startOfSpecifiedDate = currentTime - (duration * 60 * 60 * 1000)
// duration should be in range [0, 24]
// duration = Math.min(duration, 24)
startOfSpecifiedDate = currentTime - (duration * 60 * 60 * 1000)
}
// Step 4: Get the end of the specified date by adding 24 hours (in milliseconds)
const endOfSpecifiedDate = startOfSpecifiedDate + (24 * 60 * 60 * 1000)
while (isTimestampInDateRange(chats[0]?.time, startOfSpecifiedDate, endOfSpecifiedDate) && isTimestampInDateRange(chats[chats.length - 1]?.time, startOfSpecifiedDate, endOfSpecifiedDate)) {
// Step 4: Get the end of the specified date by current time
const endOfSpecifiedDate = currentTime
while (isTimestampInDateRange(chats[0]?.time, startOfSpecifiedDate, endOfSpecifiedDate) &&
isTimestampInDateRange(chats[chats.length - 1]?.time, startOfSpecifiedDate, endOfSpecifiedDate)) {
let chatHistory = await group.getChatHistory(seq, 20)
if (chatHistory.length === 1) {
if (chats[0].seq === chatHistory[0].seq) {
@ -58,45 +59,51 @@ export class Tokenizer {
}
chats.push(...chatHistory)
chats.sort(compareByTime)
seq = chatHistory[0].seq
seq = chatHistory?.[0]?.seq
if (!seq) {
break
}
if (Config.debug) {
logger.info(`拉取到${chatHistory.length}条聊天记录,当前已累计获取${chats.length}条聊天记录,继续拉...`)
}
}
chats = chats.filter(chat => isTimestampInDateRange(chat.time, startOfSpecifiedDate, endOfSpecifiedDate))
if (userId) {
chats = chats.filter(chat => chat.sender.user_id === userId)
}
return chats
}
async getKeywordTopK (groupId, topK = 100, duration = 0) {
async getKeywordTopK (e, groupId, topK = 100, duration = 0, userId) {
if (!nodejieba) {
throw new Error('未安装node-rs/jieba娱乐功能-词云统计不可用')
}
// duration represents the number of hours to go back, should in range [0, 24]
let chats = await this.getHistory(groupId, new Date(), duration)
let duration_str = duration > 0 ? `${duration}小时` : '今日'
logger.mark(`聊天记录拉取完成,获取到${duration_str}${chats.length}条聊天记录,准备分词中`)
let chats = await this.getHistory(e, groupId, new Date(), duration, userId)
let durationStr = duration > 0 ? `${duration}小时` : '今日'
logger.mark(`聊天记录拉取完成,获取到${durationStr}${chats.length}条聊天记录,准备分词中`)
const _path = process.cwd()
let stopWordsPath = `${_path}/plugins/chatgpt-plugin/utils/wordcloud/cn_stopwords.txt`
const data = fs.readFileSync(stopWordsPath)
const stopWords = String(data)?.split('\n') || []
let chatContent = chats
.map(c => c.message
//只统计文本内容
.filter(item => item.type == 'text')
.map(textItem => `${textItem.text}`)
.join("").trim()
// 只统计文本内容
.filter(item => item.type == 'text')
.map(textItem => `${textItem.text}`)
.join('').trim()
)
.map(c => {
let length = c.length
let threshold = 10
if (length < 100 && length > 50) {
threshold = 6
} else if (length <= 50 && length > 25) {
threshold = 3
} else if (length <= 25) {
threshold = 2
}
// let length = c.length
let threshold = 2
// if (length < 100 && length > 50) {
// threshold = 6
// } else if (length <= 50 && length > 25) {
// threshold = 3
// } else if (length <= 25) {
// threshold = 2
// }
return nodejieba.extract(c, threshold)
})
.reduce((acc, curr) => acc.concat(curr), [])
@ -132,6 +139,85 @@ export class Tokenizer {
}
}
export class ShamrockTokenizer extends Tokenizer {
async getHistory (e, groupId, date = new Date(), duration = 0, userId) {
logger.mark('当前使用Shamrock适配器')
if (!groupId) {
throw new Error('no valid group id')
}
let group = e.bot.pickGroup(groupId, true)
// 直接加大力度
let pageSize = 500
let chats = (await group.getChatHistory(0, pageSize, false)) || []
// Get the current timestamp
let currentTime = date.getTime()
// Step 2: Set the hours, minutes, seconds, and milliseconds to 0
date.setHours(0, 0, 0, 0)
// Step 3: Calculate the timestamp representing the start of the specified date
// duration represents the number of hours to go back
// if duration is 0, keeping the original date (start of today)
let startOfSpecifiedDate = date.getTime()
// if duration > 0, go back to the specified number of hours
if (duration > 0) {
// duration should be in range [0, 24]
// duration = Math.min(duration, 24)
startOfSpecifiedDate = currentTime - (duration * 60 * 60 * 1000)
}
// Step 4: Get the end of the specified date by currentTime
const endOfSpecifiedDate = currentTime
let cursor = chats.length
// -------------------------------------------------------
// | | |
// -------------------------------------------------------
// ^ ^
// long ago cursor+pageSize cursor current
while (isTimestampInDateRange(chats[0]?.time, startOfSpecifiedDate, endOfSpecifiedDate)) {
// 由于Shamrock消息是从最新的开始拉结束时由于动态更新一旦有人发送消息就会立刻停止所以不判断结束时间
// 拉到后面会巨卡所以增大page减少次数
pageSize = Math.floor(Math.max(cursor / 2, pageSize))
cursor = cursor + pageSize
let retries = 3
let chatHistory
while (retries >= 0) {
try {
chatHistory = await group.getChatHistory(0, cursor, false)
break
} catch (err) {
if (retries === 0) {
logger.error(err)
}
retries--
}
}
if (retries < 0) {
logger.warn('拉不动了,就这样吧')
break
}
if (chatHistory.length === 1) {
break
}
if (chatHistory.length === chats.length) {
// 没有了!再拉也没有了
break
}
let oldLength = chats.length
chats = chatHistory
// chats.sort(compareByTime)
if (Config.debug) {
logger.info(`拉取到${chats.length - oldLength}条聊天记录,当前已累计获取${chats.length}条聊天记录,继续拉...`)
}
}
chats = chats.filter(chat => isTimestampInDateRange(chat.time, startOfSpecifiedDate, endOfSpecifiedDate))
if (userId) {
chats = chats.filter(chat => chat.sender.user_id === userId)
}
return chats
}
}
function isTimestampInDateRange (timestamp, startOfSpecifiedDate, endOfSpecifiedDate) {
if (!timestamp) {
return false

View file

@ -1,11 +1,19 @@
import { Tokenizer } from './tokenizer.js'
import { ShamrockTokenizer, Tokenizer } from './tokenizer.js'
import { render } from '../common.js'
export async function makeWordcloud (e, groupId, duration = 0) {
let tokenizer = new Tokenizer()
let topK = await tokenizer.getKeywordTopK(groupId, 100, duration)
export async function makeWordcloud (e, groupId, duration = 0, userId) {
let tokenizer = getTokenizer(e)
let topK = await tokenizer.getKeywordTopK(e, groupId, 100, duration, userId)
let list = JSON.stringify(topK)
// let list = topK
console.log(list)
await render(e, 'chatgpt-plugin', 'wordcloud/index', { list })
logger.info(list)
let img = await render(e, 'chatgpt-plugin', 'wordcloud/index', { list }, { retType: 'base64' })
await e.reply(img, true)
}
function getTokenizer (e) {
if (e.adapter === 'shamrock') {
return new ShamrockTokenizer()
} else {
return new Tokenizer()
}
}