diff --git a/docSite/content/zh-cn/docs/development/upgrading/490.md b/docSite/content/zh-cn/docs/development/upgrading/490.md index 45cb07cf0..dfa63aaee 100644 --- a/docSite/content/zh-cn/docs/development/upgrading/490.md +++ b/docSite/content/zh-cn/docs/development/upgrading/490.md @@ -47,6 +47,7 @@ curl --location --request POST 'https://{{host}}/api/admin/initv490' \ 1. 知识库数据不再限制索引数量，可无限自定义。同时可自动更新输入文本的索引，不影响自定义索引。 2. Markdown 解析，增加链接后中文标点符号检测，增加空格。 3. Prompt 模式工具调用，支持思考模型。同时优化其格式检测，减少空输出的概率。 +4. Mongo 文件读取流合并，减少计算量。同时优化存储 chunks，极大提高大文件读取速度。50M PDF 读取时间提高 3 倍。 ## 🐛 修复 diff --git a/packages/service/common/file/gridfs/controller.ts b/packages/service/common/file/gridfs/controller.ts index e77f93df7..f809cc772 100644 --- a/packages/service/common/file/gridfs/controller.ts +++ b/packages/service/common/file/gridfs/controller.ts @@ -52,7 +52,9 @@ export async function uploadFile({ const stats = await fsp.stat(path); if (!stats.isFile()) return Promise.reject(`${path} is not a file`); - const readStream = fs.createReadStream(path); + const readStream = fs.createReadStream(path, { + highWaterMark: 256 * 1024 + }); // Add default metadata metadata.teamId = teamId; @@ -62,9 +64,27 @@ export async function uploadFile({ // create a gridfs bucket const bucket = getGridBucket(bucketName); + const fileSize = stats.size; + const chunkSizeBytes = (() => { + // 计算理想块大小：文件大小 ÷ 目标块数(10) + const idealChunkSize = Math.ceil(fileSize / 10); + + // 确保块大小至少为512KB + const minChunkSize = 512 * 1024; // 512KB + + // 取理想块大小和最小块大小中的较大值 + let chunkSize = Math.max(idealChunkSize, minChunkSize); + + // 将块大小向上取整到最接近的64KB的倍数，使其更整齐 + chunkSize = Math.ceil(chunkSize / (64 * 1024)) * (64 * 1024); + + return chunkSize; + })(); + const stream = bucket.openUploadStream(filename, { metadata, - contentType + contentType, + chunkSizeBytes }); // save to gridfs diff --git a/packages/service/common/file/gridfs/utils.ts b/packages/service/common/file/gridfs/utils.ts index 1e0725e75..9e376f28f 100644 --- a/packages/service/common/file/gridfs/utils.ts +++ b/packages/service/common/file/gridfs/utils.ts @@ -3,15 +3,13 @@ import { PassThrough } from 'stream'; export const gridFsStream2Buffer = (stream: NodeJS.ReadableStream) => { return new Promise((resolve, reject) => { - const chunks: Buffer[] = []; - let totalLength = 0; + const chunks: Uint8Array[] = []; stream.on('data', (chunk) => { chunks.push(chunk); - totalLength += chunk.length; }); stream.on('end', () => { - const resultBuffer = Buffer.concat(chunks, totalLength); // 一次性拼接 + const resultBuffer = Buffer.concat(chunks); // 一次性拼接 resolve(resultBuffer); }); stream.on('error', (err) => { @@ -21,25 +19,26 @@ export const gridFsStream2Buffer = (stream: NodeJS.ReadableStream) => { }; export const stream2Encoding = async (stream: NodeJS.ReadableStream) => { - const start = Date.now(); const copyStream = stream.pipe(new PassThrough()); /* get encoding */ const buffer = await (() => { return new Promise((resolve, reject) => { - let tmpBuffer: Buffer = Buffer.from([]); + const chunks: Uint8Array[] = []; + let totalLength = 0; stream.on('data', (chunk) => { - if (tmpBuffer.length < 200) { - tmpBuffer = Buffer.concat([tmpBuffer, chunk]); + if (totalLength < 200) { + chunks.push(chunk); + totalLength += chunk.length; - if (tmpBuffer.length >= 200) { - resolve(tmpBuffer); + if (totalLength >= 200) { + resolve(Buffer.concat(chunks)); } } }); stream.on('end', () => { - resolve(tmpBuffer); + resolve(Buffer.concat(chunks)); }); stream.on('error', (err) => { reject(err); diff --git a/packages/service/core/ai/audio/speech.ts b/packages/service/core/ai/audio/speech.ts index 3d82f68c6..0bacaae4e 100644 --- a/packages/service/core/ai/audio/speech.ts +++ b/packages/service/core/ai/audio/speech.ts @@ -43,13 +43,13 @@ export async function text2Speech({ const readableStream = response.body as unknown as NodeJS.ReadableStream; readableStream.pipe(res); - let bufferStore = Buffer.from([]); + const chunks: Uint8Array[] = []; readableStream.on('data', (chunk) => { - bufferStore = Buffer.concat([bufferStore, chunk]); + chunks.push(chunk); }); readableStream.on('end', () => { - onSuccess({ model, buffer: bufferStore }); + onSuccess({ model, buffer: Buffer.concat(chunks) }); }); readableStream.on('error', (e) => { onError(e); diff --git a/packages/service/core/ai/config/provider/DeepSeek.json b/packages/service/core/ai/config/provider/DeepSeek.json index df9369e38..91b748e85 100644 --- a/packages/service/core/ai/config/provider/DeepSeek.json +++ b/packages/service/core/ai/config/provider/DeepSeek.json @@ -46,8 +46,8 @@ "defaultConfig": {}, "fieldMap": {}, "type": "llm", - "showTopP": true, - "showStopSign": true + "showTopP": false, + "showStopSign": false } ] } diff --git a/packages/web/i18n/en/account_model.json b/packages/web/i18n/en/account_model.json index d409b97fd..f7143cfdb 100644 --- a/packages/web/i18n/en/account_model.json +++ b/packages/web/i18n/en/account_model.json @@ -21,6 +21,7 @@ "edit_channel": "Channel configuration", "enable_channel": "Enable", "forbid_channel": "Disabled", + "input maxToken_tip": "The model max_tokens parameter, if left blank, means that the model does not support it.", "key_type": "API key format:", "log": "Call log", "log_detail": "Log details", @@ -28,6 +29,7 @@ "log_status": "Status", "mapping": "Model Mapping", "mapping_tip": "A valid Json is required. \nThe model can be mapped when sending a request to the actual address. \nFor example:\n{\n \n \"gpt-4o\": \"gpt-4o-test\"\n\n}\n\nWhen FastGPT requests the gpt-4o model, the gpt-4o-test model is sent to the actual address, instead of gpt-4o.", + "max_temperature_tip": "If the model temperature parameter is not filled in, it means that the model does not support the temperature parameter.", "model": "Model", "model_name": "Model name", "model_test": "Model testing", diff --git a/packages/web/i18n/zh-CN/account_model.json b/packages/web/i18n/zh-CN/account_model.json index 73c3fced9..4626d717b 100644 --- a/packages/web/i18n/zh-CN/account_model.json +++ b/packages/web/i18n/zh-CN/account_model.json @@ -21,6 +21,7 @@ "edit_channel": "渠道配置", "enable_channel": "启用", "forbid_channel": "禁用", + "input maxToken_tip": "模型 max_tokens 参数，如果留空，则代表模型不支持该参数。", "key_type": "API key 格式: ", "log": "调用日志", "log_detail": "日志详情", @@ -28,6 +29,7 @@ "log_status": "状态", "mapping": "模型映射", "mapping_tip": "需填写一个有效 Json。可在向实际地址发送请求时，对模型进行映射。例如：\n{\n \"gpt-4o\": \"gpt-4o-test\"\n}\n当 FastGPT 请求 gpt-4o 模型时，会向实际地址发送 gpt-4o-test 的模型，而不是 gpt-4o。", + "max_temperature_tip": "模型 temperature 参数，不填则代表模型不支持 temperature 参数。", "model": "模型", "model_name": "模型名", "model_test": "模型测试", diff --git a/packages/web/i18n/zh-Hant/account_model.json b/packages/web/i18n/zh-Hant/account_model.json index b62b937d6..cf3803b37 100644 --- a/packages/web/i18n/zh-Hant/account_model.json +++ b/packages/web/i18n/zh-Hant/account_model.json @@ -19,6 +19,7 @@ "edit_channel": "渠道配置", "enable_channel": "啟用", "forbid_channel": "禁用", + "input maxToken_tip": "模型 max_tokens 參數，如果留空，則代表模型不支持該參數。", "key_type": "API key 格式:", "log": "調用日誌", "log_detail": "日誌詳情", @@ -26,6 +27,7 @@ "log_status": "狀態", "mapping": "模型映射", "mapping_tip": "需填寫一個有效 Json。\n可在向實際地址發送請求時，對模型進行映射。\n例如：\n{\n \n \"gpt-4o\": \"gpt-4o-test\"\n\n}\n\n當 FastGPT 請求 gpt-4o 模型時，會向實際地址發送 gpt-4o-test 的模型，而不是 gpt-4o。", + "max_temperature_tip": "模型 temperature 參數，不填則代表模型不支持 temperature 參數。", "model": "模型", "model_name": "模型名", "model_test": "模型測試", diff --git a/projects/app/src/pageComponents/account/model/AddModelBox.tsx b/projects/app/src/pageComponents/account/model/AddModelBox.tsx index 9ae33164f..582bea51c 100644 --- a/projects/app/src/pageComponents/account/model/AddModelBox.tsx +++ b/projects/app/src/pageComponents/account/model/AddModelBox.tsx @@ -38,6 +38,7 @@ import { useSystemStore } from '@/web/common/system/useSystemStore'; import QuestionTip from '@fastgpt/web/components/common/MyTooltip/QuestionTip'; import { Prompt_CQJson, Prompt_ExtractJson } from '@fastgpt/global/core/ai/prompt/agent'; import MyModal from '@fastgpt/web/components/common/MyModal'; +import FormLabel from '@fastgpt/web/components/common/MyBox/FormLabel'; export const AddModelButton = ({ onCreate, @@ -134,6 +135,14 @@ export const ModelEditModal = ({ const { runAsync: updateModel, loading: updatingModel } = useRequest2( async (data: SystemModelItemType) => { + for (const key in data) { + // @ts-ignore + const val = data[key]; + if (val === null || val === undefined || Number.isNaN(val)) { + // @ts-ignore + data[key] = ''; + } + } return putSystemModel({ model: data.model, metadata: data @@ -295,7 +304,16 @@ export const ModelEditModal = ({ {isLLMModel && ( <> - {t('common:core.ai.Max context')} + + + {t('common:core.ai.Max context')} + + - {t('account:model.max_quote')} + + + {t('account:model.max_quote')} + + - {t('common:core.chat.response.module maxToken')} + + + {t('common:core.chat.response.module maxToken')} + + + @@ -329,7 +361,12 @@ export const ModelEditModal = ({ - {t('account:model.max_temperature')} + + + {t('account:model.max_temperature')} + + +