From 42b2046f96c3200517b73b1bf2a6ad6f72fa742b Mon Sep 17 00:00:00 2001 From: Archer <545436317@qq.com> Date: Sat, 8 Feb 2025 10:44:33 +0800 Subject: [PATCH] 4.8.21 feature (#3720) * agent search demo * edit form force close image select * feat: llm params and doubao1.5 * perf: model error tip * fix: template register path * package --- .../content/zh-cn/docs/development/docker.md | 4 +- docSite/content/zh-cn/docs/development/faq.md | 6 +- .../content/zh-cn/docs/development/sealos.md | 4 + .../zh-cn/docs/development/upgrading/4821.md | 20 ++ packages/global/core/ai/model.d.ts | 5 + packages/global/core/app/type.d.ts | 20 +- packages/global/core/workflow/constants.ts | 4 + .../global/core/workflow/runtime/type.d.ts | 4 + .../workflow/template/system/aiChat/index.ts | 27 +- .../core/workflow/template/system/tools.ts | 26 +- .../core/ai/config/provider/DeepSeek.json | 4 +- .../core/ai/config/provider/Doubao.json | 88 ++++++ .../core/ai/config/provider/OpenAI.json | 3 + packages/service/core/ai/config/utils.ts | 9 +- packages/service/core/ai/utils.ts | 22 +- packages/service/core/dataset/search/agent.ts | 277 ++++++++++++++++++ .../service/core/dataset/search/controller.ts | 2 + .../dispatch/agent/runTool/functionCall.ts | 22 +- .../dispatch/agent/runTool/promptCall.ts | 18 +- .../dispatch/agent/runTool/toolChoice.ts | 20 +- .../workflow/dispatch/agent/runTool/type.d.ts | 6 +- .../core/workflow/dispatch/chat/oneapi.ts | 22 +- packages/service/worker/utils.ts | 2 +- packages/templates/register.ts | 7 +- .../common/MySlider/InputSlider.tsx | 2 +- packages/web/i18n/en/account.json | 3 + packages/web/i18n/en/app.json | 4 + packages/web/i18n/en/login.json | 9 +- packages/web/i18n/zh-CN/account.json | 3 + packages/web/i18n/zh-CN/app.json | 4 + packages/web/i18n/zh-CN/login.json | 1 + packages/web/i18n/zh-Hant/account.json | 3 + packages/web/i18n/zh-Hant/app.json | 4 + packages/web/i18n/zh-Hant/login.json | 1 + projects/app/package.json | 2 +- .../core/ai/AISettingModal/index.tsx | 164 +++++++++-- .../app/src/global/support/api/userRes.d.ts | 2 +- .../account/model/ModelConfigTable.tsx | 45 +++ .../app/detail/SimpleApp/EditForm.tsx | 41 ++- .../RenderInput/templates/SettingLLMModel.tsx | 9 +- projects/app/src/pages/_error.tsx | 4 +- .../app/src/pages/api/core/ai/model/test.ts | 20 +- .../src/pages/api/core/dataset/searchTest.ts | 2 + projects/app/src/pages/login/index.tsx | 14 +- projects/app/src/web/core/app/utils.ts | 46 ++- 45 files changed, 896 insertions(+), 109 deletions(-) create mode 100644 docSite/content/zh-cn/docs/development/upgrading/4821.md create mode 100644 packages/service/core/dataset/search/agent.ts diff --git a/docSite/content/zh-cn/docs/development/docker.md b/docSite/content/zh-cn/docs/development/docker.md index 7f9274151..22be77625 100644 --- a/docSite/content/zh-cn/docs/development/docker.md +++ b/docSite/content/zh-cn/docs/development/docker.md @@ -11,7 +11,7 @@ weight: 707 1. 基础的网络知识：端口，防火墙…… 2. Docker 和 Docker Compose 基础知识 -3. 大模型相关接口和参数 +3. 大模型相关接口和参数 4. RAG 相关知识：向量模型，向量数据库，向量检索 ## 部署架构图 @@ -211,6 +211,8 @@ docker restart oneapi ### 6. 配置模型 +务必先配置至少一组模型，否则系统无法正常使用。 + [点击查看模型配置教程](/docs/development/modelConfig/intro/) ## FAQ diff --git a/docSite/content/zh-cn/docs/development/faq.md b/docSite/content/zh-cn/docs/development/faq.md index 9d37d7750..ee5e9ea1c 100644 --- a/docSite/content/zh-cn/docs/development/faq.md +++ b/docSite/content/zh-cn/docs/development/faq.md @@ -130,7 +130,7 @@ OneAPI 的 API Key 配置错误，需要修改`OPENAI_API_KEY`环境变量，并 ## 四、常见模型问题 -### 如何检查模型问题 +### 如何检查模型可用性问题 1. 私有部署模型，先确认部署的模型是否正常。 2. 通过 CURL 请求，直接测试上游模型是否正常运行（云端模型或私有模型均进行测试） @@ -403,3 +403,7 @@ curl --location --request POST 'https://oneapi.xxxx/v1/chat/completions' \ "tool_choice": "auto" }' ``` + +### 向量检索得分大于 1 + +由于模型没有归一化导致的。目前仅支持归一化的模型。 \ No newline at end of file diff --git a/docSite/content/zh-cn/docs/development/sealos.md b/docSite/content/zh-cn/docs/development/sealos.md index 0a23e5ce7..4911aadc3 100644 --- a/docSite/content/zh-cn/docs/development/sealos.md +++ b/docSite/content/zh-cn/docs/development/sealos.md @@ -60,6 +60,10 @@ FastGPT 使用了 one-api 项目来管理模型池，其可以兼容 OpenAI 、A ### 3. 配置模型 +### 4. 配置模型 + +务必先配置至少一组模型，否则系统无法正常使用。 + [点击查看模型配置教程](/docs/development/modelConfig/intro/) ## 收费 diff --git a/docSite/content/zh-cn/docs/development/upgrading/4821.md b/docSite/content/zh-cn/docs/development/upgrading/4821.md new file mode 100644 index 000000000..887543aff --- /dev/null +++ b/docSite/content/zh-cn/docs/development/upgrading/4821.md @@ -0,0 +1,20 @@ +--- +title: 'V4.8.21(进行中)' +description: 'FastGPT V4.8.21 更新说明' +icon: 'upgrade' +draft: false +toc: true +weight: 804 +--- + + + +## 完整更新内容 + +1. +2. 新增 - LLM 模型支持 top_p, response_format, json_schema 参数。 +3. 新增 - Doubao1.5 模型预设。 +4. 优化 - 模型未配置时错误提示。 +5. 修复 - 简易模式，切换到其他非视觉模型时候，会强制关闭图片识别。 +6. 修复 - o1,o3 模型，在测试时候字段映射未生效导致报错。 +7. 修复 - 公众号对话空指针异常。 \ No newline at end of file diff --git a/packages/global/core/ai/model.d.ts b/packages/global/core/ai/model.d.ts index 235dd0e6f..6ebacc383 100644 --- a/packages/global/core/ai/model.d.ts +++ b/packages/global/core/ai/model.d.ts @@ -26,11 +26,16 @@ type BaseModelItemType = { export type LLMModelItemType = PriceType & BaseModelItemType & { type: ModelTypeEnum.llm; + // Model params maxContext: number; maxResponse: number; quoteMaxToken: number; maxTemperature?: number; + showTopP?: boolean; + responseFormatList?: string[]; + showStopSign?: boolean; + censor?: boolean; vision?: boolean; reasoning?: boolean; diff --git a/packages/global/core/app/type.d.ts b/packages/global/core/app/type.d.ts index 1008344d2..52b6964fb 100644 --- a/packages/global/core/app/type.d.ts +++ b/packages/global/core/app/type.d.ts @@ -74,13 +74,17 @@ export type AppDetailType = AppSchema & { export type AppSimpleEditFormType = { // templateId: string; aiSettings: { - model: string; - systemPrompt?: string | undefined; - temperature?: number; - maxToken?: number; - isResponseAnswerText: boolean; + [NodeInputKeyEnum.aiModel]: string; + [NodeInputKeyEnum.aiSystemPrompt]?: string | undefined; + [NodeInputKeyEnum.aiChatTemperature]?: number; + [NodeInputKeyEnum.aiChatMaxToken]?: number; + [NodeInputKeyEnum.aiChatIsResponseText]: boolean; maxHistories: number; - [NodeInputKeyEnum.aiChatReasoning]?: boolean; + [NodeInputKeyEnum.aiChatReasoning]?: boolean; // Is open reasoning mode + [NodeInputKeyEnum.aiChatTopP]?: number; + [NodeInputKeyEnum.aiChatStopSign]?: string; + [NodeInputKeyEnum.aiChatResponseFormat]?: string; + [NodeInputKeyEnum.aiChatJsonSchema]?: string; }; dataset: { datasets: SelectedDatasetType; @@ -119,6 +123,10 @@ export type SettingAIDataType = { maxHistories?: number; [NodeInputKeyEnum.aiChatVision]?: boolean; // Is open vision mode [NodeInputKeyEnum.aiChatReasoning]?: boolean; // Is open reasoning mode + [NodeInputKeyEnum.aiChatTopP]?: number; + [NodeInputKeyEnum.aiChatStopSign]?: string; + [NodeInputKeyEnum.aiChatResponseFormat]?: string; + [NodeInputKeyEnum.aiChatJsonSchema]?: string; }; // variable diff --git a/packages/global/core/workflow/constants.ts b/packages/global/core/workflow/constants.ts index 9aa4620bd..270bb30c2 100644 --- a/packages/global/core/workflow/constants.ts +++ b/packages/global/core/workflow/constants.ts @@ -142,6 +142,10 @@ export enum NodeInputKeyEnum { aiChatVision = 'aiChatVision', stringQuoteText = 'stringQuoteText', aiChatReasoning = 'aiChatReasoning', + aiChatTopP = 'aiChatTopP', + aiChatStopSign = 'aiChatStopSign', + aiChatResponseFormat = 'aiChatResponseFormat', + aiChatJsonSchema = 'aiChatJsonSchema', // dataset datasetSelectList = 'datasets', diff --git a/packages/global/core/workflow/runtime/type.d.ts b/packages/global/core/workflow/runtime/type.d.ts index fdf1bfeba..cc6d664c9 100644 --- a/packages/global/core/workflow/runtime/type.d.ts +++ b/packages/global/core/workflow/runtime/type.d.ts @@ -221,6 +221,10 @@ export type AIChatNodeProps = { [NodeInputKeyEnum.aiChatIsResponseText]: boolean; [NodeInputKeyEnum.aiChatVision]?: boolean; [NodeInputKeyEnum.aiChatReasoning]?: boolean; + [NodeInputKeyEnum.aiChatTopP]?: number; + [NodeInputKeyEnum.aiChatStopSign]?: string; + [NodeInputKeyEnum.aiChatResponseFormat]?: string; + [NodeInputKeyEnum.aiChatJsonSchema]?: string; [NodeInputKeyEnum.aiChatQuoteRole]?: AiChatQuoteRoleType; [NodeInputKeyEnum.aiChatQuoteTemplate]?: string; diff --git a/packages/global/core/workflow/template/system/aiChat/index.ts b/packages/global/core/workflow/template/system/aiChat/index.ts index 103d5f311..5d5cc66cc 100644 --- a/packages/global/core/workflow/template/system/aiChat/index.ts +++ b/packages/global/core/workflow/template/system/aiChat/index.ts @@ -63,14 +63,12 @@ export const AiChatModule: FlowNodeTemplateType = { key: NodeInputKeyEnum.aiChatTemperature, renderTypeList: [FlowNodeInputTypeEnum.hidden], // Set in the pop-up window label: '', - value: undefined, valueType: WorkflowIOValueTypeEnum.number }, { key: NodeInputKeyEnum.aiChatMaxToken, renderTypeList: [FlowNodeInputTypeEnum.hidden], // Set in the pop-up window label: '', - value: undefined, valueType: WorkflowIOValueTypeEnum.number }, @@ -98,6 +96,30 @@ export const AiChatModule: FlowNodeTemplateType = { valueType: WorkflowIOValueTypeEnum.boolean, value: true }, + { + key: NodeInputKeyEnum.aiChatTopP, + renderTypeList: [FlowNodeInputTypeEnum.hidden], + label: '', + valueType: WorkflowIOValueTypeEnum.number + }, + { + key: NodeInputKeyEnum.aiChatStopSign, + renderTypeList: [FlowNodeInputTypeEnum.hidden], + label: '', + valueType: WorkflowIOValueTypeEnum.string + }, + { + key: NodeInputKeyEnum.aiChatResponseFormat, + renderTypeList: [FlowNodeInputTypeEnum.hidden], + label: '', + valueType: WorkflowIOValueTypeEnum.string + }, + { + key: NodeInputKeyEnum.aiChatJsonSchema, + renderTypeList: [FlowNodeInputTypeEnum.hidden], + label: '', + valueType: WorkflowIOValueTypeEnum.string + }, // settings modal --- { ...Input_Template_System_Prompt, @@ -108,7 +130,6 @@ export const AiChatModule: FlowNodeTemplateType = { Input_Template_History, Input_Template_Dataset_Quote, Input_Template_File_Link_Prompt, - { ...Input_Template_UserChatInput, toolDescription: i18nT('workflow:user_question') } ], outputs: [ diff --git a/packages/global/core/workflow/template/system/tools.ts b/packages/global/core/workflow/template/system/tools.ts index f1472e495..8ef75d9f5 100644 --- a/packages/global/core/workflow/template/system/tools.ts +++ b/packages/global/core/workflow/template/system/tools.ts @@ -43,14 +43,12 @@ export const ToolModule: FlowNodeTemplateType = { key: NodeInputKeyEnum.aiChatTemperature, renderTypeList: [FlowNodeInputTypeEnum.hidden], // Set in the pop-up window label: '', - value: undefined, valueType: WorkflowIOValueTypeEnum.number }, { key: NodeInputKeyEnum.aiChatMaxToken, renderTypeList: [FlowNodeInputTypeEnum.hidden], // Set in the pop-up window label: '', - value: undefined, valueType: WorkflowIOValueTypeEnum.number }, { @@ -60,6 +58,30 @@ export const ToolModule: FlowNodeTemplateType = { valueType: WorkflowIOValueTypeEnum.boolean, value: true }, + { + key: NodeInputKeyEnum.aiChatTopP, + renderTypeList: [FlowNodeInputTypeEnum.hidden], + label: '', + valueType: WorkflowIOValueTypeEnum.number + }, + { + key: NodeInputKeyEnum.aiChatStopSign, + renderTypeList: [FlowNodeInputTypeEnum.hidden], + label: '', + valueType: WorkflowIOValueTypeEnum.string + }, + { + key: NodeInputKeyEnum.aiChatResponseFormat, + renderTypeList: [FlowNodeInputTypeEnum.hidden], + label: '', + valueType: WorkflowIOValueTypeEnum.string + }, + { + key: NodeInputKeyEnum.aiChatJsonSchema, + renderTypeList: [FlowNodeInputTypeEnum.hidden], + label: '', + valueType: WorkflowIOValueTypeEnum.string + }, { ...Input_Template_System_Prompt, diff --git a/packages/service/core/ai/config/provider/DeepSeek.json b/packages/service/core/ai/config/provider/DeepSeek.json index 1aee2c1ab..5838de644 100644 --- a/packages/service/core/ai/config/provider/DeepSeek.json +++ b/packages/service/core/ai/config/provider/DeepSeek.json @@ -5,7 +5,7 @@ "model": "deepseek-chat", "name": "Deepseek-chat", "maxContext": 64000, - "maxResponse": 4096, + "maxResponse": 8000, "quoteMaxToken": 60000, "maxTemperature": 1.5, "vision": false, @@ -25,7 +25,7 @@ "model": "deepseek-reasoner", "name": "Deepseek-reasoner", "maxContext": 64000, - "maxResponse": 4096, + "maxResponse": 8000, "quoteMaxToken": 60000, "maxTemperature": null, "vision": false, diff --git a/packages/service/core/ai/config/provider/Doubao.json b/packages/service/core/ai/config/provider/Doubao.json index 64924b0e4..f8f124083 100644 --- a/packages/service/core/ai/config/provider/Doubao.json +++ b/packages/service/core/ai/config/provider/Doubao.json @@ -1,6 +1,94 @@ { "provider": "Doubao", "list": [ + { + "model": "Doubao-1.5-lite-32k", + "name": "Doubao-1.5-lite-32k", + "maxContext": 32000, + "maxResponse": 4000, + "quoteMaxToken": 32000, + "maxTemperature": 1, + "vision": false, + "toolChoice": true, + "functionCall": false, + "defaultSystemChatPrompt": "", + "datasetProcess": true, + "usedInClassify": true, + "customCQPrompt": "", + "usedInExtractFields": true, + "usedInQueryExtension": true, + "customExtractPrompt": "", + "usedInToolCall": true, + "defaultConfig": {}, + "fieldMap": {}, + "type": "llm" + }, + { + "model": "Doubao-1.5-pro-32k", + "name": "Doubao-1.5-pro-32k", + "maxContext": 32000, + "maxResponse": 4000, + "quoteMaxToken": 32000, + "maxTemperature": 1, + "vision": false, + "toolChoice": true, + "functionCall": false, + "defaultSystemChatPrompt": "", + "datasetProcess": true, + "usedInClassify": true, + "customCQPrompt": "", + "usedInExtractFields": true, + "usedInQueryExtension": true, + "customExtractPrompt": "", + "usedInToolCall": true, + "defaultConfig": {}, + "fieldMap": {}, + "type": "llm" + }, + { + "model": "Doubao-1.5-pro-256k", + "name": "Doubao-1.5-pro-256k", + "maxContext": 256000, + "maxResponse": 12000, + "quoteMaxToken": 256000, + "maxTemperature": 1, + "vision": false, + "toolChoice": true, + "functionCall": false, + "defaultSystemChatPrompt": "", + "datasetProcess": true, + "usedInClassify": true, + "customCQPrompt": "", + "usedInExtractFields": true, + "usedInQueryExtension": true, + "customExtractPrompt": "", + "usedInToolCall": true, + "defaultConfig": {}, + "fieldMap": {}, + "type": "llm" + }, + { + "model": "Doubao-1.5-vision-pro-32k", + "name": "Doubao-1.5-vision-pro-32k", + "maxContext": 32000, + "maxResponse": 4000, + "quoteMaxToken": 32000, + "maxTemperature": 1, + "vision": true, + "toolChoice": true, + "functionCall": false, + "defaultSystemChatPrompt": "", + "datasetProcess": true, + "usedInClassify": true, + "customCQPrompt": "", + "usedInExtractFields": true, + "usedInQueryExtension": true, + "customExtractPrompt": "", + "usedInToolCall": true, + "defaultConfig": {}, + "fieldMap": {}, + "type": "llm" + }, { "model": "Doubao-lite-4k", "name": "Doubao-lite-4k", diff --git a/packages/service/core/ai/config/provider/OpenAI.json b/packages/service/core/ai/config/provider/OpenAI.json index a852cf22d..e1c137390 100644 --- a/packages/service/core/ai/config/provider/OpenAI.json +++ b/packages/service/core/ai/config/provider/OpenAI.json @@ -8,6 +8,9 @@ "maxResponse": 16000, "quoteMaxToken": 60000, "maxTemperature": 1.2, + "showTopP": true, + "responseFormatList": ["text", "json_object", "json_schema"], + "showStopSign": true, "vision": true, "toolChoice": true, "functionCall": true, diff --git a/packages/service/core/ai/config/utils.ts b/packages/service/core/ai/config/utils.ts index d303bc5b4..bd03c317e 100644 --- a/packages/service/core/ai/config/utils.ts +++ b/packages/service/core/ai/config/utils.ts @@ -31,10 +31,12 @@ import { delay } from '@fastgpt/global/common/system/utils'; export const loadSystemModels = async (init = false) => { const getProviderList = () => { const currentFileUrl = new URL(import.meta.url); - const modelsPath = path.join( - path.dirname(currentFileUrl.pathname.replace(/^\/+/, '')), - 'provider' + const filePath = decodeURIComponent( + process.platform === 'win32' + ? currentFileUrl.pathname.substring(1) // Remove leading slash on Windows + : currentFileUrl.pathname ); + const modelsPath = path.join(path.dirname(filePath), 'provider'); return fs.readdirSync(modelsPath) as string[]; }; @@ -150,6 +152,7 @@ export const loadSystemModels = async (init = false) => { console.error('Load models error', error); // @ts-ignore global.systemModelList = undefined; + return Promise.reject(error); } }; diff --git a/packages/service/core/ai/utils.ts b/packages/service/core/ai/utils.ts index 40951a372..6ff350b8b 100644 --- a/packages/service/core/ai/utils.ts +++ b/packages/service/core/ai/utils.ts @@ -37,9 +37,14 @@ export const computedTemperature = ({ return temperature; }; -type CompletionsBodyType = +type CompletionsBodyType = ( | ChatCompletionCreateParamsNonStreaming - | ChatCompletionCreateParamsStreaming; + | ChatCompletionCreateParamsStreaming +) & { + response_format?: any; + json_schema?: string; + stop?: string; +}; type InferCompletionsBody = T extends { stream: true } ? ChatCompletionCreateParamsStreaming : ChatCompletionCreateParamsNonStreaming; @@ -53,6 +58,10 @@ export const llmCompletionsBodyFormat = ( return body as InferCompletionsBody; } + const response_format = body.response_format; + const json_schema = body.json_schema ?? undefined; + const stop = body.stop ?? undefined; + const requestBody: T = { ...body, temperature: @@ -62,7 +71,14 @@ export const llmCompletionsBodyFormat = ( temperature: body.temperature }) : undefined, - ...modelData?.defaultConfig + ...modelData?.defaultConfig, + response_format: response_format + ? { + type: response_format, + json_schema + } + : undefined, + stop: stop?.split('|') }; // field map diff --git a/packages/service/core/dataset/search/agent.ts b/packages/service/core/dataset/search/agent.ts new file mode 100644 index 000000000..c5f9cea6d --- /dev/null +++ b/packages/service/core/dataset/search/agent.ts @@ -0,0 +1,277 @@ +import { chats2GPTMessages } from '@fastgpt/global/core/chat/adapt'; +import { ChatItemType } from '@fastgpt/global/core/chat/type'; +import { DatasetSearchModeEnum } from '@fastgpt/global/core/dataset/constants'; +import { getLLMModel } from '../../ai/model'; +import { filterGPTMessageByMaxContext } from '../../chat/utils'; +import { replaceVariable } from '@fastgpt/global/common/string/tools'; +import { createChatCompletion } from '../../ai/config'; +import { llmCompletionsBodyFormat } from '../../ai/utils'; +import { ChatCompletionMessageParam } from '@fastgpt/global/core/ai/type'; +import { SearchDataResponseItemType } from '@fastgpt/global/core/dataset/type'; +import { searchDatasetData } from './controller'; + +type SearchDatasetDataProps = { + queries: string[]; + histories: ChatItemType[]; + teamId: string; + model: string; + similarity?: number; // min distance + limit: number; // max Token limit + datasetIds: string[]; + searchMode?: `${DatasetSearchModeEnum}`; + usingReRank?: boolean; + reRankQuery: string; + + /* + { + tags: { + $and: ["str1","str2"], + $or: ["str1","str2",null] null means no tags + }, + createTime: { + $gte: 'xx', + $lte: 'xxx' + } + } + */ + collectionFilterMatch?: string; +}; + +const analyzeQuery = async ({ query, histories }: { query: string; histories: ChatItemType[] }) => { + const modelData = getLLMModel('gpt-4o-mini'); + + const systemFewShot = ` +## 知识背景 +FastGPT 是低代码AI应用构建平台，支持通过语义相似度实现精准数据检索。用户正在利用该功能开发数据检索应用。 + +## 任务目标 +基于用户历史对话和知识背景，生成多维度检索方案，确保覆盖核心语义及潜在关联维度。 + +## 工作流程 +1. 问题解构阶段 + [意图识别] 提取用户问题的核心实体和关系： + - 显性需求：直接提及的关键词 + - 隐性需求：可能涉及的关联概念 + [示例] 若问题为"推荐手机"，需考虑价格、品牌、使用场景等维度 + +2. 完整性校验阶段 + [完整性评估] 检查是否缺失核心实体和关系： + - 主语完整 + - 多实体关系准确 + [维度扩展] 检查是否需要补充： + □ 时间范围 □ 地理限定 □ 比较维度 + □ 专业术语 □ 同义词替换 □ 场景参数 + +3. 检索生成阶段 + [组合策略] 生成包含以下要素的查询序列： + ① 基础查询（核心关键词） + ② 扩展查询（核心+同义词） + ③ 场景查询（核心+场景限定词） + ④ 逆向查询（相关技术/对比对象） + +## 输出规范 +格式要求： +1. 每个查询为完整陈述句 +2. 包含至少1个核心词+1个扩展维度 +3. 按查询范围从宽到窄排序 + +禁止项： +- 使用问句形式 +- 包含解决方案描述 +- 超出话题范围的假设 + +## 执行示例 +用户问题："如何优化数据检索速度" + +查询内容： +1. FastGPT 数据检索速度优化的常用方法 +2. FastGPT 大数据量下的语义检索性能提升方案 +3. FastGPT API 响应时间的优化指标 + +## 任务开始 +`.trim(); + const filterHistories = await filterGPTMessageByMaxContext({ + messages: chats2GPTMessages({ messages: histories, reserveId: false }), + maxContext: modelData.maxContext - 1000 + }); + + const messages = [ + { + role: 'system', + content: systemFewShot + }, + ...filterHistories, + { + role: 'user', + content: query + } + ] as any; + + const { response: result } = await createChatCompletion({ + body: llmCompletionsBodyFormat( + { + stream: false, + model: modelData.model, + temperature: 0.1, + messages + }, + modelData + ) + }); + let answer = result.choices?.[0]?.message?.content || ''; + + // Extract queries from the answer by line number + const queries = answer + .split('\n') + .map((line) => { + const match = line.match(/^\d+\.\s*(.+)$/); + return match ? match[1].trim() : null; + }) + .filter(Boolean) as string[]; + + if (queries.length === 0) { + return [answer]; + } + + return queries; +}; +const checkQuery = async ({ + queries, + histories, + searchResult +}: { + queries: string[]; + histories: ChatItemType[]; + searchResult: SearchDataResponseItemType[]; +}) => { + const modelData = getLLMModel('gpt-4o-mini'); + + const systemFewShot = ` +## 知识背景 +FastGPT 是低代码AI应用构建平台，支持通过语义相似度实现精准数据检索。用户正在利用该功能开发数据检索应用。 + +## 查询结果 +${searchResult.map((item) => item.q + item.a).join('---\n---')} + +## 任务目标 +检查"检索结果"是否覆盖用户的问题，如果无法覆盖用户问题，则再次生成检索方案。 + +## 工作流程 +1. 检查检索结果是否覆盖用户的问题 +2. 如果检索结果覆盖用户问题，则直接输出："Done" +3. 如果无法覆盖用户问题，则结合用户问题和检索结果，生成进一步的检索方案，进行深度检索 + +## 输出规范 + +1. 每个查询均为完整的查询语句 +2. 通过序号来表示多个检索内容 + +## 输出示例1 +Done + +## 输出示例2 +1. 环界云计算的办公地址 +2. 环界云计算的注册地址在哪里 + +## 任务开始 +`.trim(); + const filterHistories = await filterGPTMessageByMaxContext({ + messages: chats2GPTMessages({ messages: histories, reserveId: false }), + maxContext: modelData.maxContext - 1000 + }); + + const messages = [ + { + role: 'system', + content: systemFewShot + }, + ...filterHistories, + { + role: 'user', + content: queries.join('\n') + } + ] as any; + console.log(messages); + const { response: result } = await createChatCompletion({ + body: llmCompletionsBodyFormat( + { + stream: false, + model: modelData.model, + temperature: 0.1, + messages + }, + modelData + ) + }); + let answer = result.choices?.[0]?.message?.content || ''; + console.log(answer); + if (answer.includes('Done')) { + return []; + } + + const nextQueries = answer + .split('\n') + .map((line) => { + const match = line.match(/^\d+\.\s*(.+)$/); + return match ? match[1].trim() : null; + }) + .filter(Boolean) as string[]; + + return nextQueries; +}; +export const agentSearchDatasetData = async ({ + searchRes = [], + tokens = 0, + ...props +}: SearchDatasetDataProps & { + searchRes?: SearchDataResponseItemType[]; + tokens?: number; +}) => { + const query = props.queries[0]; + + const searchResultList: SearchDataResponseItemType[] = []; + let searchQueries: string[] = []; + + // 1. agent 分析问题 + searchQueries = await analyzeQuery({ query, histories: props.histories }); + + // 2. 检索内容 + 检查 + let retryTimes = 3; + while (true) { + retryTimes--; + if (retryTimes < 0) break; + + console.log(searchQueries, '--'); + const { searchRes: searchRes2, tokens: tokens2 } = await searchDatasetData({ + ...props, + queries: searchQueries + }); + // console.log(searchRes2.map((item) => item.q)); + // deduplicate and merge search results + const uniqueResults = searchRes2.filter((item) => { + return !searchResultList.some((existingItem) => existingItem.id === item.id); + }); + searchResultList.push(...uniqueResults); + if (uniqueResults.length === 0) break; + + const checkResult = await checkQuery({ + queries: searchQueries, + histories: props.histories, + searchResult: searchRes2 + }); + + if (checkResult.length > 0) { + searchQueries = checkResult; + } else { + break; + } + } + + console.log(searchResultList.length); + return { + searchRes: searchResultList, + tokens: 0, + usingSimilarityFilter: false, + usingReRank: false + }; +}; diff --git a/packages/service/core/dataset/search/controller.ts b/packages/service/core/dataset/search/controller.ts index df458f632..def8613d3 100644 --- a/packages/service/core/dataset/search/controller.ts +++ b/packages/service/core/dataset/search/controller.ts @@ -23,8 +23,10 @@ import json5 from 'json5'; import { MongoDatasetCollectionTags } from '../tag/schema'; import { readFromSecondary } from '../../../common/mongo/utils'; import { MongoDatasetDataText } from '../data/dataTextSchema'; +import { ChatItemType } from '@fastgpt/global/core/chat/type'; type SearchDatasetDataProps = { + histories?: ChatItemType[]; teamId: string; model: string; similarity?: number; // min distance diff --git a/packages/service/core/workflow/dispatch/agent/runTool/functionCall.ts b/packages/service/core/workflow/dispatch/agent/runTool/functionCall.ts index fa7e27187..1ae995a13 100644 --- a/packages/service/core/workflow/dispatch/agent/runTool/functionCall.ts +++ b/packages/service/core/workflow/dispatch/agent/runTool/functionCall.ts @@ -46,7 +46,15 @@ export const runToolWithFunctionCall = async ( externalProvider, stream, workflowStreamResponse, - params: { temperature, maxToken, aiChatVision } + params: { + temperature, + maxToken, + aiChatVision, + aiChatTopP, + aiChatStopSign, + aiChatResponseFormat, + aiChatJsonSchema + } } = workflowProps; // Interactive @@ -204,12 +212,18 @@ export const runToolWithFunctionCall = async ( const requestBody = llmCompletionsBodyFormat( { model: toolModel.model, - temperature, - max_tokens, + stream, messages: requestMessages, functions, - function_call: 'auto' + function_call: 'auto', + + temperature, + max_tokens, + top_p: aiChatTopP, + stop: aiChatStopSign, + response_format: aiChatResponseFormat, + json_schema: aiChatJsonSchema }, toolModel ); diff --git a/packages/service/core/workflow/dispatch/agent/runTool/promptCall.ts b/packages/service/core/workflow/dispatch/agent/runTool/promptCall.ts index 331496852..58e95a059 100644 --- a/packages/service/core/workflow/dispatch/agent/runTool/promptCall.ts +++ b/packages/service/core/workflow/dispatch/agent/runTool/promptCall.ts @@ -54,7 +54,15 @@ export const runToolWithPromptCall = async ( externalProvider, stream, workflowStreamResponse, - params: { temperature, maxToken, aiChatVision } + params: { + temperature, + maxToken, + aiChatVision, + aiChatTopP, + aiChatStopSign, + aiChatResponseFormat, + aiChatJsonSchema + } } = workflowProps; if (interactiveEntryToolParams) { @@ -215,10 +223,14 @@ export const runToolWithPromptCall = async ( const requestBody = llmCompletionsBodyFormat( { model: toolModel.model, + stream, + messages: requestMessages, temperature, max_tokens, - stream, - messages: requestMessages + top_p: aiChatTopP, + stop: aiChatStopSign, + response_format: aiChatResponseFormat, + json_schema: aiChatJsonSchema }, toolModel ); diff --git a/packages/service/core/workflow/dispatch/agent/runTool/toolChoice.ts b/packages/service/core/workflow/dispatch/agent/runTool/toolChoice.ts index b2b845f75..5bbfc838f 100644 --- a/packages/service/core/workflow/dispatch/agent/runTool/toolChoice.ts +++ b/packages/service/core/workflow/dispatch/agent/runTool/toolChoice.ts @@ -93,7 +93,15 @@ export const runToolWithToolChoice = async ( stream, externalProvider, workflowStreamResponse, - params: { temperature, maxToken, aiChatVision } + params: { + temperature, + maxToken, + aiChatVision, + aiChatTopP, + aiChatStopSign, + aiChatResponseFormat, + aiChatJsonSchema + } } = workflowProps; if (maxRunToolTimes <= 0 && response) { @@ -263,12 +271,16 @@ export const runToolWithToolChoice = async ( const requestBody = llmCompletionsBodyFormat( { model: toolModel.model, - temperature, - max_tokens, stream, messages: requestMessages, tools, - tool_choice: 'auto' + tool_choice: 'auto', + temperature, + max_tokens, + top_p: aiChatTopP, + stop: aiChatStopSign, + response_format: aiChatResponseFormat, + json_schema: aiChatJsonSchema }, toolModel ); diff --git a/packages/service/core/workflow/dispatch/agent/runTool/type.d.ts b/packages/service/core/workflow/dispatch/agent/runTool/type.d.ts index 8d70bc227..64ecd79fc 100644 --- a/packages/service/core/workflow/dispatch/agent/runTool/type.d.ts +++ b/packages/service/core/workflow/dispatch/agent/runTool/type.d.ts @@ -16,12 +16,16 @@ export type DispatchToolModuleProps = ModuleDispatchProps<{ [NodeInputKeyEnum.history]?: ChatItemType[]; [NodeInputKeyEnum.userChatInput]: string; + [NodeInputKeyEnum.fileUrlList]?: string[]; [NodeInputKeyEnum.aiModel]: string; [NodeInputKeyEnum.aiSystemPrompt]: string; [NodeInputKeyEnum.aiChatTemperature]: number; [NodeInputKeyEnum.aiChatMaxToken]: number; [NodeInputKeyEnum.aiChatVision]?: boolean; - [NodeInputKeyEnum.fileUrlList]?: string[]; + [NodeInputKeyEnum.aiChatTopP]?: number; + [NodeInputKeyEnum.aiChatStopSign]?: string; + [NodeInputKeyEnum.aiChatResponseFormat]?: string; + [NodeInputKeyEnum.aiChatJsonSchema]?: string; }> & { messages: ChatCompletionMessageParam[]; toolNodes: ToolNodeItemType[]; diff --git a/packages/service/core/workflow/dispatch/chat/oneapi.ts b/packages/service/core/workflow/dispatch/chat/oneapi.ts index 4b8cc865c..5f6f5ab56 100644 --- a/packages/service/core/workflow/dispatch/chat/oneapi.ts +++ b/packages/service/core/workflow/dispatch/chat/oneapi.ts @@ -89,6 +89,11 @@ export const dispatchChatCompletion = async (props: ChatProps): Promise, Response = any> { run(data: Props) { // watch memory - addLog.debug(`${this.name} worker queueLength: ${this.workerQueue.length}`); + // addLog.debug(`${this.name} worker queueLength: ${this.workerQueue.length}`); return new Promise((resolve, reject) => { /* diff --git a/packages/templates/register.ts b/packages/templates/register.ts index fae3334ea..5b6d33423 100644 --- a/packages/templates/register.ts +++ b/packages/templates/register.ts @@ -7,7 +7,12 @@ import { AppTemplateSchemaType } from '@fastgpt/global/core/app/type'; const getTemplateNameList = () => { const currentFileUrl = new URL(import.meta.url); - const templatesPath = path.join(path.dirname(currentFileUrl.pathname), 'src'); + const filePath = decodeURIComponent( + process.platform === 'win32' + ? currentFileUrl.pathname.substring(1) // Remove leading slash on Windows + : currentFileUrl.pathname + ); + const templatesPath = path.join(path.dirname(filePath), 'src'); return fs.readdirSync(templatesPath) as string[]; }; diff --git a/packages/web/components/common/MySlider/InputSlider.tsx b/packages/web/components/common/MySlider/InputSlider.tsx index a8726e433..a23cb5383 100644 --- a/packages/web/components/common/MySlider/InputSlider.tsx +++ b/packages/web/components/common/MySlider/InputSlider.tsx @@ -27,7 +27,7 @@ const InputSlider = ({ valLen * 0.8 + min, valLen * 0.985 + min ]; - }, []); + }, [max, min]); return ( diff --git a/packages/web/i18n/en/account.json b/packages/web/i18n/en/account.json index 1bef764cd..6290f70d1 100644 --- a/packages/web/i18n/en/account.json +++ b/packages/web/i18n/en/account.json @@ -55,6 +55,9 @@ "model.request_auth_tip": "When making a request to a custom request address, carry the request header: Authorization: Bearer xxx to make the request.", "model.request_url": "Custom url", "model.request_url_tip": "If you fill in this value, you will initiate a request directly without passing. \nYou need to follow the API format of Openai and fill in the full request address, such as\n\nLLM: {Host}}/v1/Chat/Completions\n\nEmbedding: {host}}/v1/embeddings\n\nSTT: {Host}/v1/Audio/Transcriptions\n\nTTS: {Host}}/v1/Audio/Speech\n\nRERARARARARARARANK: {Host}}/v1/RERARARARARARARARARARANK", + "model.response_format": "Response format", + "model.show_stop_sign": "Display stop sequence parameters", + "model.show_top_p": "Show Top-p parameters", "model.test_model": "Model testing", "model.tool_choice": "Tool choice", "model.tool_choice_tag": "ToolCall", diff --git a/packages/web/i18n/en/app.json b/packages/web/i18n/en/app.json index c3ca4a379..e70ae1170 100644 --- a/packages/web/i18n/en/app.json +++ b/packages/web/i18n/en/app.json @@ -110,12 +110,16 @@ "publish_success": "Publish Successful", "question_guide_tip": "After the conversation, 3 guiding questions will be generated for you.", "reasoning_response": "Output thinking", + "response_format": "Response format", "saved_success": "Saved successfully! \nTo use this version externally, click Save and Publish", "search_app": "Search apps", "setting_app": "Workflow", "setting_plugin": "Workflow", + "show_top_p_tip": "An alternative method of temperature sampling, called Nucleus sampling, the model considers the results of tokens with TOP_P probability mass quality. \nTherefore, 0.1 means that only tokens containing the highest probability quality are considered. \nThe default is 1.", "simple_tool_tips": "This plugin contains special inputs and is not currently supported for invocation by simple applications.", "source_updateTime": "Update time", + "stop_sign": "Stop", + "stop_sign_placeholder": "Multiple serial numbers are separated by |, for example: aaa|stop", "stream_response": "Stream", "stream_response_tip": "Turning this switch off forces the model to use non-streaming mode and will not output content directly. \nIn the output of the AI ​​reply, the content output by this model can be obtained for secondary processing.", "temperature": "Temperature", diff --git a/packages/web/i18n/en/login.json b/packages/web/i18n/en/login.json index 8dfe1b826..39cbb57be 100644 --- a/packages/web/i18n/en/login.json +++ b/packages/web/i18n/en/login.json @@ -1,20 +1,21 @@ { "Chinese_ip_tip": "It is detected that you are a mainland Chinese IP, click to jump to visit the mainland China version.", "Login": "Login", + "agree": "agree", + "cookies_tip": " This website uses cookies to provide a better service experience. By continuing to use the site, you agree to our Cookie Policy.", "forget_password": "Find Password", "login_failed": "Login failed", "login_success": "Login successful", + "model_not_config": "It is detected that the system has not configured the model, please configure the model before using it", "no_remind": "Don't remind again", "password_condition": "Password maximum 60 characters", "password_tip": "Password must be at least 6 characters long and contain at least two combinations: numbers, letters, or special characters", "policy_tip": "By using this service, you agree to our", "privacy": "Privacy Policy", + "privacy_policy": "Privacy Policy", "redirect": "Jump", "register": "Register", "root_password_placeholder": "The root user password is the value of the environment variable DEFAULT_ROOT_PSW", "terms": "Terms", - "use_root_login": "Log in as root user", - "agree": "agree", - "cookies_tip": " This website uses cookies to provide a better service experience. By continuing to use the site, you agree to our Cookie Policy.", - "privacy_policy": "Privacy Policy" + "use_root_login": "Log in as root user" } diff --git a/packages/web/i18n/zh-CN/account.json b/packages/web/i18n/zh-CN/account.json index d9043f010..8604427c3 100644 --- a/packages/web/i18n/zh-CN/account.json +++ b/packages/web/i18n/zh-CN/account.json @@ -55,6 +55,9 @@ "model.request_auth_tip": "向自定义请求地址发起请求时候，携带请求头：Authorization: Bearer xxx 进行请求", "model.request_url": "自定义请求地址", "model.request_url_tip": "如果填写该值，则会直接向该地址发起请求，不经过 OneAPI。需要遵循 OpenAI 的 API格式，并填写完整请求地址，例如：\nLLM: {{host}}/v1/chat/completions\nEmbedding: {{host}}/v1/embeddings\nSTT: {{host}}/v1/audio/transcriptions\nTTS: {{host}}/v1/audio/speech\nRerank: {{host}}/v1/rerank", + "model.response_format": "响应格式", + "model.show_stop_sign": "展示停止序列参数", + "model.show_top_p": "展示 Top-p 参数", "model.test_model": "模型测试", "model.tool_choice": "支持工具调用", "model.tool_choice_tag": "工具调用", diff --git a/packages/web/i18n/zh-CN/app.json b/packages/web/i18n/zh-CN/app.json index d75cd9ea5..b1f41b3aa 100644 --- a/packages/web/i18n/zh-CN/app.json +++ b/packages/web/i18n/zh-CN/app.json @@ -110,12 +110,16 @@ "publish_success": "发布成功", "question_guide_tip": "对话结束后，会为你生成 3 个引导性问题。", "reasoning_response": "输出思考", + "response_format": "回复格式", "saved_success": "保存成功！如需在外部使用该版本，请点击“保存并发布”", "search_app": "搜索应用", "setting_app": "应用配置", "setting_plugin": "插件配置", + "show_top_p_tip": "用温度采样的替代方法，称为Nucleus采样，该模型考虑了具有TOP_P概率质量质量的令牌的结果。因此，0.1表示仅考虑包含最高概率质量的令牌。默认为 1。", "simple_tool_tips": "该插件含有特殊输入，暂不支持被简易应用调用", "source_updateTime": "更新时间", + "stop_sign": "停止序列", + "stop_sign_placeholder": "多个序列号通过 | 隔开，例如：aaa|stop", "stream_response": "流输出", "stream_response_tip": "关闭该开关，可以强制模型使用非流模式，并且不会直接进行内容输出。可以在 AI 回复的输出中，获取本次模型输出的内容进行二次处理。", "temperature": "温度", diff --git a/packages/web/i18n/zh-CN/login.json b/packages/web/i18n/zh-CN/login.json index f3cc74a89..30d157e06 100644 --- a/packages/web/i18n/zh-CN/login.json +++ b/packages/web/i18n/zh-CN/login.json @@ -6,6 +6,7 @@ "forget_password": "忘记密码?", "login_failed": "登录异常", "login_success": "登录成功", + "model_not_config": "检测到系统未配置模型，请先配置模型后再使用", "no_remind": "不再提醒", "password_condition": "密码最多 60 位", "password_tip": "密码至少 6 位，且至少包含两种组合：数字、字母或特殊字符", diff --git a/packages/web/i18n/zh-Hant/account.json b/packages/web/i18n/zh-Hant/account.json index 66ce5b25c..b619ba6f1 100644 --- a/packages/web/i18n/zh-Hant/account.json +++ b/packages/web/i18n/zh-Hant/account.json @@ -54,6 +54,9 @@ "model.request_auth_tip": "向自訂請求地址發起請求時候，攜帶請求頭：Authorization: Bearer xxx 進行請求", "model.request_url": "自訂請求地址", "model.request_url_tip": "如果填寫該值，則會直接向該地址發起請求，不經過 OneAPI。\n需要遵循 OpenAI 的 API格式，並填寫完整請求地址，例如：\n\nLLM: {{host}}/v1/chat/completions\n\nEmbedding: {{host}}/v1/embeddings\n\nSTT: {{host}}/v1/audio/transcriptions\n\nTTS: {{host}}/v1/audio/speech\n\nRerank: {{host}}/v1/rerank", + "model.response_format": "響應格式", + "model.show_stop_sign": "展示停止序列參數", + "model.show_top_p": "展示 Top-p 參數", "model.test_model": "模型測試", "model.tool_choice": "支援工具調用", "model.tool_choice_tag": "工具調用", diff --git a/packages/web/i18n/zh-Hant/app.json b/packages/web/i18n/zh-Hant/app.json index c2b7651f3..e5221b588 100644 --- a/packages/web/i18n/zh-Hant/app.json +++ b/packages/web/i18n/zh-Hant/app.json @@ -110,12 +110,16 @@ "publish_success": "發布成功", "question_guide_tip": "對話結束後，會為你產生 3 個引導性問題。", "reasoning_response": "輸出思考", + "response_format": "回复格式", "saved_success": "保存成功！\n如需在外部使用該版本，請點擊“儲存並發布”", "search_app": "搜尋應用程式", "setting_app": "應用程式設定", "setting_plugin": "外掛設定", + "show_top_p_tip": "用溫度採樣的替代方法，稱為Nucleus採樣，該模型考慮了具有TOP_P概率質量質量的令牌的結果。\n因此，0.1表示僅考慮包含最高概率質量的令牌。\n默認為 1。", "simple_tool_tips": "該插件含有特殊輸入，暫不支持被簡易應用調用", "source_updateTime": "更新時間", + "stop_sign": "停止序列", + "stop_sign_placeholder": "多個序列號通過 | 隔開，例如：aaa|stop", "stream_response": "流輸出", "stream_response_tip": "關閉該開關​​，可以強制模型使用非流模式，並且不會直接進行內容輸出。\n可在 AI 回覆的輸出中，取得本次模型輸出的內容進行二次處理。", "temperature": "溫度", diff --git a/packages/web/i18n/zh-Hant/login.json b/packages/web/i18n/zh-Hant/login.json index 51bafc851..31b11e8e8 100644 --- a/packages/web/i18n/zh-Hant/login.json +++ b/packages/web/i18n/zh-Hant/login.json @@ -6,6 +6,7 @@ "forget_password": "忘記密碼？", "login_failed": "登入失敗", "login_success": "登入成功", + "model_not_config": "檢測到系統未配置模型，請先配置模型後再使用", "no_remind": "不再提醒", "password_condition": "密碼最多 60 個字元", "password_tip": "密碼至少 6 位，且至少包含兩種組合：數字、字母或特殊字符", diff --git a/projects/app/package.json b/projects/app/package.json index 0288929ab..a04eb75a0 100644 --- a/projects/app/package.json +++ b/projects/app/package.json @@ -1,6 +1,6 @@ { "name": "app", - "version": "4.8.20", + "version": "4.8.21", "private": false, "scripts": { "dev": "next dev", diff --git a/projects/app/src/components/core/ai/AISettingModal/index.tsx b/projects/app/src/components/core/ai/AISettingModal/index.tsx index 5059119c7..998bf35fe 100644 --- a/projects/app/src/components/core/ai/AISettingModal/index.tsx +++ b/projects/app/src/components/core/ai/AISettingModal/index.tsx @@ -18,7 +18,8 @@ import { Thead, Tr, Table, - FlexProps + FlexProps, + Input } from '@chakra-ui/react'; import { useSystemStore } from '@/web/common/system/useSystemStore'; import { NodeInputKeyEnum } from '@fastgpt/global/core/workflow/constants'; @@ -31,13 +32,15 @@ import { getWebLLMModel } from '@/web/common/system/utils'; import MyIcon from '@fastgpt/web/components/common/Icon'; import dynamic from 'next/dynamic'; import InputSlider from '@fastgpt/web/components/common/MySlider/InputSlider'; +import MySelect from '@fastgpt/web/components/common/MySelect'; +import JsonEditor from '@fastgpt/web/components/common/Textarea/JsonEditor'; const ModelPriceModal = dynamic(() => import('@/components/core/ai/ModelTable').then((mod) => mod.ModelPriceModal) ); const FlexItemStyles: FlexProps = { - mt: 5, + mt: 4, alignItems: 'center', h: '35px' }; @@ -68,7 +71,7 @@ const AIChatSettingsModal = ({ const [refresh, setRefresh] = useState(false); const { feConfigs } = useSystemStore(); - const { handleSubmit, getValues, setValue, watch } = useForm({ + const { handleSubmit, getValues, setValue, watch, register } = useForm({ defaultValues: defaultData }); const model = watch('model'); @@ -88,6 +91,17 @@ const AIChatSettingsModal = ({ const llmSupportTemperature = typeof selectedModel?.maxTemperature === 'number'; const llmSupportReasoning = !!selectedModel?.reasoning; + const topP = watch(NodeInputKeyEnum.aiChatTopP); + const llmSupportTopP = !!selectedModel?.showTopP; + + const stopSign = watch(NodeInputKeyEnum.aiChatStopSign); + const llmSupportStopSign = !!selectedModel?.showStopSign; + + const responseFormat = watch(NodeInputKeyEnum.aiChatResponseFormat); + const jsonSchema = watch(NodeInputKeyEnum.aiChatJsonSchema); + const llmSupportResponseFormat = + !!selectedModel?.responseFormatList && selectedModel?.responseFormatList.length > 0; + const tokenLimit = useMemo(() => { return selectedModel?.maxResponse || 4096; }, [selectedModel?.maxResponse]); @@ -146,7 +160,7 @@ const AIChatSettingsModal = ({ )} + {llmSupportTopP && ( + + + + Top_p + + + { + setValue(NodeInputKeyEnum.aiChatTopP, e.target.checked ? 1 : undefined); + }} + /> + + + { + setValue(NodeInputKeyEnum.aiChatTopP, e); + setRefresh(!refresh); + }} + /> + + + )} + {llmSupportStopSign && ( + + + + {t('app:stop_sign')} + + { + setValue(NodeInputKeyEnum.aiChatStopSign, e.target.checked ? '' : undefined); + }} + /> + + + + + + )} + {llmSupportResponseFormat && selectedModel?.responseFormatList && ( + + + {t('app:response_format')} + { + setValue( + NodeInputKeyEnum.aiChatResponseFormat, + e.target.checked ? selectedModel?.responseFormatList?.[0] : undefined + ); + }} + /> + + + + isDisabled={responseFormat === undefined} + size={'sm'} + bg={'myGray.25'} + list={selectedModel.responseFormatList.map((item) => ({ + value: item, + label: item + }))} + value={responseFormat} + onchange={(e) => { + setValue(NodeInputKeyEnum.aiChatResponseFormat, e); + }} + /> + + + )} + {/* Json schema */} + {responseFormat === 'json_schema' && ( + + + JSON Schema + + + { + setValue(NodeInputKeyEnum.aiChatJsonSchema, e); + }} + bg={'myGray.25'} + /> + + + )} {llmSupportReasoning && ( @@ -306,25 +424,6 @@ const AIChatSettingsModal = ({ )} - {showResponseAnswerText && ( - - - - {t('app:stream_response')} - - - { - const value = e.target.checked; - setValue(NodeInputKeyEnum.aiChatIsResponseText, value); - setRefresh((state) => !state); - }} - /> - - - )} {showVisionSwitch && ( @@ -349,6 +448,25 @@ const AIChatSettingsModal = ({ )} + {showResponseAnswerText && ( + + + + {t('app:stream_response')} + + + { + const value = e.target.checked; + setValue(NodeInputKeyEnum.aiChatIsResponseText, value); + setRefresh((state) => !state); + }} + /> + + + )}