🔍 搜尋結果:user'

🔍 搜尋結果:user'

使用 AI-copilot(Next.js、gpt4、LangChain 和 CopilotKit)建立電子表格應用程式

**長話短說** -------- 在本文中,您將學習如何建立人工智慧驅動的電子表格應用程式,該應用程式允許您使用簡單的英語命令執行各種會計功能並輕鬆與資料互動。 我們將介紹如何: - 使用 Next.js 建立 Web 應用程式, - 使用 React Spreadsheet 建立電子表格應用程式,以及 - 使用 CopilotKit 將 AI 整合到軟體應用程式中。 - 讓電子表格更容易使用、更有趣 ![圖片描述](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/ruxylrd07dga5ngw98np.gif) --- CopilotKit:建構應用內人工智慧副駕駛的框架 ========================== CopilotKit是一個[開源的AI副駕駛平台](https://github.com/CopilotKit/CopilotKit)。我們可以輕鬆地將強大的人工智慧整合到您的 React 應用程式中。 建造: - ChatBot:上下文感知的應用內聊天機器人,可以在應用程式內執行操作 💬 - CopilotTextArea:人工智慧驅動的文字字段,具有上下文感知自動完成和插入功能📝 - 聯合代理:應用程式內人工智慧代理,可以與您的應用程式和使用者互動🤖 ![圖片描述](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/x3us3vc140aun0dvrdof.gif) {% cta https://github.com/CopilotKit/CopilotKit %} Star CopilotKit ⭐️ {% endcta %} (請原諒 AI 的拼字錯誤並給 CopilotKit 加上星號:) 現在回到文章! --- 先決條件 ---- 要完全理解本教程,您需要對 React 或 Next.js 有基本的了解。 以下是建立人工智慧驅動的電子表格應用程式所需的工具: - [React Spreadsheet](https://github.com/iddan/react-spreadsheet) - 一個簡單的包,使我們能夠在 React 應用程式中加入電子表格。 - [OpenAI API](https://platform.openai.com/api-keys) - 提供 API 金鑰,使我們能夠使用 ChatGPT 模型執行各種任務。 - [Tavily AI](https://tavily.com/) - 一個搜尋引擎,使人工智慧代理能夠在應用程式中進行研究並存取即時知識。 - [CopilotKit](https://github.com/CopilotKit) - 一個開源副駕駛框架,用於建立自訂 AI 聊天機器人、應用程式內 AI 代理程式和文字區域。 專案設定和套件安裝 --------- 首先,透過在終端機中執行以下程式碼片段來建立 Next.js 應用程式: ``` npx create-next-app spreadsheet-app ``` 選擇您首選的配置設定。在本教學中,我們將使用 TypeScript 和 Next.js App Router。 ![圖片描述](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/f2zg8z22tdgmlv4wmfil.png) 接下來,安裝[OpenAI 函式庫](https://platform.openai.com/docs/introduction)、 [Heroicons](https://www.npmjs.com/package/@heroicons/react)和[React Spreadsheet](https://github.com/iddan/react-spreadsheet)套件及其相依性: ``` npm install openai react-spreadsheet scheduler @heroicons/react ``` 最後,安裝 CopilotKit 軟體套件。這些套件使我們能夠從 React 狀態檢索資料並將 AI copilot 新增至應用程式。 ``` npm install @copilotkit/react-ui @copilotkit/react-textarea @copilotkit/react-core @copilotkit/backend ``` 恭喜!您現在已準備好建立應用程式。 --- 建立電子表格應用程式 ---------- 在本節中,我將引導您使用 React Spreadsheet 建立電子表格應用程式。 該應用程式分為兩個元件: `Sidebar`和`SingleSpreadsheet` 。 要設定這些元件,請導航至 Next.js 應用程式資料夾並建立一個包含以下檔案的`components`資料夾: ``` cd app mkdir components && cd components touch Sidebar.tsx SingleSpreadsheet.tsx ``` ![圖片描述](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/bd9crph3qdenb2eikfoh.png) 將新建立的元件匯入到**`app/page.tsx`**檔案中。 ``` "use client"; import React, { useState } from "react"; //👇🏻 import the components import { SpreadsheetData } from "./types"; import Sidebar from "./components/Sidebar"; import SingleSpreadsheet from "./components/SingleSpreadsheet"; const Main = () => { return ( <div className='flex'> <p>Hello world</p> </div> ); }; export default Main; ``` 接下來,建立將包含電子表格資料的 React 狀態,並將它們作為 props 傳遞到元件中。 ``` const Main = () => { //👇🏻 holds the title and data within a spreadsheet const [spreadsheets, setSpreadsheets] = React.useState<SpreadsheetData[]>([ { title: "Spreadsheet 1", data: [ [{ value: "" }, { value: "" }, { value: "" }], [{ value: "" }, { value: "" }, { value: "" }], [{ value: "" }, { value: "" }, { value: "" }], ], }, ]); //👇🏻 represents the index of a spreadsheet const [selectedSpreadsheetIndex, setSelectedSpreadsheetIndex] = useState(0); return ( <div className='flex'> <Sidebar spreadsheets={spreadsheets} selectedSpreadsheetIndex={selectedSpreadsheetIndex} setSelectedSpreadsheetIndex={setSelectedSpreadsheetIndex} /> <SingleSpreadsheet spreadsheet={spreadsheets[selectedSpreadsheetIndex]} setSpreadsheet={(spreadsheet) => { setSpreadsheets((prev) => { console.log("setSpreadsheet", spreadsheet); const newSpreadsheets = [...prev]; newSpreadsheets[selectedSpreadsheetIndex] = spreadsheet; return newSpreadsheets; }); }} /> </div> ); }; ``` 此程式碼片段建立了 React 狀態,用於保存電子表格資料及其索引,並將它們作為 props 傳遞到元件中。 `Sidebar`元件接受所有可用的電子表格, `SingleSpreadsheet`元件接收所有電子表格,包括更新電子表格資料的`setSpreadsheet`函數。 將下面的程式碼片段複製到`Sidebar.tsx`檔案中。它顯示應用程式中的所有電子表格,並允許使用者在它們之間進行切換。 ``` import React from "react"; import { SpreadsheetData } from "../types"; interface SidebarProps { spreadsheets: SpreadsheetData[]; selectedSpreadsheetIndex: number; setSelectedSpreadsheetIndex: (index: number) => void; } const Sidebar = ({ spreadsheets, selectedSpreadsheetIndex, setSelectedSpreadsheetIndex, }: SidebarProps) => { return ( <div className='w-64 h-screen bg-gray-800 text-white overflow-auto p-5'> <ul> {spreadsheets.map((spreadsheet, index) => ( <li key={index} className={`mb-4 cursor-pointer ${ index === selectedSpreadsheetIndex ? "ring-2 ring-blue-500 ring-inset p-3 rounded-lg" : "p-3" }`} onClick={() => setSelectedSpreadsheetIndex(index)} > {spreadsheet.title} </li> ))} </ul> </div> ); }; export default Sidebar; ``` ![圖片描述](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/c36qccuwa0knqokag5mk.gif) 更新`SingleSpreadsheet.tsx`文件,如下所示: ``` import React from "react"; import Spreadsheet from "react-spreadsheet"; import { SpreadsheetData, SpreadsheetRow } from "../types"; interface MainAreaProps { spreadsheet: SpreadsheetData; setSpreadsheet: (spreadsheet: SpreadsheetData) => void; } //👇🏻 adds a new row to the spreadsheet const addRow = () => { const numberOfColumns = spreadsheet.rows[0].length; const newRow: SpreadsheetRow = []; for (let i = 0; i < numberOfColumns; i++) { newRow.push({ value: "" }); } setSpreadsheet({ ...spreadsheet, rows: [...spreadsheet.rows, newRow], }); }; //👇🏻 adds a new column to the spreadsheet const addColumn = () => { const spreadsheetData = [...spreadsheet.data]; for (let i = 0; i < spreadsheet.data.length; i++) { spreadsheet.data[i].push({ value: "" }); } setSpreadsheet({ ...spreadsheet, data: spreadsheetData, }); }; const SingleSpreadsheet = ({ spreadsheet, setSpreadsheet }: MainAreaProps) => { return ( <div className='flex-1 overflow-auto p-5'> {/** -- Spreadsheet title ---*/} <div className='flex items-start'> {/** -- Spreadsheet rows and columns---*/} {/** -- Add column button ---*/} </div> {/** -- Add row button ---*/} </div> ); }; export default SingleSpreadsheet; ``` - 從上面的程式碼片段來看, ``` - The `SingleSpreadsheet.tsx` file includes the addRow and addColumn functions. ``` ``` - The `addRow` function calculates the current number of rows, adds a new row, and updates the spreadsheet accordingly. ``` ``` - Similarly, the `addColumn` function adds a new column to the spreadsheet. ``` ``` - The `SingleSpreadsheet` component renders placeholders for the user interface elements. ``` 更新`SingleSpreadsheet`元件以呈現電子表格標題、其資料以及新增行和列按鈕。 ``` return ( <div className='flex-1 overflow-auto p-5'> {/** -- Spreadsheet title ---*/} <input type='text' value={spreadsheet.title} className='w-full p-2 mb-5 text-center text-2xl font-bold outline-none bg-transparent' onChange={(e) => setSpreadsheet({ ...spreadsheet, title: e.target.value }) } /> {/** -- Spreadsheet rows and columns---*/} <div className='flex items-start'> <Spreadsheet data={spreadsheet.data} onChange={(data) => { console.log("data", data); setSpreadsheet({ ...spreadsheet, data: data as any }); }} /> {/** -- Add column button ---*/} <button className='bg-blue-500 text-white rounded-lg ml-6 w-8 h-8 mt-0.5' onClick={addColumn} > + </button> </div> {/** -- Add row button ---*/} <button className='bg-blue-500 text-white rounded-lg w-8 h-8 mt-5 ' onClick={addRow} > + </button> </div> ); ``` 為了確保一切按預期工作,請在`app`資料夾中建立一個`types.ts`文件,其中包含應用程式中聲明的所有靜態類型。 ``` export interface Cell { value: string; } export type SpreadsheetRow = Cell[]; export interface SpreadsheetData { title: string; rows: SpreadsheetRow[]; } ``` 恭喜! 🎉 您的電子表格應用程式應該可以完美執行。在接下來的部分中,您將了解如何新增 AI 副駕駛,以使用 CopilotKit 自動執行各種任務。 --- 使用 CopilotKit 改進應用程式功能 ---------------------- 在這裡,您將學習如何將 AI 副駕駛加入到電子表格應用程式,以使用 CopilotKit 自動執行複雜的操作。 CopilotKit 提供前端和[後端](https://docs.copilotkit.ai/getting-started/quickstart-backend)套件。它們使您能夠插入 React 狀態並使用 AI 代理在後端處理應用程式資料。 首先,我們將 CopilotKit React 元件新增到應用程式前端。 ### 將 CopilotKit 加入前端 在`app/page.tsx`中,將以下程式碼片段加入`Main`元件的頂部。 ``` import "@copilotkit/react-ui/styles.css"; import { CopilotKit } from "@copilotkit/react-core"; import { CopilotSidebar } from "@copilotkit/react-ui"; import { INSTRUCTIONS } from "./instructions"; const HomePage = () => { return ( <CopilotKit url='/api/copilotkit'> <CopilotSidebar instructions={INSTRUCTIONS} labels={{ initial: "Welcome to the spreadsheet app! How can I help you?", }} defaultOpen={true} clickOutsideToClose={false} > <Main /> </CopilotSidebar> </CopilotKit> ); }; const Main = () => { //--- Main component // }; export default HomePage; ``` - 從上面的程式碼片段來看, ``` - I imported the CopilotKit, its sidebar component, and CSS file to use its frontend components within the application. ``` ``` - The [CopilotKit component](https://docs.copilotkit.ai/reference/CopilotKit) accepts a `url` prop that represents the API server route where CopilotKit will be configured. ``` ``` - The Copilot component also renders the [CopilotSidebar component](https://docs.copilotkit.ai/reference/CopilotSidebar) , allowing users to provide custom instructions to the AI copilot within the application. ``` ``` - Lastly, you can export the `HomePage` component containing the `CopilotSidebar` and the `Main` components. ``` 從上面的程式碼片段中,您會注意到`CopilotSidebar`元件有一個`instructions`屬性。此屬性使您能夠為 CopilotKit 提供額外的上下文或指導。 因此,在`app`資料夾中建立`instructions.ts`檔案並將這些[命令](https://github.com/CopilotKit/spreadsheet-demo/blob/main/src/app/instructions.ts)複製到該檔案中。 ![圖片描述](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/whs8k2ly9as7j4wn6h5o.png) 接下來,您需要將 CopilotKit 插入應用程式的狀態以存取應用程式的資料。為了實現這一點,CopilotKit 提供了兩個鉤子: [useCopilotAction](https://docs.copilotkit.ai/reference/useCopilotAction)和[useMakeCopilotReadable](https://docs.copilotkit.ai/reference/useMakeCopilotReadable) 。 [useCopilotAction](https://docs.copilotkit.ai/reference/useCopilotAction)掛鉤可讓您定義 CopilotKit 執行的動作。它接受包含以下參數的物件: - `name` - 操作的名稱。 - `description` - 操作的描述。 - `parameters` - 包含所需參數清單的陣列。 - `render` - 預設的自訂函數或字串。 - `handler` - 由操作觸發的可執行函數。 ``` useCopilotAction({ name: "sayHello", description: "Say hello to someone.", parameters: [ { name: "name", type: "string", description: "name of the person to say greet", }, ], render: "Process greeting message...", handler: async ({ name }) => { alert(`Hello, ${name}!`); }, }); ``` [useMakeCopilotReadable](https://docs.copilotkit.ai/reference/useMakeCopilotReadable)掛鉤向 CopilotKit 提供應用程式狀態。 ``` import { useMakeCopilotReadable } from "@copilotkit/react-core"; const appState = ...; useMakeCopilotReadable(JSON.stringify(appState)); ``` 現在,讓我們回到電子表格應用程式。在`SingleSpreadsheet`元件中,將應用程式狀態傳遞到 CopilotKit 中,如下所示。 ``` import { useCopilotAction, useMakeCopilotReadable, } from "@copilotkit/react-core"; const SingleSpreadsheet = ({ spreadsheet, setSpreadsheet }: MainAreaProps) => { //👇🏻 hook for providing the application state useMakeCopilotReadable( "This is the current spreadsheet: " + JSON.stringify(spreadsheet) ); // --- other lines of code }; ``` 接下來,您需要在`SingleSpreadsheet`元件中新增兩個操作,該元件在使用者更新電子表格資料並使用 CopilotKit 新增資料行時執行。 在繼續之前,請在`app`資料夾中建立一個包含`canonicalSpreadsheetData.ts`檔案的`utils`資料夾。 ``` cd app mkdir utils && cd utils touch canonicalSpreadsheetData.ts ``` 將下面的程式碼片段複製到檔案中。它接受對電子表格所做的更新,並將其轉換為電子表格中資料行所需的格式。 ``` import { SpreadsheetRow } from "../types" export interface RowLike { cells: CellLike[] | undefined; } export interface CellLike { value: string; } export function canonicalSpreadsheetData( rows: RowLike[] | undefined ): SpreadsheetRow[] { const canonicalRows: SpreadsheetRow[] = []; for (const row of rows || []) { const canonicalRow: SpreadsheetRow = []; for (const cell of row.cells || []) { canonicalRow.push({value: cell.value}); } canonicalRows.push(canonicalRow); } return canonicalRows; } ``` 現在,讓我們使用`SingleSpreadsheet`元件中的`useCopilotAction`掛鉤建立操作。複製下面的第一個操作: ``` import { canonicalSpreadsheetData } from "../utils/canonicalSpreadsheetData"; import { PreviewSpreadsheetChanges } from "./PreviewSpreadsheetChanges"; import { SpreadsheetData, SpreadsheetRow } from "../types"; import { useCopilotAction } from "@copilotkit/react-core"; useCopilotAction({ name: "suggestSpreadsheetOverride", description: "Suggest an override of the current spreadsheet", parameters: [ { name: "rows", type: "object[]", description: "The rows of the spreadsheet", attributes: [ { name: "cells", type: "object[]", description: "The cells of the row", attributes: [ { name: "value", type: "string", description: "The value of the cell", }, ], }, ], }, { name: "title", type: "string", description: "The title of the spreadsheet", required: false, }, ], render: (props) => { const { rows } = props.args const newRows = canonicalSpreadsheetData(rows); return ( <PreviewSpreadsheetChanges preCommitTitle="Replace contents" postCommitTitle="Changes committed" newRows={newRows} commit={(rows) => { const updatedSpreadsheet: SpreadsheetData = { title: spreadsheet.title, rows: rows, }; setSpreadsheet(updatedSpreadsheet); }} /> ) }, handler: ({ rows, title }) => { // Do nothing. // The preview component will optionally handle committing the changes. }, }); ``` 上面的程式碼片段執行使用者的任務並使用 CopilotKit 產生 UI 功能顯示結果預覽。 `suggestSpreadsheetOverride`操作傳回一個自訂元件 ( `PreviewSpreadsheetChanges` ),該元件接受以下內容為 props: - 要新增到電子表格的新資料行, - 一些文字 - `preCommitTitle`和`postCommitTitle` ,以及 - 更新電子表格的`commit`函數。 您很快就會學會如何使用它們。 在元件資料夾中建立`PreviewSpreadsheetChanges`元件,並將下列程式碼片段複製到檔案中: ``` import { CheckCircleIcon } from '@heroicons/react/20/solid' import { SpreadsheetRow } from '../types'; import { useState } from 'react'; import Spreadsheet from 'react-spreadsheet'; export interface PreviewSpreadsheetChanges { preCommitTitle: string; postCommitTitle: string; newRows: SpreadsheetRow[]; commit: (rows: SpreadsheetRow[]) => void; } export function PreviewSpreadsheetChanges(props: PreviewSpreadsheetChanges) { const [changesCommitted, setChangesCommitted] = useState(false); const commitChangesButton = () => { return ( <button className="inline-flex items-center gap-x-2 rounded-md bg-indigo-600 px-3.5 py-2.5 text-sm font-semibold text-white shadow-sm hover:bg-indigo-500 focus-visible:outline focus-visible:outline-2 focus-visible:outline-offset-2 focus-visible:outline-indigo-600" onClick={() => { props.commit(props.newRows); setChangesCommitted(true); }} > {props.preCommitTitle} </button> ); } const changesCommittedButtonPlaceholder = () => { return ( <button className=" inline-flex items-center gap-x-2 rounded-md bg-gray-100 px-3.5 py-2.5 text-sm font-semibold text-green-600 shadow-sm cursor-not-allowed" disabled > {props.postCommitTitle} <CheckCircleIcon className="-mr-0.5 h-5 w-5" aria-hidden="true" /> </button> ); } return ( <div className="flex flex-col"> <Spreadsheet data={props.newRows} /> <div className="mt-5"> {changesCommitted ? changesCommittedButtonPlaceholder() : commitChangesButton() } </div> </div> ); } ``` `PreviewSpreadsheetChanges`元件傳回一個電子表格,其中包含從請求產生的資料和一個按鈕(帶有`preCommitTitle`文字),該按鈕允許您將這些變更提交到主電子表格表(透過觸發`commit`函數)。這可確保您在將結果新增至電子表格之前對結果感到滿意。 將下面的第二個操作加入到`SingleSpreadsheet`元件。 ``` useCopilotAction({ name: "appendToSpreadsheet", description: "Append rows to the current spreadsheet", parameters: [ { name: "rows", type: "object[]", description: "The new rows of the spreadsheet", attributes: [ { name: "cells", type: "object[]", description: "The cells of the row", attributes: [ { name: "value", type: "string", description: "The value of the cell", }, ], }, ], }, ], render: (props) => { const status = props.status; const { rows } = props.args const newRows = canonicalSpreadsheetData(rows); return ( <div> <p>Status: {status}</p> <Spreadsheet data={newRows} /> </div> ) }, handler: ({ rows }) => { const canonicalRows = canonicalSpreadsheetData(rows); const updatedSpreadsheet: SpreadsheetData = { title: spreadsheet.title, rows: [...spreadsheet.rows, ...canonicalRows], }; setSpreadsheet(updatedSpreadsheet); }, }); ``` `appendToSpreadsheet`操作透過在電子表格中新增資料行來更新電子表格。 以下是操作的簡短示範: \[https://www.youtube.com/watch?v=kGQ9xl5mSoQ\] 最後,在`Main`元件中新增一個操作,以便在使用者提供指令時建立一個新的電子表格。 ``` useCopilotAction({ name: "createSpreadsheet", description: "Create a new spreadsheet", parameters: [ { name: "rows", type: "object[]", description: "The rows of the spreadsheet", attributes: [ { name: "cells", type: "object[]", description: "The cells of the row", attributes: [ { name: "value", type: "string", description: "The value of the cell", }, ], }, ], }, { name: "title", type: "string", description: "The title of the spreadsheet", }, ], render: (props) => { const { rows, title } = props.args; const newRows = canonicalSpreadsheetData(rows); return ( <PreviewSpreadsheetChanges preCommitTitle="Create spreadsheet" postCommitTitle="Spreadsheet created" newRows={newRows} commit={ (rows) => { const newSpreadsheet: SpreadsheetData = { title: title || "Untitled Spreadsheet", rows: rows, }; setSpreadsheets((prev) => [...prev, newSpreadsheet]); setSelectedSpreadsheetIndex(spreadsheets.length); }} /> ); }, handler: ({ rows, title }) => { // Do nothing. // The preview component will optionally handle committing the changes. }, }); ``` 恭喜!您已成功為此應用程式建立所需的操作。現在,讓我們將應用程式連接到 Copilotkit 後端。 ### 將 Tavily AI 和 OpenAI 加入到 CopilotKit 在本教程的開頭,我向您介紹了[Tavily AI](https://tavily.com/) (一個為 AI 代理提供知識的搜尋引擎)和 OpenAI(一個使我們能夠存取[GPT-4 AI 模型的](https://openai.com/gpt-4)庫)。 在本部分中,您將了解如何取得 Tavily 和 OpenAI API 金鑰並將它們整合到 CopilotKit 中以建立高級智慧應用程式。 造訪[Tavily AI 網站](https://app.tavily.com/sign-in),建立一個帳戶,然後將您的 API 金鑰複製到您專案的`.env.local`檔案中。 ![圖片描述](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/w8shpxr5hh9r9kggk1jv.png) 接下來,導覽至[OpenAI 開發者平台](https://platform.openai.com/api-keys),建立 API 金鑰,並將其複製到`.env.local`檔案中。 ![圖片描述](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/f4ugtobr70wg6z6ru3cj.png) 以下是`.env.local`檔案的預覽,其中包括 API 金鑰並指定要使用的 OpenAI 模型。請注意,存取 GPT-4 模型需要[訂閱 ChatGPT Plus](https://openai.com/chatgpt/pricing) 。 ``` TAVILY_API_KEY=<your_API_key> OPENAI_MODEL=gpt-4-1106-preview OPENAI_API_KEY=<your_API_key> ``` 回到我們的應用程式,您需要為 Copilot 建立 API 路由。因此,建立一個包含`route.ts`的`api/copilotkit`資料夾並新增一個`tavily.ts`檔案。 ``` cd app mkdir api && cd api mkdir copilotkit && cd copilotkit touch route.ts tavily.ts ``` 在`tavily.ts`檔案中建立一個函數,該函數接受使用者的查詢,使用 Tavily Search API 對查詢進行研究,並使用[OpenAI GPT-4 模型](https://openai.com/gpt-4)總結結果。 ``` import OpenAI from "openai"; export async function research(query: string) { //👇🏻 sends the request to the Tavily Search API const response = await fetch("https://api.tavily.com/search", { method: "POST", headers: { "Content-Type": "application/json", }, body: JSON.stringify({ api_key: process.env.TAVILY_API_KEY, query, search_depth: "basic", include_answer: true, include_images: false, include_raw_content: false, max_results: 20, }), }); //👇🏻 the response const responseJson = await response.json(); const openai = new OpenAI({ apiKey: process.env.OPENAI_API_KEY! }); //👇🏻 passes the response into the OpenAI GPT-4 model const completion = await openai.chat.completions.create({ messages: [ { role: "system", content: `Summarize the following JSON to answer the research query \`"${query}"\`: ${JSON.stringify( responseJson )} in plain English.`, }, ], model: process.env.OPENAI_MODEL, }); //👇🏻 returns the result return completion.choices[0].message.content; } ``` 最後,您可以透過將使用者的查詢傳遞到函數中並向 CopilotKit 提供其回應來執行`route.ts`檔案中的`research`函數。 ``` import { CopilotBackend, OpenAIAdapter } from "@copilotkit/backend"; import { Action } from "@copilotkit/shared"; import { research } from "./tavily"; //👇🏻 carries out a research on the user's query const researchAction: Action<any> = { name: "research", description: "Call this function to conduct research on a certain query.", parameters: [ { name: "query", type: "string", description: "The query for doing research. 5 characters or longer. Might be multiple words", }, ], handler: async ({ query }) => { console.log("Research query: ", query); const result = await research(query); console.log("Research result: ", result); return result; }, }; export async function POST(req: Request): Promise<Response> { const actions: Action<any>[] = []; if (process.env.TAVILY_API_KEY!) { actions.push(researchAction); } const copilotKit = new CopilotBackend({ actions: actions, }); const openaiModel = process.env.OPENAI_MODEL; return copilotKit.response(req, new OpenAIAdapter({ model: openaiModel })); } ``` 恭喜!您已完成本教學的專案。 結論 -- [CopilotKit](https://copilotkit.ai/)是一款令人難以置信的工具,可讓您在幾分鐘內將 AI Copilot 加入到您的產品中。無論您是對人工智慧聊天機器人和助理感興趣,還是對複雜任務的自動化感興趣,CopilotKit 都能讓您輕鬆實現。 如果您需要建立 AI 產品或將 AI 工具整合到您的軟體應用程式中,您應該考慮 CopilotKit。 您可以在 GitHub 上找到本教學的源程式碼: https://github.com/CopilotKit/spreadsheet-demo 感謝您的閱讀! --- 原文出處:https://dev.to/copilotkit/build-an-ai-powered-spreadsheet-app-nextjs-langchain-copilotkit-109d

React 元件設計模式 - 第 1 部分

React 設計模式是開發人員用來解決使用 React 建立應用程式時遇到的常見問題和挑戰的既定實踐和解決方案。這些模式封裝了針對重複出現的設計問題的可重複使用解決方案,從而提高了可維護性、可擴展性和效率。它們提供了一種結構化的方法來組織元件、管理狀態、處理資料流和最佳化效能。 **請考慮以下 6 種 React 設計模式:** 1. 容器和呈現模式 2. HOC(高階元件)模式 3. 複合元件模式 4. 提供者模式(使用提供者進行資料管理) 5. 狀態減速器模式 6. 元件組成模式 1. 容器和呈現模式 ---------- 在此模式中,容器元件負責管理資料和狀態邏輯。它們從外部來源獲取資料,在必要時對其進行操作,並將其作為 props 傳遞給展示元件。它們通常連接到外部服務、Redux 儲存或上下文提供者。 另一方面,展示元件僅專注於 UI 元素的展示。他們透過 props 從容器元件接收資料,並以視覺上吸引人的方式呈現它。展示元件通常是無狀態的功能元件或純元件,這使得它們更容易測試和重複使用。 **讓我們考慮一個複雜的範例來說明這些模式:** 假設我們正在建立一個社群媒體儀表板,用戶可以在其中查看朋友的貼文並與他們互動。以下是我們建立元件的方式: **容器元件(FriendFeedContainer):** 該元件將負責從 API 獲取有關好友貼文的資料、處理任何必要的資料轉換以及管理提要的狀態。它將相關資料傳遞給展示元件。 ``` import React, { useState, useEffect } from 'react'; import FriendFeed from './FriendFeed'; const FriendFeedContainer = () => { const [friendPosts, setFriendPosts] = useState([]); useEffect(() => { // Fetch friend posts from API const fetchFriendPosts = async () => { const posts = await fetch('https://api.example.com/friend-posts'); const data = await posts.json(); setFriendPosts(data); }; fetchFriendPosts(); }, []); return <FriendFeed posts={friendPosts} />; }; export default FriendFeedContainer; ``` **展示元件(FriendFeed):** 該元件將從其父容器元件 (FriendFeedContainer) 接收好友貼文資料作為 props,並以視覺上吸引人的方式呈現它們。 ``` import React from 'react'; const FriendFeed = ({ posts }) => { return ( <div> <h2>Friend Feed</h2> <ul> {posts.map(post => ( <li key={post.id}> <p>{post.content}</p> <p>Posted by: {post.author}</p> </li> ))} </ul> </div> ); }; export default FriendFeed; ``` 透過以這種方式建立我們的元件,我們將獲取資料和管理狀態的問題與 UI 渲染邏輯分開。這種分離使得我們的 React 應用程式在擴充時可以更輕鬆地進行測試、重複使用和維護。 2.HOC(高階元件)模式 ------------- 高階元件 (HOC) 是 React 中的一種模式,可讓您跨多個元件重複使用元件邏輯。它們是接受元件並傳回具有附加功能的新元件的函數。 為了示範 HOC 在具有 React hook 的社群媒體儀表板範例中的使用,讓我們考慮一個場景,其中您有多個元件需要從 API 取得使用者資料。您可以建立一個 HOC 來處理資料獲取並將獲取的資料作為 props 傳遞給包裝的元件,而不是在每個元件中重複取得邏輯。 這是一個基本範例: ``` import React, { useState, useEffect } from 'react'; // Define a higher-order component for fetching user data const withUserData = (WrappedComponent) => { return (props) => { const [userData, setUserData] = useState(null); const [loading, setLoading] = useState(true); useEffect(() => { // Simulate fetching user data from an API const fetchData = async () => { try { const response = await fetch('https://api.example.com/user'); const data = await response.json(); setUserData(data); setLoading(false); } catch (error) { console.error('Error fetching user data:', error); setLoading(false); } }; fetchData(); }, []); return ( <div> {loading ? ( <p>Loading...</p> ) : ( <WrappedComponent {...props} userData={userData} /> )} </div> ); }; }; // Create a component to display user data const UserProfile = ({ userData }) => { return ( <div> <h2>User Profile</h2> {userData && ( <div> <p>Name: {userData.name}</p> <p>Email: {userData.email}</p> {/* Additional user data fields */} </div> )} </div> ); }; // Wrap the UserProfile component with the withUserData HOC const UserProfileWithUserData = withUserData(UserProfile); // Main component where you can render the wrapped component const SocialMediaDashboard = () => { return ( <div> <h1>Social Media Dashboard</h1> <UserProfileWithUserData /> </div> ); }; export default SocialMediaDashboard; ``` 在這個例子中: - `withUserData`是一個高階元件,用於處理從 API 取得使用者資料。它包裝傳遞的元件 ( `WrappedComponent` ) 並將取得的使用者資料作為 prop ( `userData` ) 提供給它。 - `UserProfile`是一個功能元件,它接收`userData`屬性並顯示使用者設定檔資訊。 - `UserProfileWithUserData`是透過使用`withUserData`包裝`UserProfile`傳回的元件。 - `SocialMediaDashboard`是主要元件,您可以在其中呈現`UserProfileWithUserData`或任何其他需要使用者資料的元件。 使用此模式,您可以輕鬆地跨社交媒體儀表板應用程式中的多個元件重複使用資料取得邏輯,而無需重複程式碼。 3. 複合元件模式 --------- React 中的複合元件模式是一種設計模式,可讓您建立協同工作以形成有凝聚力的 UI 的元件,同時仍保持明確的關注點分離並提供自訂元件行為和外觀的靈活性。 在此模式中,父元件充當一個或多個子元件(稱為「複合元件」)的容器。這些子元件協同工作以實現特定的功能或行為。複合元件的關鍵特徵是它們透過父元件彼此共享狀態和功能。 以下是使用 hooks 在 React 中實作複合元件模式的簡單範例: ``` import React, { useState } from 'react'; // Parent component that holds the compound components const Toggle = ({ children }) => { const [isOn, setIsOn] = useState(false); // Function to toggle the state const toggle = () => { setIsOn((prevIsOn) => !prevIsOn); }; // Clone the children and pass the toggle function and state to them const childrenWithProps = React.Children.map(children, (child) => { if (React.isValidElement(child)) { return React.cloneElement(child, { isOn, toggle }); } return child; }); return <div>{childrenWithProps}</div>; }; // Child component for the toggle button const ToggleButton = ({ isOn, toggle }) => { return ( <button onClick={toggle}> {isOn ? 'Turn Off' : 'Turn On'} </button> ); }; // Child component for the toggle status const ToggleStatus = ({ isOn }) => { return <p>The toggle is {isOn ? 'on' : 'off'}.</p>; }; // Main component where you use the compound components const App = () => { return ( <Toggle> <ToggleStatus /> <ToggleButton /> </Toggle> ); }; export default App; ``` 在這個例子中: - `Toggle`是保存複合元件( `ToggleButton`和`ToggleStatus` )的父元件。 - `ToggleButton`是負責渲染切換按鈕的子元件。 - `ToggleStatus`是另一個負責顯示切換狀態的子元件。 - `Toggle`元件管理狀態 ( `isOn` ) 並提供`toggle`功能來控制狀態。它克隆其子級並將`isOn`狀態和`toggle`函數作為 props 傳遞給它們。 透過使用複合元件模式,您可以建立可重複使用和可組合的元件,封裝複雜的 UI 邏輯,同時仍允許自訂和靈活性。 ### 4. Provider Pattern(使用Provider進行資料管理) React 中的提供者模式是一種設計模式,用於跨多個元件管理和共用應用程式狀態或資料。它涉及建立一個封裝狀態或資料的提供者元件,並透過 React 的上下文 API 將其提供給其後代元件。 讓我們透過一個範例來說明 React 中用於管理使用者身份驗證資料的 Provider 模式: ``` // UserContext.js import React, { createContext, useState } from 'react'; // Create a context for user data const UserContext = createContext(); // Provider component export const UserProvider = ({ children }) => { const [user, setUser] = useState(null); // Function to login user const login = (userData) => { setUser(userData); }; // Function to logout user const logout = () => { setUser(null); }; return ( <UserContext.Provider value={{ user, login, logout }}> {children} </UserContext.Provider> ); }; export default UserContext; ``` 在這個例子中: - 我們使用 React 的`createContext`函數來建立一個名為`UserContext`上下文。此上下文將用於跨元件共享用戶資料和與身份驗證相關的功能。 - 我們定義一個`UserProvider`元件作為`UserContext`的提供者。該元件使用`useState`鉤子管理使用者狀態,並提供`login`和`logout`等方法來更新使用者狀態。 - 在`UserProvider`內部,我們用`UserContext.Provider`包裝`children` ,並將`user`狀態以及`login`和`logout`函數作為提供者的值傳遞。 - 現在,任何需要存取使用者資料或驗證相關功能的元件都可以使用`useContext`掛鉤來使用`UserContext` 。 讓我們建立一個使用上下文中的使用者資料的元件: ``` // UserProfile.js import React, { useContext } from 'react'; import UserContext from './UserContext'; const UserProfile = () => { const { user, logout } = useContext(UserContext); return ( <div> {user ? ( <div> <h2>Welcome, {user.username}!</h2> <button onClick={logout}>Logout</button> </div> ) : ( <div> <h2>Please log in</h2> </div> )} </div> ); }; export default UserProfile; ``` 在此元件中: 我們匯入`UserContext`並使用`useContext`鉤子來存取`UserProvider`提供的使用者資料和`logout`功能。 根據使用者是否登錄,我們呈現不同的 UI 元素。 最後,我們用`UserProvider`包裝我們的應用程式,以使用戶資料和身份驗證相關的功能可供所有元件使用: ``` // App.js import React from 'react'; import { UserProvider } from './UserContext'; import UserProfile from './UserProfile'; const App = () => { return ( <UserProvider> <div> <h1>My App</h1> <UserProfile /> </div> </UserProvider> ); }; export default App; ``` 透過這種方式,Provider 模式允許我們跨多個元件管理和共享應用程式狀態或資料,而無需進行 prop 鑽取,從而使我們的程式碼更乾淨、更易於維護。 --- 原文出處:https://dev.to/fpaghar/react-component-design-patterns-part-1-5f0g

建立完美人工智慧應用所需的所有工具。

過去十年來,人工智慧世界取得了長足發展。 人工智慧無所不在,從語音助理到軟體開發,如果我們正確使用它,它會非常有幫助。 在這樣的世界中,製作 AI 應用程式是有利可圖的,因此我在這裡介紹 25 個開源專案,您可以使用它們來製作 AI 應用程式並將其提升到新的水平。 其中有一些令人興奮的概念,例如使用語音合成與 3D 角色進行互動式溝通。堅持到底。 將會有大量的資源、文章、專案想法、指南等可供參考。 讓我們涵蓋這一切! --- 1. [Taipy](https://github.com/Avaiga/taipy) - 將資料和人工智慧演算法整合到生產就緒的 Web 應用程式中。 ---------------------------------------------------------------------------- ![打字](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/deak7rre409rzv5j5viv.png) Taipy 是一個開源 Python 庫,可用於輕鬆的端到端應用程式開發,具有假設分析、智慧管道執行、內建調度和部署工具。 我相信你們大多數人都不明白 Taipy 用於為基於 Python 的應用程式建立 GUI 介面並改進資料流管理。 因此,您可以繪製資料集的圖表,並使用類似 GUI 的滑桿來提供使用其他實用功能來處理資料的選項。 雖然 Streamlit 是一種流行的工具,但在處理大型資料集時,其效能可能會顯著下降,這使得它在生產級使用上不切實際。 另一方面,Taipy 在不犧牲性能的情況下提供了簡單性和易用性。透過嘗試 Taipy,您將親身體驗其用戶友好的介面和高效的資料處理。 在底層,Taipy 利用各種函式庫來簡化開發並增強功能。 ![圖書館](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/n9xts3nof4uapr7dakrl.png) 開始使用以下命令。 ``` pip install taipy ``` 我們來談談最新的[Taipy v3.1 版本](https://docs.taipy.io/en/latest/relnotes/)。 最新版本使得在 Taipy 的多功能零件物件中可視化任何 HTML 或 Python 物件成為可能。 這意味著[Folium](https://python-visualization.github.io/folium/latest/) 、 [Bokeh](https://bokeh.org/) 、 [Vega-Altair](https://altair-viz.github.io/)和[Matplotlib](https://matplotlib.org/)等程式庫現在可用於視覺化。 這也帶來了對[Plotly python 的](https://plotly.com/python/)原生支持,使繪製圖表變得更加容易。 ![陰謀蟒蛇](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/xdewvex88md09hvu3s80.png) 他們還使用分散式運算提高了效能,但最好的部分是 Taipy,它的所有依賴項現在都與 Python 3.12 完全相容,因此您可以在使用 Taipy 進行專案的同時使用最新的工具和程式庫。 您可以閱讀[文件](https://docs.taipy.io/en/latest/)。 例如,您可以看到[聊天演示](https://docs.taipy.io/en/release-3.1/gallery/llm/5_chatbot/),它使用 OpenAI 的 GPT-4 API 來產生對您的訊息的回應。您可以輕鬆更改程式碼以使用任何其他 API 或模型。 ![聊天演示](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/kug1mclhmzyad0hjchif.png) 另一個有用的事情是,Taipy 團隊提供了一個名為[Taipy Studio](https://docs.taipy.io/en/latest/manuals/studio/)的 VSCode 擴充功能來加速 Taipy 應用程式的建置。 ![太皮工作室](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/kc1umm5hcxes0ydbuspb.png) 您也可以使用 Taipy 雲端部署應用程式。 如果您想閱讀部落格來了解程式碼庫結構,您可以閱讀 HuggingFace 的[使用 Taipy 在 Python 中為您的 LLM 建立 Web 介面](https://huggingface.co/blog/Alex1337/create-a-web-interface-for-your-llm-in-python)。 嘗試新技術通常很困難,但 Taipy 提供了[10 多個演示教程](https://docs.taipy.io/en/release-3.1/gallery/),其中包含程式碼和適當的文件供您遵循。 例如,一些現場演示範例: - [新冠儀表板](https://covid-dashboard.taipy.cloud/Country) - [推文生成](https://tweet-generation.taipy.cloud/) - [資料視覺化](https://production-planning.taipy.cloud/Data-Visualization) - [即時人臉辨識](https://face-recognition.taipy.cloud/) Taipy 在 GitHub 上有 7k+ Stars,並且處於`v3`版本,因此它們正在不斷改進。 ![利桑·阿爾·蓋布](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/m8etards1b7qfpbk2scr.png) https://github.com/Avaiga/taipy Star Taipy ⭐️ --- 2. [Supabase](https://github.com/supabase/supabase) - 開源 Firebase 替代品。 ---------------------------------------------------------------------- ![蘇帕貝斯](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/an2b9aqiij0j2tml1c6b.png) 要建立AI應用程式,您需要一個後端,而Supabase作為優秀的後端服務提供者可以滿足這一需求。 開始使用以下 npm 指令 (Next.js)。 ``` npx create-next-app -e with-supabase ``` 這就是使用 CRUD 操作的方式。 ``` import { createClient } from '@supabase/supabase-js' // Initialize const supabaseUrl = 'https://chat-room.supabase.co' const supabaseKey = 'public-anon-key' const supabase = createClient(supabaseUrl, supabaseKey) // Create a new chat room const newRoom = await supabase .from('rooms') .insert({ name: 'Supabase Fan Club', public: true }) // Get public rooms and their messages const publicRooms = await supabase .from('rooms') .select(` name, messages ( text ) `) .eq('public', true) // Update multiple users const updatedUsers = await supabase .from('users') .eq('account_type', 'paid') .update({ highlight_color: 'gold' }) ``` 您可以閱讀[文件](https://supabase.com/docs)。 您可以使用身份驗證、即時、邊緣功能、儲存等功能建立一個速度極快的應用程式。 Supabase 涵蓋了這一切! Supabase 也提供了幾個入門套件,例如[Nextjs 與 LangChain](https://github.com/langchain-ai/langchain-nextjs-template) 、 [Stripe 與 Nextjs](https://github.com/vercel/nextjs-subscription-payments)或[AI Chatbot](https://github.com/supabase-community/vercel-ai-chatbot) 。 Supabase 在 GitHub 上擁有超過 63,000 顆星,並且擁有大量提交超過 27,000 次的貢獻者。 https://github.com/supabase/supabase 明星 Supabase ⭐️ --- 3. [Chatwoot](https://github.com/chatwoot/chatwoot) - 即時聊天、電子郵件支援、全通路服務台並擁有您的資料。 -------------------------------------------------------------------------------- ![查特伍德](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/bpgjh0hdr5u5cpf2kdn7.png) Chatwoot 連接流行的客戶溝通管道,如電子郵件、網站即時聊天、Facebook、Twitter、WhatsApp、Instagram、Line 等。這有助於您從單一儀表板跨管道提供一致的客戶體驗。 這在各種情況下都可能很重要,例如當您圍繞人工智慧應用程式建立社群時。 ![聊天特烏功能](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/l0u3z2cdqvzhqb94h5zm.png) 您可以閱讀[文件](https://www.chatwoot.com/docs/product)來發現各種整合選項,以便更輕鬆地管理整個生態系統。 他們在每個整合中都有非常詳細的文件和快照範例,例如[帶有 WhatsApp Cloud API 的 WhatsApp 通道](https://www.chatwoot.com/docs/product/channels/whatsapp/whatsapp-cloud)。您可以根據需要一鍵式或自架部署到 Heroku。 他們在 GitHub 上擁有 18k+ Stars,並且發布了`v3.6`版本。 https://github.com/chatwoot/chatwoot 明星 Chatwoot ⭐️ --- 4. [CopilotKit](https://github.com/CopilotKit/CopilotKit) - 在數小時內為您的產品提供 AI Copilot。 ------------------------------------------------------------------------------------ ![副駕駛套件](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/nzuxjfog2ldam3csrl62.png) 您可以使用兩個 React 元件將關鍵 AI 功能整合到 React 應用程式中。它們還提供內建(完全可自訂)Copilot 原生 UX 元件,例如`<CopilotKit />` 、 `<CopilotPopup />` 、 `<CopilotSidebar />` 、 `<CopilotTextarea />` 。 開始使用以下 npm 指令。 ``` npm i @copilotkit/react-core @copilotkit/react-ui @copilotkit/react-textarea ``` 這是整合 CopilotTextArea 的方法。 ``` import { CopilotTextarea } from "@copilotkit/react-textarea"; import { useState } from "react"; export function SomeReactComponent() { const [text, setText] = useState(""); return ( <> <CopilotTextarea className="px-4 py-4" value={text} onValueChange={(value: string) => setText(value)} placeholder="What are your plans for your vacation?" autosuggestionsConfig={{ textareaPurpose: "Travel notes from the user's previous vacations. Likely written in a colloquial style, but adjust as needed.", chatApiConfigs: { suggestionsApiConfig: { forwardedParams: { max_tokens: 20, stop: [".", "?", "!"], }, }, }, }} /> </> ); } ``` 您可以閱讀[文件](https://docs.copilotkit.ai/getting-started/quickstart-textarea)。 基本概念是在幾分鐘內建立可用於基於 LLM 的全端應用程式的 AI 聊天機器人。 https://github.com/CopilotKit/CopilotKit Star CopilotKit ⭐️ --- 5. [DALL·E Mini](https://github.com/borisdayma/dalle-mini) - 根據文字提示產生圖像。 ------------------------------------------------------------------------ ![從文字生成圖像](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/mco3wf4nzc5j245aizpu.png) OpenAI 擁有第一個令人印象深刻的模型,用於使用 DALL·E 生成圖像。 Craiyon/DALL·E mini 嘗試使用開源模型重現這些結果。 如果您想知道這個名字,DALL-E mini 應母公司的要求更名為 Craiyon,並以更易於存取的網路應用程式格式使用類似的技術。 您可以在[Craiyon](https://www.craiyon.com/)上使用該模型。 ![蠟筆](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/ntjhsr9f7t1y0idlysjw.png) 開始使用以下命令(用於開發)。 ``` pip install dalle-mini ``` 您可以閱讀[文件](https://github.com/borisdayma/dalle-mini?tab=readme-ov-file#development)。 您可以閱讀[DALL-E Mini 解釋](https://wandb.ai/dalle-mini/dalle-mini/reports/DALL-E-Mini-Explained-with-Demo--Vmlldzo4NjIxODA)來了解有關資料集、架構和所涉及演算法的更多資訊。 您可以閱讀[最佳真實感 AI 圖像和提示的終極指南](https://www.craiyon.com/blog/ultimate-guide-best-ai-art-photorealistic-images-and-prompts),以便更好地理解優質資源。 DALL·E Mini 在 GitHub 上擁有 14k+ Stars,目前處於`v0.1`版本。 https://github.com/borisdayma/dalle-mini 明星 DALL·E Mini ⭐️ --- 6. [Deepgram](https://github.com/deepgram) - 將語音 AI 建置到您的應用程式中。 --------------------------------------------------------------- ![深度圖](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/32enxrtcwqk6g81eazay.png) 從新創公司到 NASA,Deepgram API 每天都用於轉錄和理解數百萬分鐘的音訊。快速、準確、可擴展且經濟高效。 它為開發人員提供語音到文字和音訊智慧模型。 ![深度圖選項](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/rdc3tqg7fvt3sw6ktle7.png) 儘管他們有免費增值模式,但免費套餐的限制足以讓您入門。 可視化效果更上一層樓。您可以檢查即時串流媒體回應或音訊檔案並比較音訊的智慧程度。 ![串流媒體](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/4wcvzzrqzn94gxe594hf.png) ![情緒分析](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/uw6wkhzg7g6vgq7lphri.png) 您可以閱讀[文件](https://developers.deepgram.com/docs/introduction)。 您也可以閱讀 Deepgram 撰寫的[關於如何將語音辨識新增至您的 React 和 Node.js 專案的](https://deepgram.com/learn/how-to-add-speech-recognition-to-your-react-project)範例部落格。 如果您想嘗試 API 來親自了解模型的靈活性,請查看他們的[API Playground](https://playground.deepgram.com/?smart_format=true&language=en&model=nova-2) 。 https://github.com/deepgram 明星 Deepgram ⭐️ --- 7. [InvokeAI](https://github.com/invoke-ai/InvokeAI) - 領先的穩定擴散模型創意引擎。 --------------------------------------------------------------------- ![呼叫人工智慧](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/a1uira3ta4ufauefp0ff.png) 關於 InvokeAI 是 Stable Diffusion(開源文字到圖像和圖像到圖像生成器)的實現。 它可以在 Windows、Mac 和 Linux 機器上執行,並在 RAM 低至 4 GB 的 GPU 卡上執行。 此解決方案提供業界領先的WebUI,支援透過CLI進行終端使用,並作為多種商業產品的基礎。 ![呼叫ai](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/g5802r0wtxlbkqdtclce.png) 您可以閱讀有關[安裝和硬體要求](https://invoke-ai.github.io/InvokeAI/installation/INSTALLATION/)、[如何安裝不同型號](https://invoke-ai.github.io/InvokeAI/installation/050_INSTALLING_MODELS/)以及最重要的[自動安裝的資訊](https://invoke-ai.github.io/InvokeAI/installation/010_INSTALL_AUTOMATED/)。 令人興奮的功能是能夠使用另一個圖像生成圖像,如[文件](https://invoke-ai.github.io/InvokeAI/features/IMG2IMG/)中所述。 InvokeAI 在 GitHub 上有近 21k 顆星, https://github.com/invoke-ai/InvokeAI 明星 InvokeAI ⭐️ --- 8. [OpenAI](https://github.com/openai) - 您所需要的一切。 ------------------------------------------------- ![開放人工智慧](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/k02duibi54zzzddck91z.png) Gemini by Google 和 OpenAI 非常受歡迎,但我們在此列表中專注於 OpenAI。 如果您想了解更多訊息,可以在 Medium 上閱讀[Google AI Gemini API in web using React 🤖](https://generativeai.pub/google-gemini-api-in-web-using-react-7e5bf0bf0abc) 。這很簡單,也很切中要害。 透過 OpenAI,您可以使用 DALL·E(根據文字描述建立原創、逼真的圖像和藝術)、Whisper(語音辨識模型)和 GPT-4。在評論中告訴我們關於索拉的事吧! 您可以使用簡單的 API 開始建置。 ``` completion = openai.ChatCompletion.create( model="gpt-3.5-turbo", messages=[ {"role": "system", "content": "You are a helpful assistant."}, {"role": "user", "content": "What are some famous astronomical observatories?"} ] ) ``` 您可以閱讀[文件](https://platform.openai.com/docs/introduction)。它提供瞭如此多的選項來建立非常酷的東西! ![文件概述](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/o9yi0tar96jxi4pkni81.png) 甚至 Stripe 也使用 GPT-4 來改善使用者體驗。 例如,您可以建立[Assistant 應用程式](https://platform.openai.com/docs/assistants/overview)並查看[API 遊樂場](https://platform.openai.com/playground/p/default-chat?model=text-davinci-003)以更好地理解它。 ![GPT-3](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/t89658h4drhy4a8zf3xs.png) 如果您需要指南,可以閱讀 Dzone 的[Integrating ChatGPT With ReactJS](https://dzone.com/articles/integrating-chatgpt-with-reactjs-a-comprehensive-g) 。 其間,OpenAI收購了Sora,獲得了壟斷地位。你怎麼認為? https://github.com/openai 明星 OpenAI ⭐️ --- 9. [DeepFaceLab](https://github.com/iperov/DeepFaceLab) - 用於建立深度贗品的領先軟體。 ------------------------------------------------------------------------ ![深臉實驗室](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/g32stb7uo201msv3jn8f.png) DeepFaceLab 是製作 Deepfakes 的頂級開源工具。 Deepfakes 是透過深度學習製作的經過修改的圖像和影片。它們經常被用來交換圖片或剪輯中的臉孔,有時是為了開玩笑,但也有出於有害的原因。 DeepFaceLab,用Python建置,是一個強大的deepfake工具。它可以改變媒體中的臉孔,甚至消除皺紋和老化跡象。 這些是您可以使用 DeepFaceLab 執行的一些操作。 - 換臉。 ![更換臉部](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/86jnuec9l6eaalwf9w51.png) - [臉部抗衰老 - YouTube](https://www.youtube.com/watch?v=Ddx5B-84ebo) 。 ![圖片描述](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/axh2e6117felh4zhoh3p.png) - 更換頭部。 ![更換頭部](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/nyvbncox7k1u28nait50.png) - 操縱嘴唇。 您可以使用這個基本教學來了解[如何有效地使用 DeepFaceLab](https://www.youtube.com/watch?v=kOIMXt8KK8M)來完成這些事情。 您可以在[YouTube](https://www.youtube.com/channel/UCGf4OlX_aTt8DlrgiH3jN3g/videos)上看到使用此 DeepLab 演算法的影片。 不幸的是,DeepFaceLab 中沒有「讓一切正常」按鈕,但值得根據您的特定需求了解其工作流程。 儘管它於 2023 年 11 月 9 日存檔,在 GitHub 上有近 44k+ 顆星,但由於其大量的教程和可靠的演算法,它仍然是您的 AI 應用程式的可靠選擇。 https://github.com/iperov/DeepFaceLab 明星 DeepFaceLab ⭐️ --- 10. [Detectron2](https://github.com/facebookresearch/detectron2) - 基於 PyTorch 的模組化物件偵測庫。 ---------------------------------------------------------------------------------------- ![探測器2](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/jxe7wuf8v8y7e039ziel.png) Detectron2 是 Facebook AI Research 的下一代函式庫,提供最先進的偵測和分割演算法。它是 Detectron 和 maskrcnn-benchmark 的後繼者。 它支援 Facebook 上的多個電腦視覺研究專案和生產應用程式。 使用此[YouTube 教學](https://www.youtube.com/watch?v=eUSgtfK4ivk)將 Detectron2 與 Facebook 開發者倡導者的機器學習結合使用。 Detectron2 旨在支援各種最先進的物件偵測和分割模型,同時也適應不斷發展的前沿研究領域。 您可以閱讀[如何入門](https://detectron2.readthedocs.io/en/latest/tutorials/getting_started.html)以及 [元博客](https://ai.meta.com/blog/-detectron2-a-pytorch-based-modular-object-detection-library-/),其中深入介紹了 Detectron 的目標。 舊版的 Detectron 使用的是 Caffe,因此很難與後來結合 Caffe2 和 PyTorch 的程式碼變更一起使用。為了回應社群回饋,Facebook AI 發布了 Detectron2 作為更新的、更容易使用的版本。 Detectron2 配備了用於物件偵測的先進演算法,例如 DensePose 和全景特徵金字塔網路。 此外,Detectron2 還可以進行語義分割和全景分割,這有助於更準確地偵測和分割影像和影片中的物件。 Detectron2 不僅支援使用邊界框和實例分割遮罩進行物件偵測,還可以預測人體姿勢,與 Detectron 類似。 它們在 GitHub 儲存庫上擁有 28k+ Stars,並在 GitHub 上被 1.6k+ 開發人員使用。 https://github.com/facebookresearch/detectron2 Star Detectron2 ⭐️ --- [11.FastAI-](https://github.com/fastai/fastai)深度學習庫。 ---------------------------------------------------- ![你真好](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/6qvxqd22q3qamqtvwe6p.png) Fastai 是一個多功能的深度學習庫,旨在滿足從業者和研究人員的需求。它為從業者提供了高級元件,以便他們在常見的深度學習任務中快速獲得一流的結果。 同時,它為研究人員提供低階元件來實驗和開發新方法。 Detectron2 透過其分層架構實現了易用性和靈活性之間的平衡。 該架構將複雜的深度學習技術分解為可管理的抽象,簡潔地利用了 Python 的動態特性和 PyTorch 的靈活性。 它建構在較低層級 API 的層次結構之上,這些 API 提供可組合的建構塊。這樣,想要重寫部分高級 API 或加入特定行為以滿足其需求的用戶無需學習如何使用最低級別。 ![架構API](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/kfooe2mxrh3xplcxeg75.png) [安裝 pyTorch](https://pytorch.org/get-started/locally/)後即可開始使用以下命令。 ``` conda install -c fastai fastai ``` 您可以閱讀[文件](https://docs.fast.ai/)。 它們針對初學者、中級和專家的[教程](https://docs.fast.ai/tutorial.html)有不同的起點。 如果您想為 FastAI 做出貢獻,您應該閱讀他們的[程式碼風格指南](https://docs.fast.ai/dev/style.html)。 如果您更喜歡影片,可以在 YouTube 上觀看傑里米霍華德 (Jeremy Howard) 撰寫的[課程“0”:程式設計師實用深度學習 (fastai)](https://www.youtube.com/watch?v=gGxe2mN3kAg) 。 它們在 GitHub 上擁有超過 25,000 顆星,並已被 GitHub 上超過 16,000 名開發人員使用。 https://github.com/fastai/fastai 明星 FastAI ⭐️ --- 12.[穩定擴散](https://github.com/CompVis/stable-diffusion)- 潛在文字到影像擴散模型。 -------------------------------------------------------------------- ![穩定擴散](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/63worvztgs1cmy2owtkf.png) > 什麼是穩定擴散? 穩定擴散是指生成模型中使用的一種技術,特別是在文字到圖像合成的背景下,其中將資訊從文字描述轉移到圖像的過程是逐漸且平滑地完成的。 在潛在文字到影像擴散模型中,穩定擴散可確保來自文字描述的訊息在整個模型的潛在空間中一致地擴散或傳播。這種擴散過程有助於產生與給定文字輸入相符的高品質和逼真的圖像。 穩定的擴散機制確保模型在生成過程中不會出現突然的跳躍或不穩定。我希望這能解決問題! 下載和採樣穩定擴散的簡單方法是使用[擴散器庫](https://github.com/huggingface/diffusers/tree/main#new--stable-diffusion-is-now-fully-compatible-with-diffusers)。 ``` # make sure you're logged in with `huggingface-cli login` from torch import autocast from diffusers import StableDiffusionPipeline pipe = StableDiffusionPipeline.from_pretrained( "CompVis/stable-diffusion-v1-4", use_auth_token=True ).to("cuda") prompt = "a photo of an astronaut riding a horse on mars" with autocast("cuda"): image = pipe(prompt)["sample"][0] image.save("astronaut_rides_horse.png") ``` 您可以閱讀[研究論文](https://ommer-lab.com/research/latent-diffusion-models/)以及有關[穩定擴散影像修改](https://github.com/CompVis/stable-diffusion?tab=readme-ov-file#image-modification-with-stable-diffusion)的更多資訊。 例如,這是輸入。 ![輸入](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/zpvxxhrrvthd8w1a0rrl.png) 這是放大一點後的輸出。 ![輸出](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/gzqvd06kse8ifhzry0la.png) Stable Diffusion v1 是一種特定的模型配置,它採用 860M UNet 和 CLIP ViT-L/14 文字編碼器進行擴散模型,並具有下採樣因子 8 自動編碼器。該模型在 256x256 影像上進行了預訓練,隨後在 512x512 影像上進行了微調。 他們在 GitHub 儲存庫上擁有大約 64k+ Stars。 https://github.com/CompVis/stable-diffusion 恆星穩定擴散 ⭐️ --- 13. [Mocap Drones](https://github.com/jyjblrd/Mocap-Drones) - 用於房間規模追蹤的低成本動作捕捉系統。 --------------------------------------------------------------------------------- ![動作捕捉無人機](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/3hq4hnzbx2wtxboehosi.png) 該專案需要 SFM(運動結構)OpenCV 模組,這需要您從原始程式碼編譯 OpenCV。 從`computer_code`目錄中,執行此命令來安裝節點相依性。 ``` yarn install yarn run dev // to start the web server. ``` 您將獲得前端介面的 URL 視圖。 開啟一個單獨的終端機視窗並執行命令`python3 api/index.py`來啟動後端伺服器。此伺服器負責接收攝影機串流並執行動作捕捉計算。 架構如下。 ![建築學](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/jt6e3f32scak65wfdp8s.png) 您可以觀看此[YouTube 影片](https://www.youtube.com/watch?v=0ql20JKrscQ)來了解 Mocap 無人機的工作原理,也可以觀看該專案所有者的[部落格](https://joshuabird.com/blog/post/mocap-drones)。 https://www.youtube.com/watch?v=0ql20JKrscQ 您可以閱讀[文件](https://github.com/jyjblrd/Mocap-Drones?tab=readme-ov-file#runing-the-code)。 這是一個最近的開源專案,在 GitHub 儲存庫上擁有 900 多個 star。 https://github.com/jyjblrd/Mocap-Drones 明星動捕無人機 ⭐️ --- 14. [Whisper Speech](https://github.com/collabora/WhisperSpeech) - 透過反轉 Whisper 建構的文字轉語音系統。 ------------------------------------------------------------------------------------------- ![低聲講話](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/hpawahh7aqsh1pnsnu76.png) 該模型與穩定擴散類似,但用於語音,功能強大且高度可自訂。 該團隊確保使用經過適當許可的語音錄音,並且所有程式碼都是開源的,使該模型對於商業應用程式來說是安全的。 目前,這些模型是在英語 LibreLight 資料集上進行訓練的。 您可以進一步研究[架構](https://github.com/collabora/WhisperSpeech?tab=readme-ov-file#architecture)。 ![建築學](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/hnfqick2y1yoxgkmwlk6.png) 您可以聽到[範例聲音](https://github.com/collabora/WhisperSpeech/assets/107984/aa5a1e7e-dc94-481f-8863-b022c7fd7434)並使用[colab](https://colab.research.google.com/drive/1xxGlTbwBmaY6GKA24strRixTXGBOlyiw)自行嘗試。 它們相當新,在 GitHub 上有大約 3k+ 的星星。 https://github.com/collabora/WhisperSpeech 星語語音 ⭐️ --- 15. [eSpeak NG](https://github.com/espeak-ng/espeak-ng) - 支援一百多種語言和口音的語音合成器。 ---------------------------------------------------------------------------- ![電子說](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/a28zdxcr1jthb5bht2fi.png) eSpeak NG 是一款緊湊型開源軟體文字語音合成器,適用於 Linux、Windows、Android 和其他作業系統。它支援 100 多種語言和口音。它基於 Jonathan Duddington 建立的 eSpeak 引擎。 您可以閱讀各種系統上的[安裝指南](https://github.com/espeak-ng/espeak-ng/blob/master/docs/guide.md)。 對於類似 Debian 的發行版(例如 Ubuntu、Mint 等)。您可以使用此命令。 ``` sudo apt-get install espeak-ng ``` 您可以查看[支援的語言](https://github.com/espeak-ng/espeak-ng/blob/master/docs/languages.md)清單、閱讀[文件](https://github.com/espeak-ng/espeak-ng/tree/master?tab=readme-ov-file#documentation)並查看[功能](https://github.com/espeak-ng/espeak-ng/tree/master?tab=readme-ov-file#features)。 該模型將文字轉換為音素程式碼,表明其作為另一個語音合成引擎前端的潛在能力。 他們在 GitHub 上有 2700+ 顆星星, https://github.com/espeak-ng/espeak-ng 明星 eSpeak NG ⭐️ --- 16.[聊天機器人 UI](https://github.com/mckaywrigley/chatbot-ui) - 每個模型的人工智慧聊天。 ------------------------------------------------------------------------ ![圖片描述](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/k8smowkv6scq9lujjeab.png) 我們都使用過 ChatGPT,這個專案可以幫助我們為任何 AI 聊天機器人設定使用者介面。少一麻煩! 你可以閱讀[安裝指南](https://github.com/mckaywrigley/chatbot-ui?tab=readme-ov-file#1-install-docker)來安裝 docker、supabase CLI 和其他東西。 您可以閱讀<a href="">文件</a>並查看[演示](https://twitter.com/mckaywrigley/status/1738273242283151777?s=20)。 這在底層使用了 Supabase (Postgres),這就是我們之前討論它的原因。 我沒有討論 Vercel AI 聊天機器人,因為它與此機器人相比是一個相當新的比較。 Chatbot UI 在 GitHub 上擁有大約 25k+ Stars,因此它仍然是開發人員為任何聊天機器人建立 UI 介面的首選。 https://github.com/mckaywrigley/chatbot-ui 明星聊天機器人 UI ⭐️ --- 17. [GPT-4 & LangChain](https://github.com/mayooear/gpt4-pdf-chatbot-langchain) - 用於大型 PDF 文件的 GPT4 和 LangChain 聊天機器人。 -------------------------------------------------------------------------------------------------------------------------- ![聊天架構](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/0pe0xehimhyw2mfubzu9.png) 這可用於新的 GPT-4 API 來為多個大型 PDF 檔案建立 chatGPT 聊天機器人。 該系統是使用 LangChain、Pinecone、Typescript、OpenAI 和 Next.js 建構的。 LangChain 是一個簡化可擴展 AI/LLM 應用程式和聊天機器人開發的框架。 Pinecone 用作向量存儲,用於以文字格式儲存嵌入和 PDF,以便以後檢索類似文件。 您可以閱讀涉及複製、安裝依賴項和設定環境 API 金鑰[的開發指南](https://github.com/mayooear/gpt4-pdf-chatbot-langchain?tab=readme-ov-file#development)。 您可以觀看[YouTube 影片](https://www.youtube.com/watch?v=ih9PBGVVOO4),了解如何遵循和使用它。 他們在 GitHub 上擁有 14k+ Stars,僅提交了 34 次。在您的下一個人工智慧應用程式中嘗試! https://github.com/mayooear/gpt4-pdf-chatbot-langchain 明星 GPT-4 和 Langchain ⭐️ --- 18. [Amica](https://github.com/semperai/amica) - 允許您在瀏覽器中輕鬆與 3D 角色聊天。 --------------------------------------------------------------------- ![朋友](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/2nvizcn717h3cteocft5.png) Amica 是一個開源接口,用於透過語音合成和語音辨識與 3D 角色進行互動式通訊。 您可以匯入 VRM 文件,調整聲音以適合角色,並產生包含情緒表達的回應文字。 他們使用 Three.js、OpenAI、Whisper、Bakllava 等進行視覺處理。您可以閱讀[Amica 的工作原理](https://docs.heyamica.com/overview/how-amica-works)及其所涉及的[核心概念](https://docs.heyamica.com/overview/core-concepts)。 您可以克隆該存儲庫並使用它來[開始](https://docs.heyamica.com/getting-started/installation)。 ``` npm i npm run dev ``` 您可以閱讀[文件](https://docs.heyamica.com/)並查看[演示](https://amica.arbius.ai/),這真是太棒了:D ![示範](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/92iv9y2auly6tvenee82.png) 您可以觀看這段簡短的影片,了解它的功能。 https://www.youtube.com/watch?v=hUxAEnFiXH8 Amica 使用 Tauri 建立桌面應用程式。別擔心,我們在此清單的後面部分介紹了金牛座。 他們在 GitHub 上有 400 多個 Star,看起來非常容易使用。 https://github.com/semperai/amica Star Amica ⭐️ --- 19. [Hugging Face Transformers](https://github.com/huggingface/transformers) - 適用於 Pytorch、TensorFlow 和 JAX 的最先進的機器學習。 ---------------------------------------------------------------------------------------------------------------------- ![擁抱變形金剛臉](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/c3acbf1f145jihy4pqar.png) Hugging Face Transformers 可以輕鬆存取最先進的預訓練模型和演算法,用於文字分類、語言生成和問答等任務。該庫建置在 PyTorch 和 TensorFlow 之上,允許用戶以最少的努力將高級 NLP 功能無縫整合到他們的應用程式中。 憑藉大量預訓練模型和支援社區,Hugging Face Transformers 簡化了基於 NLP 的解決方案的開發。 這些模型可用於執行 100 多種語言的文本相關任務,例如文字分類、資訊擷取、問答、摘要、翻譯和文字生成。 它們還可以處理與影像相關的任務,例如影像分類、物件偵測和分割,以及與音訊相關的任務,例如語音辨識和音訊分類。 他們還可以執行各種模式的多任務處理,包括表格問答、光學字元辨識、從掃描文件中提取資訊、視訊分類和視覺問答。 您可以看到大量可用的[模型](https://huggingface.co/models)。 您可以瀏覽[文件](https://huggingface.co/docs/transformers/task_summary)以取得完整的目標並向您展示可以執行的各種任務的範例。 例如,使用管道的一種方法是用於影像分割。 ``` from transformers import pipeline segmenter = pipeline(task="image-segmentation") preds = segmenter( "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/pipeline-cat-chonk.jpeg" ) preds = [{"score": round(pred["score"], 4), "label": pred["label"]} for pred in preds] print(*preds, sep="\n") ``` Transformer 得到了 Jax、PyTorch 和 TensorFlow 這三個最廣泛使用的深度學習庫的支持,並且它們之間可以無縫整合。這種整合可以使用一個庫輕鬆訓練模型,然後加載它們以使用另一個庫進行推理。 它們在 GitHub 上擁有大約 120k+ 星,並被 142k+ 大量開發人員使用。試試看! https://github.com/huggingface/transformers 明星抱臉變形金剛 ⭐️ --- 20. [LLAMA](https://github.com/facebookresearch/llama) - LLaMA 模型的推理程式碼。 ------------------------------------------------------------------------ ![來電](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/bia2hnh4i79w9ljj1c4l.png) Llama 2 是 Facebook Research 開發的尖端技術,使個人、創作者、研究人員和各種規模的企業能夠使用大型語言模型負責任地實驗、創新和擴展他們的想法。 最新版本包括模型權重以及預訓練和微調 Llama 語言模型的起始程式碼,參數範圍從 7B 到 70B。 開始使用涵蓋以下步驟的[安裝指南](https://github.com/facebookresearch/llama?tab=readme-ov-file#quick-start)。 - 克隆並下載儲存庫。 - 安裝所需的依賴項。 - 從 Meta 網站註冊並下載模型。 - 執行提供的腳本來下載模型。 - 使用提供的命令在本地執行所需的模型。 您可以觀看由 ZeroToMastery 製作的關於什麼是美洲駝的[YouTube 影片](https://www.youtube.com/watch?v=OqZ0CSKzu10)。 您也可以在[Hugging Face](https://huggingface.co/meta-llama)和[Meta 官方頁面](https://llama.meta.com/)上查看型號清單和更多資訊。 Ollama 基於 llama,在 GitHub 上擁有 50k+ star。請參閱文件並使用此模型進行更多研究。 https://github.com/facebookresearch/llama 明星 LLAMA ⭐️ --- 21. [Fonoster](https://github.com/fonoster/fonoster) - Twilio 的開源替代品。 --------------------------------------------------------------------- ![福諾斯特](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/pruup1a8yibepdi40fjk.png) Fonoster Inc. 研究了一種創新的可編程電信堆棧,該堆疊將為企業提供完全基於雲端的實用程序,將電話服務與網路連接起來。 根據您想要實現的目標,有多種開始方法。 開始使用以下 npm 指令。 ``` npm install @fonoster/websdk // CDN is also available ``` 例如,您可以透過以下方式將 Fonoster 與 Google Speech API 結合使用。 (您將需要服務帳戶的金鑰) ``` npm install @fonoster/googleasr @fonoster/googletts ``` 這是您可以配置語音伺服器以使用插件的方法。 ``` const { VoiceServer } = require("@fonoster/voice"); const GoogleTTS = require("@fonoster/googletts"); const GoogleASR = require("@fonoster/googleasr"); const voiceServer = new VoiceServer(); const speechConfig = { keyFilename: "./google.json" }; // Set the server to use the speech APIS voiceServer.use(new GoogleTTS(speechConfig)); voiceServer.use(new GoogleASR(speechConfig)); voiceServer.listen(async(req, res) => { console.log(req); await res.answer(); // To use this verb you MUST have a TTS plugin const speech = await res.gather(); await res.say("You said " + speech); await res.hangup(); }); ``` 您可以閱讀[文件](https://fonoster.com/docs/overview/)。 他們提供了一個足以入門的免費套餐。 他們在 GitHub 上擁有大約 6k+ 顆星,並發布了 250 多個版本。 https://github.com/fonoster/fonoster 明星 Fonoster ⭐️ --- 22. [DIPY](https://github.com/dipy/dipy) - Python 中的 paragon 3D/4D+ 成像庫。 ------------------------------------------------------------------------ ![下降](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/l2y1ztg24l2wc1kq5u0g.png) DIPY 是 Python 中領先的 3D/4D+ 成像庫。它包含用於空間歸一化、訊號處理、機器學習、統計分析和醫學影像視覺化的各種方法。 此外,它還包含計算解剖學的專門方法,包括擴散、灌注和結構成像。 您可以開始使用。 ``` pip install dipy // run this in python console import dipy print(dipy.get_info()) ``` 如果您使用的是 anaconda 或其他系統,您可以閱讀完整的[安裝指南](https://docs.dipy.org/stable/examples_built/quick_start/quick_start.html#sphx-glr-examples-built-quick-start-quick-start-py)。 您可以閱讀[文件](https://docs.dipy.org/stable/)並存取他們的[YouTube 頻道](https://www.youtube.com/c/diffusionimaginginpython)。 你可以看看詳細的[例子](https://docs.dipy.org/stable/examples_built/index.html)。 ![例子](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/3b6x3jotln0chpoycmci.png) 他們的下載量超過 428k,並且在 GitHub 儲存庫上擁有 600 多個 Star。 https://github.com/dipy/dipy 明星 DIPY ⭐️ --- 23. [Elastic Search](https://github.com/elastic/elasticsearch) - 免費開放式、分散式、RESTful 搜尋引擎。 ---------------------------------------------------------------------------------------- ![彈性搜尋](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/ouw3u41qdkfjvt999lnv.png) ![資料擬合](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/tsn875yov9bmklfg9aqc.png) Elasticsearch 是一種分散式、RESTful 搜尋和分析引擎,能夠解決大量使用案例。 作為 Elastic Stack 的核心,它集中儲存您的資料,以實現閃電般的快速搜尋、微調的相關性以及可輕鬆擴展的強大分析。 他們闡述了使用 ElasticSearch 的用例。 ![用例](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/sp4qf45yzulbi4c7dire.png) Elasticsearch 使用標準 RESTful API 和 JSON。我們也使用多種語言(例如 Java、Python、.NET、SQL 和 PHP)來建立和維護客戶端。 該結構如下。 ``` const { Client } = require('@elastic/elasticsearch') const client = new Client({ node: 'http://localhost:9200' }) client .search({ index: 'social-*', body: { query: { match: { message: 'myProduct' } }, aggs: { top_10_states: { terms: { field: 'state', size: 10 } } } } }) .then(({ body }) => { const { hits } = body.hits console.log(hits) }) .catch(console.error) ``` 您可以閱讀<a href="">文件</a>並查看[功能清單](https://www.elastic.co/elasticsearch/features)。 儘管具有有用的功能,Elastic Search 的主要缺點是缺乏免費套餐。但是,您仍然可以利用免費試用版來探索和了解開源專案的架構。 Elastic Search 在 GitHub 上擁有超過 67k+ 的星星和近 1900 名貢獻者,並且處於`v8`版本中,正在不斷發展和改進。 https://github.com/elastic/elasticsearch 明星 Elastic Search ⭐️ --- 24. [Tauri](https://github.com/tauri-apps/tauri) - 使用 Web 前端建立更小、更快且安全的桌面應用程式。 ------------------------------------------------------------------------------ ![困難](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/7z6iilytnkaw5d3uj6zv.png) Tauri 是一個工具包,旨在幫助開發人員利用幾乎任何可用的前端框架為主要桌面平台建立應用程式。其核心是使用 Rust 開發的,而 CLI 利用 Node.js,提供了一種真正的多語言方法來開發和維護卓越的應用程式。 Tauri 應用程式中的使用者介面目前利用 Tao 作為 macOS、Windows、Linux、Android 和 iOS 上的視窗處理庫。 為了渲染您的應用程式,Tauri 使用 WRY,這是一個為系統 Web 視圖提供統一介面的程式庫。它在 macOS 和 iOS 上利用 WKWebView、在 Windows 上利用 WebView2、在 Linux 上利用 WebKitGTK 以及在 Android 上利用 Android System WebView。 您可以使用 Vite、HTML/CSS/JS、Next.js、Svelte 等等。 開始使用以下 npm 指令。 ``` npm create tauri-app@latest ``` 您可以閱讀[文件](https://tauri.app/v1/guides/getting-started/prerequisites)並查看 Tauri 提供的[功能清單](https://tauri.app/v1/guides/features/)。 您甚至可以使用 Tauri 建立自己的 CLI,這有多酷:) 團隊提供了[YouTube 影片](https://www.youtube.com/watch?v=UxTJeEbZX-0&t=2s),讓您了解更多關於 Tauri 的訊息。 他們在 GitHub 上擁有超過 75k 顆星星,並發布了 800 多個版本。 https://github.com/tauri-apps/tauri 金牛座之星 ⭐️ --- 25. [AutoGPT](https://github.com/Significant-Gravitas/AutoGPT) - 比 ChatGPT 更令人興奮。 --------------------------------------------------------------------------------- ![自動gpt](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/3hjamyxzkhy7luwsi9vp.png) AutoGPT 的核心在於其主要專案,即由大型語言模型 (LLM) 驅動的半自治代理,旨在為您執行任何任務。 AutoGPT 計畫由[四個主要部分](https://docs.agpt.co/#agent)組成: - 代理 – 也稱為“AutoGPT” - 基準 – 又稱 agbenchmark - 熔爐 - 前端 了解如何使用 OpenAI 金鑰[設定 AutoGPT](https://docs.agpt.co/autogpt/setup/) 。 您可以觀看[Fireship 發布的有關 AutoGPT 的 YouTube 影片](https://www.youtube.com/watch?v=_rGXIXyNqpk)。 https://www.youtube.com/watch?v=\_rGXIXyNqpk 您也可以觀看 Sentral Media 提供的[AutoGPT 教學](https://www.youtube.com/watch?v=FeIIaJUN-4A)。 您可以閱讀[文件](https://docs.agpt.co/)並查看[專案板](https://github.com/orgs/Significant-Gravitas/projects/1),以了解目前正在開發的內容。 即使您對 AI 不太了解,您也可以嘗試 AutoGPT 以了解如何節省時間並建立很酷的東西。 由於如此出色的用例和自動化功能,他們在 GitHub Repo 上擁有大約 159k+ 的星星。 https://github.com/Significant-Gravitas/AutoGPT 明星 AutoGPT ⭐️ --- 還沒結束。 現在,讓我們探索一些有價值的資源,這些資源將幫助您學習新概念並製作更好的人工智慧應用程式。 我們會保持簡單。不掛! - [人工智慧 (AI) 課程、書籍、視訊講座和論文](https://github.com/owainlewis/awesome-artificial-intelligence) - [機器學習/深度學習/AI + Web3 - 教程](https://github.com/TarrySingh/Artificial-Intelligence-Deep-Learning-Machine-Learning-Tutorials) - [ML 初學者](https://github.com/microsoft/ML-For-Beginners)- 12 週、26 節課程、52 個測驗,適合所有人的經典機器學習。 - [機器學習框架、函式庫和軟體](https://github.com/josephmisiti/awesome-machine-learning) - [如何製作人工智慧:逐步指南 - Revelo](https://www.revelo.com/blog/how-to-make-an-ai) 希望這將幫助您學習更多概念! --- 我希望您在列表中找到有用的東西。 我介紹了一些很棒的開源專案,它們可以將您的 AI 應用程式提升到一個新的水平。 人工智慧正在改變世界,最好與人工智慧保持朋友關係,而不是忽視它。 利用它來提高生產力,並抓住機會開發一些非凡的東西。 如果您想以最佳方式改進您的專案,有些開源專案比其他專案更有用,尤其是 Taipy 和 AutoGPT。 請發表評論,讓我們知道哪個專案最讓您感到驚訝。 祝你有美好的一天!直到下一次。 在 GitHub 上關注我。 https://github.com/Anmol-Baranwal 關注 Taipy 以了解更多此類內容。 https://dev.to/taipy --- 原文出處:https://dev.to/taipy/all-the-tools-i-need-to-build-a-perfect-ai-app-2oeh

我正在建立一個全端應用程式:以下是我將要使用的庫......

您可以使用無數的框架和函式庫來改進您的全端應用程式。 我們將介紹令人興奮的概念,例如應用程式內通知、使用 React 製作影片、從為開發人員提供的電子郵件 API 到在瀏覽器中建立互動式音樂。 那我們就開始吧。 (不要忘記為這些庫加註星標以表示您的支持)。 ![圖片描述](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/qqoipyuoxgb83swyoo4a.gif) https://github.com/CopilotKit/CopilotKit --- 1. [CopilotKit](https://github.com/CopilotKit/CopilotKit) - 在數小時內為您的產品提供 AI Copilot。 ------------------------------------------------------------------------------------ ![副駕駛套件](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/nzuxjfog2ldam3csrl62.png) 您可以使用兩個 React 元件將關鍵 AI 功能整合到 React 應用程式中。它們還提供內建(完全可自訂)Copilot 原生 UX 元件,例如`<CopilotKit />` 、 `<CopilotPopup />` 、 `<CopilotSidebar />` 、 `<CopilotTextarea />` 。 開始使用以下 npm 指令。 ``` npm i @copilotkit/react-core @copilotkit/react-ui @copilotkit/react-textarea ``` 這是整合 CopilotTextArea 的方法。 ``` import { CopilotTextarea } from "@copilotkit/react-textarea"; import { useState } from "react"; export function SomeReactComponent() { const [text, setText] = useState(""); return ( <> <CopilotTextarea className="px-4 py-4" value={text} onValueChange={(value: string) => setText(value)} placeholder="What are your plans for your vacation?" autosuggestionsConfig={{ textareaPurpose: "Travel notes from the user's previous vacations. Likely written in a colloquial style, but adjust as needed.", chatApiConfigs: { suggestionsApiConfig: { forwardedParams: { max_tokens: 20, stop: [".", "?", "!"], }, }, }, }} /> </> ); } ``` 您可以閱讀[文件](https://docs.copilotkit.ai/getting-started/quickstart-textarea)。 基本概念是在幾分鐘內建立可用於基於 LLM 的全端應用程式的 AI 聊天機器人。 https://github.com/CopilotKit/CopilotKit --- 2. [Storybook](https://github.com/storybookjs/storybook) - UI 開發、測試和文件變得簡單。 --------------------------------------------------------------------------- ![故事書](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/78rfum1ydisn51qhb408.png) Storybook 是一個用於獨立建立 UI 元件和頁面的前端工作坊。它有助於 UI 開發、測試和文件編制。 他們在 GitHub 上有超過 57,000 次提交、81,000 多個 star 和 1300 多個版本。 這是您為專案建立簡單元件的方法。 ``` import type { Meta, StoryObj } from '@storybook/react'; import { YourComponent } from './YourComponent'; //👇 This default export determines where your story goes in the story list const meta: Meta<typeof YourComponent> = { component: YourComponent, }; export default meta; type Story = StoryObj<typeof YourComponent>; export const FirstStory: Story = { args: { //👇 The args you need here will depend on your component }, }; ``` 您可以閱讀[文件](https://storybook.js.org/docs/get-started/setup)。 如今,UI 除錯起來很痛苦,因為它們與業務邏輯、互動狀態和應用程式上下文糾纏在一起。 Storybook 提供了一個獨立的 iframe 來渲染元件,而不會受到應用程式業務邏輯和上下文的干擾。這可以幫助您將開發重點放在元件的每個變體上,甚至是難以觸及的邊緣情況。 https://github.com/storybookjs/storybook --- 3. [Appwrite](https://github.com/appwrite/appwrite) - 您的後端減少麻煩。 --------------------------------------------------------------- ![應用程式寫入](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/8x568uz21seyygw6b72z.png) ![帶有 appwrite 的 sdk 列表](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/cp7k8qnamsluto7eifpl.png) Appwrite 的開源平台可讓您將身份驗證、資料庫、函數和儲存體新增至您的產品中,並建立任何規模的任何應用程式、擁有您的資料並使用您喜歡的編碼語言和工具。 他們有很好的貢獻指南,甚至不厭其煩地詳細解釋架構。 開始使用以下 npm 指令。 ``` npm install appwrite ``` 您可以像這樣建立一個登入元件。 ``` "use client"; import { useState } from "react"; import { account, ID } from "./appwrite"; const LoginPage = () => { const [loggedInUser, setLoggedInUser] = useState(null); const [email, setEmail] = useState(""); const [password, setPassword] = useState(""); const [name, setName] = useState(""); const login = async (email, password) => { const session = await account.createEmailSession(email, password); setLoggedInUser(await account.get()); }; const register = async () => { await account.create(ID.unique(), email, password, name); login(email, password); }; const logout = async () => { await account.deleteSession("current"); setLoggedInUser(null); }; if (loggedInUser) { return ( <div> <p>Logged in as {loggedInUser.name}</p> <button type="button" onClick={logout}> Logout </button> </div> ); } return ( <div> <p>Not logged in</p> <form> <input type="email" placeholder="Email" value={email} onChange={(e) => setEmail(e.target.value)} /> <input type="password" placeholder="Password" value={password} onChange={(e) => setPassword(e.target.value)} /> <input type="text" placeholder="Name" value={name} onChange={(e) => setName(e.target.value)} /> <button type="button" onClick={() => login(email, password)}> Login </button> <button type="button" onClick={register}> Register </button> </form> </div> ); }; export default LoginPage; ``` 您可以閱讀[文件](https://appwrite.io/docs)。 Appwrite 可以非常輕鬆地建立具有開箱即用的擴充功能的可擴展後端應用程式。 https://github.com/appwrite/appwrite --- 4. [Wasp](https://github.com/wasp-lang/wasp) - 用於 React、node.js 和 prisma 的類似 Rails 的框架。 --------------------------------------------------------------------------------------- ![黃蜂](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/fi2mwazueoc3ezjx8a9q.png) 使用 React 和 Node.js 開發全端 Web 應用程式的最快方法。這不是一個想法,而是一種建立瘋狂快速全端應用程式的不同方法。 這是將其整合到元件中的方法。 ``` import getRecipes from "@wasp/queries/getRecipes"; import { useQuery } from "@wasp/queries"; import type { User } from "@wasp/entities"; export function HomePage({ user }: { user: User }) { // Due to full-stack type safety, `recipes` will be of type `Recipe[]` here. const { data: recipes, isLoading } = useQuery(getRecipes); // Calling our query here! if (isLoading) { return <div>Loading...</div>; } return ( <div> <h1>Recipes</h1> <ul> {recipes ? recipes.map((recipe) => ( <li key={recipe.id}> <div>{recipe.title}</div> <div>{recipe.description}</div> </li> )) : 'No recipes defined yet!'} </ul> </div> ); } ``` 您可以閱讀[文件](https://wasp-lang.dev/docs)。 https://github.com/wasp-lang/wasp --- 5. [Novu](https://github.com/novuhq/novu) - 將應用程式內通知新增至您的應用程式! -------------------------------------------------------------- ![再次](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/716b7biilet4auudjlcu.png) Novu 提供開源通知基礎架構和功能齊全的嵌入式通知中心。 這就是如何使用`React`建立 novu 元件以用於應用程式內通知。 ``` import { NovuProvider, PopoverNotificationCenter, NotificationBell, } from "@novu/notification-center"; function App() { return ( <> <NovuProvider subscriberId={process.env.REACT_APP_SUB_ID} applicationIdentifier={process.env.REACT_APP_APP_ID} > <PopoverNotificationCenter> {({ unseenCount }) => <NotificationBell unseenCount={unseenCount} />} </PopoverNotificationCenter> </NovuProvider> </> ); } export default App; ``` 您可以閱讀[文件](https://docs.novu.co/getting-started/introduction)。 https://github.com/novuhq/novu --- 6. [Remotion](https://github.com/remotion-dev/remotion) - 使用 React 以程式設計方式製作影片。 ------------------------------------------------------------------------------- ![遠端](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/wmnrxhsc7b9mm5oagflm.png) 使用 React 建立真正的 MP4 影片,使用伺服器端渲染和參數化擴展影片製作。 開始使用以下 npm 指令。 ``` npm init video ``` 它為您提供了一個幀號和一個空白畫布,您可以在其中使用 React 渲染任何您想要的內容。 這是一個範例 React 元件,它將當前幀渲染為文字。 ``` import { AbsoluteFill, useCurrentFrame } from "remotion";   export const MyComposition = () => { const frame = useCurrentFrame();   return ( <AbsoluteFill style={{ justifyContent: "center", alignItems: "center", fontSize: 100, backgroundColor: "white", }} > The current frame is {frame}. </AbsoluteFill> ); }; ``` 您可以閱讀[文件](https://www.remotion.dev/docs/)。 過去兩年,remotion 團隊因製作 GitHub Wrapped 而聞名。 https://github.com/remotion-dev/remotion --- [7.NocoDB](https://github.com/nocodb/nocodb) - Airtable 的替代品。 ------------------------------------------------------------- ![諾科資料庫](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/iw3tchfgyzehye5c39xq.png) Airtable 的免費開源替代品是 NocoDB。它可以使用任何 MySQL、PostgreSQL、SQL Server、SQLite 或 MariaDB 資料庫製作智慧型電子表格。 其主要目標是讓強大的計算工具得到更廣泛的使用。 開始使用以下 npx 指令。 ``` npx create-nocodb-app ``` 您可以閱讀[文件](https://docs.nocodb.com/)。 NocoDB 的建立是為了為世界各地的數位企業提供強大的開源和無程式碼資料庫介面。 您可以非常快速地將airtable資料匯入NocoDB。 https://github.com/nocodb/nocodb --- 8.[新穎](https://github.com/steven-tey/novel)- 所見即所得編輯器,具有人工智慧自動完成功能。 ------------------------------------------------------------------- ![小說](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/uo34vd9twpxcpbpzgchi.png) 它使用`Next.js` 、 `Vercel AI SDK` 、 `Tiptap`作為文字編輯器。 開始使用以下 npm 指令。 ``` npm i novel ``` 您可以這樣使用它。有多種選項可用於改進您的應用程式。 ``` import { Editor } from "novel"; export default function App() { return <Editor />; } ``` https://github.com/steven-tey/novel --- 9. [Blitz](https://github.com/blitz-js/blitz) - 缺少 NextJS 的全端工具包。 ----------------------------------------------------------------- ![閃電戰](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/vz6ineg1o7xyv7pwbuqn.png) Blitz 繼承了 Next.js 的不足,為全球應用程式的交付和擴展提供了經過實戰考驗的函式庫和約定。 開始使用以下 npm 指令。 ``` npm install -g blitz ``` 這就是您如何使用 Blitz 建立新頁面。 ``` const NewProjectPage: BlitzPage = () => { const router = useRouter() const [createProjectMutation] = useMutation(createProject) return ( <div> <h1>Create New Project</h1> <ProjectForm submitText="Create Project" schema={CreateProject} onSubmit={async (values) => { // This is equivalent to calling the server function directly const project = await createProjectMutation(values) // Notice the 'Routes' object Blitz provides for routing router.push(Routes.ProjectsPage({ projectId: project.id })) }} /> </div> ); }; NewProjectPage.authenticate = true NewProjectPage.getLayout = (page) => <Layout>{page}</Layout> export default NewProjectPage ``` 您可以閱讀[文件](https://blitzjs.com/docs/get-started)。 它使建築物改善了數倍。 ![閃電戰](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/cc4mb5wdksjv1ybx71co.png) https://github.com/blitz-js/blitz --- 10. [Supabase](https://github.com/supabase/supabase) - 開源 Firebase 替代品。 ----------------------------------------------------------------------- ![蘇帕貝斯](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/ksfygjhrzhmsg9cnvobs.png) 我們大多數人都已經預料到 SUPABASE 會出現在這裡,因為它實在是太棒了。 開始使用以下 npm 指令 (Next.js)。 ``` npx create-next-app -e with-supabase ``` 這是使用 supabase 建立用戶的方法。 ``` import { createClient } from '@supabase/supabase-js' // Initialize const supabaseUrl = 'https://chat-room.supabase.co' const supabaseKey = 'public-anon-key' const supabase = createClient(supabaseUrl, supabaseKey) // Create a new user const { user, error } = await supabase.auth.signUp({ email: '[email protected]', password: 'example-password', }) ``` 您可以閱讀[文件](https://supabase.com/docs)。 您可以使用身份驗證、即時、邊緣功能、儲存等功能建立一個速度極快的應用程式。 Supabase 涵蓋了這一切! 他們還提供了一些入門套件,例如 AI 聊天機器人和 Stripe 訂閱。 https://github.com/supabase/supabase --- [11.Refine](https://github.com/refinedev/refine) - 企業開源重組工具。 ------------------------------------------------------------ ![精煉](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/qx0kd6t2jzdtf90k5ke3.png) 建立具有無與倫比的靈活性的管理面板、儀表板和 B2B 應用程式 您可以在一分鐘內使用單一 CLI 命令進行設定。 它具有適用於 15 多個後端服務的連接器,包括 Hasura、Appwrite 等。 開始使用以下 npm 指令。 ``` npm create refine-app@latest ``` 這就是使用 Refine 新增登入資訊的簡單方法。 ``` import { useLogin } from "@refinedev/core"; const { login } = useLogin(); ``` 您可以閱讀[文件](https://refine.dev/docs/)。 https://github.com/refinedev/refine --- 12. [Zenstack](https://github.com/zenstackhq/zenstack) - 資料庫到 API 和 UI 只需幾分鐘。 ----------------------------------------------------------------------------- ![禪斯塔克](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/3b6n2ea3jeeva6uujoex.png) TypeScript 工具包,透過強大的存取控制層增強 Prisma ORM,並釋放其全端開發的全部功能。 開始使用以下 npx 指令。 ``` npx zenstack@latest init ``` 這是透過伺服器適配器建立 RESTful API 的方法。 ``` // pages/api/model/[...path].ts import { requestHandler } from '@zenstackhq/next'; import { enhance } from '@zenstackhq/runtime'; import { getSessionUser } from '@lib/auth'; import { prisma } from '@lib/db'; // Mount Prisma-style APIs: "/api/model/post/findMany", "/api/model/post/create", etc. // Can be configured to provide standard RESTful APIs (using JSON:API) instead. export default requestHandler({ getPrisma: (req, res) => enhance(prisma, { user: getSessionUser(req, res) }), }); ``` 您可以閱讀[文件](https://zenstack.dev/docs/welcome)。 https://github.com/zenstackhq/zenstack --- 13. [Buildship](https://github.com/rowyio/buildship) - 低程式碼視覺化後端建構器。 -------------------------------------------------------------------- ![建造船](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/rzlrynz5xephv4t9layd.png) 對於您正在使用無程式碼應用程式建構器(FlutterFlow、Webflow、Framer、Adalo、Bubble、BravoStudio...)或前端框架(Next.js、React、Vue...)建立的應用程式,您需要一個後端來支援可擴展的 API、安全工作流程、自動化等。BuildShip 為您提供了一種完全視覺化的方式,可以在易於使用的完全託管體驗中可擴展地建立這些後端任務。 這意味著您不需要在雲端平台上爭論或部署東西、執行 DevOps 等。只需立即建置和交付 🚀 https://github.com/rowyio/buildship --- 14. [Taipy](https://github.com/Avaiga/taipy) - 將資料和人工智慧演算法整合到生產就緒的 Web 應用程式中。 ----------------------------------------------------------------------------- ![打字](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/ohv3johuz92lsaux52oq.png) Taipy 是一個開源 Python 庫,用於輕鬆的端到端應用程式開發, 具有假設分析、智慧管道執行、內建調度和部署工具。 開始使用以下命令。 ``` pip install taipy ``` 這是一個典型的Python函數,也是過濾器場景中使用的唯一任務。 ``` def filter_genre(initial_dataset: pd.DataFrame, selected_genre): filtered_dataset = initial_dataset[initial_dataset['genres'].str.contains(selected_genre)] filtered_data = filtered_dataset.nlargest(7, 'Popularity %') return filtered_data ``` 您可以閱讀[文件](https://docs.taipy.io/en/latest/)。 他們還有很多可供您建立的[演示應用程式教學](https://docs.taipy.io/en/latest/knowledge_base/demos/)。 https://github.com/Avaiga/taipy --- 15. [LocalForage](https://github.com/localForage/localForage) - 改進了離線儲存。 ------------------------------------------------------------------------ ![當地飼料](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/4hlrka5pybvmgmo2djel.png) LocalForage 是一個 JavaScript 函式庫,它透過使用非同步資料儲存和簡單的、類似 localStorage 的 API 來改善 Web 應用程式的離線體驗。它允許開發人員儲存多種類型的資料而不僅僅是字串。 開始使用以下 npm 指令。 ``` npm install localforage ``` 只需包含 JS 檔案並開始使用 localForage。 ``` <script src="localforage.js"></script> ``` 您可以閱讀[文件](https://localforage.github.io/localForage/#installation)。 https://github.com/localForage/localForage --- 16. [Zod](https://github.com/colinhacks/zod) - 使用靜態類型推斷的 TypeScript-first 模式驗證。 ------------------------------------------------------------------------------- ![佐德](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/1s6zvmqr0lv93vsrhofs.png) Zod 的目標是透過最大限度地減少重複的類型聲明來對開發人員友好。使用 Zod,您聲明一次驗證器,Zod 將自動推斷靜態 TypeScript 類型。將更簡單的類型組合成複雜的資料結構很容易。 開始使用以下 npm 指令。 ``` npm install zod ``` 這是您在建立字串架構時自訂一些常見錯誤訊息的方法。 ``` const name = z.string({ required_error: "Name is required", invalid_type_error: "Name must be a string", }); ``` 您可以閱讀[文件](https://zod.dev/)。 它適用於 Node.js 和所有現代瀏覽器 https://github.com/colinhacks/zod --- 17.[多普勒](https://github.com/DopplerHQ)- 管理你的秘密。 ----------------------------------------------- ![多普勒](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/gycxnuiiwsvibryrytlc.png) 您可以透過在具有開發、暫存和生產環境的專案中組織機密來消除機密蔓延。 開始使用以下指令 (MacOS)。 ``` $ brew install dopplerhq/cli/doppler $ doppler --version ``` 這是安裝 Doppler CLI[的 GitHub Actions 工作流程](https://github.com/DopplerHQ/cli-action)。 您可以閱讀[文件](https://docs.doppler.com/docs/start)。 ``` name: Example action on: [push] jobs: my-job: runs-on: ubuntu-latest steps: - name: Install CLI uses: dopplerhq/cli-action@v3 - name: Do something with the CLI run: doppler secrets --only-names env: DOPPLER_TOKEN: ${{ secrets.DOPPLER_TOKEN }} ``` https://github.com/DopplerHQ --- 18. [FastAPI](https://github.com/tiangolo/fastapi) - 高效能、易於學習、快速編碼、可用於生產。 ------------------------------------------------------------------------- ![快速API](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/h2awncoia6255ycl95lk.png) FastAPI 是一個現代、快速(高效能)的 Web 框架,用於基於標準 Python 類型提示使用 Python 3.8+ 建立 API。 開始使用以下命令。 ``` $ pip install fastapi ``` 這是您開始使用 FastAPI 的方式。 ``` from typing import Union from fastapi import FastAPI app = FastAPI() @app.get("/") def read_root(): return {"Hello": "World"} @app.get("/items/{item_id}") def read_item(item_id: int, q: Union[str, None] = None): return {"item_id": item_id, "q": q} ``` 您的編輯器將自動完成屬性並了解它們的類型,這是使用 FastAPI 的最佳功能之一。 您可以閱讀[文件](https://fastapi.tiangolo.com/)。 https://github.com/tiangolo/fastapi --- 19. [Flowise](https://github.com/FlowiseAI/Flowise) - 拖放 UI 來建立您的客製化 LLM 流程。 ---------------------------------------------------------------------------- ![流動](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/ct732wv07pvwx0nmavp5.png) Flowise 是一款開源 UI 視覺化工具,用於建立客製化的 LLM 編排流程和 AI 代理程式。 開始使用以下 npm 指令。 ``` npm install -g flowise npx flowise start OR npx flowise start --FLOWISE_USERNAME=user --FLOWISE_PASSWORD=1234 ``` 這就是整合 API 的方式。 ``` import requests url = "/api/v1/prediction/:id" def query(payload): response = requests.post( url, json = payload ) return response.json() output = query({ question: "hello!" )} ``` 您可以閱讀[文件](https://docs.flowiseai.com/)。 https://github.com/FlowiseAI/Flowise --- 20. [Scrapy](https://github.com/scrapy/scrapy) - Python 的快速進階網頁爬行和抓取框架.. ------------------------------------------------------------------------ ![鬥志旺盛](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/k1b2y1hzdsphw43b6v7b.png) Scrapy 是一個快速的高級網路爬行和網頁抓取框架,用於爬行網站並從頁面中提取結構化資料。它可用於多種用途,從資料探勘到監控和自動化測試。 開始使用以下命令。 ``` pip install scrapy ``` 建造並執行您的網路蜘蛛。 ``` pip install scrapy cat > myspider.py <<EOF import scrapy class BlogSpider(scrapy.Spider): name = 'blogspider' start_urls = ['https://www.zyte.com/blog/'] def parse(self, response): for title in response.css('.oxy-post-title'): yield {'title': title.css('::text').get()} for next_page in response.css('a.next'): yield response.follow(next_page, self.parse) EOF scrapy runspider myspider.py ``` 您可以閱讀[文件](https://scrapy.org/doc/)。 它擁有大約 50k+ 的星星,因此對於網頁抓取來說具有巨大的可信度。 https://github.com/scrapy/scrapy --- 21. [Tone](https://github.com/Tonejs/Tone.js) - 在瀏覽器中製作互動式音樂。 ------------------------------------------------------------- ![音調.js](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/fokxsoblaohgs4tx75g3.png) 開始使用以下 npm 指令。 ``` npm install tone ``` 這是您開始使用 Tone.js 的方法 ``` // To import Tone.js: import * as Tone from 'tone' //create a synth and connect it to the main output (your speakers) const synth = new Tone.Synth().toDestination(); //play a middle 'C' for the duration of an 8th note synth.triggerAttackRelease("C4", "8n"); ``` 您可以閱讀[文件](https://github.com/Tonejs/Tone.js?tab=readme-ov-file#installation)。 https://github.com/Tonejs/Tone.js --- 22. [Spacetime](https://github.com/spencermountain/spacetime) - 輕量級 javascript 時區庫。 ----------------------------------------------------------------------------------- ![時空](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/abfyfuzt4nw4h7b8usab.png) 您可以計算遠端時區的時間;支持夏令時、閏年和半球。按季度、季節、月份、週來定位時間.. 開始使用以下 npm 指令。 ``` npm install spacetime ``` 您可以這樣使用它。 ``` <script src="https://unpkg.com/spacetime"></script> <script> var d = spacetime('March 1 2012', 'America/New_York') //set the time d = d.time('4:20pm') d = d.goto('America/Los_Angeles') d.time() //'1:20pm' </script> ``` https://github.com/spencermountain/spacetime --- 23. [Mermaid](https://github.com/mermaid-js/mermaid) - 從類似 markdown 的文字產生圖表。 ---------------------------------------------------------------------------- ![美人魚](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/ggubn86xv7fznxol6fw7.png) 您可以使用 Markdown with Mermaid 等文字產生流程圖或序列圖等圖表。 這就是建立圖表的方法。 ``` sequenceDiagram Alice->>John: Hello John, how are you? loop Healthcheck John->>John: Fight against hypochondria end Note right of John: Rational thoughts! John-->>Alice: Great! John->>Bob: How about you? Bob-->>John: Jolly good! ``` 它將做出如下圖。 ![圖表](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/bbuo2ey5q2x3sjwywizg.png) 您可以閱讀[VS Code](https://docs.mermaidchart.com/plugins/visual-studio-code)的[文件](https://mermaid.js.org/intro/getting-started.html)和外掛程式。 請參閱[即時編輯器](https://mermaid.live/edit#pako:eNpVkE1PwzAMhv9KlvM-2AZj62EIxJd24ADXXLzEbaKlcUkdUDX1v5MONomcnNevXz32UWoyKAvZ4mfCoPHRQRWhVuHeO42T7XZHNhTiFb0nMdRjYelbQETRUbpTwRM1uQ2erbaoDyqI_AbnZfjZVZYFVOBCy8J2DWlLwUQHKmAwKrwRo4gnF5Xid-gd2FEAL9hSyp12pMIpNcee2ArxEhH4LG-3D7TPoAPcnhL_4WVxcgHZkfedqIjMSI5ljbEGZ_LyxwFaSbZYo5JFLg3Eg5Iq9NkHiemjC1oWHBOOZWoM8PlQ_8Un45iiLErwbRY9gcH8PUrumuHKlWs5J2oKpasGPUWfZcvctMVsNrSnlWOb9lNN9ax1xkJk-7VZzVaL1RoWS1zdLuFmuTR6P9-sy8X1vDS3V_MFyL7vfwD_bJ1W)中的範例。 https://github.com/mermaid-js/mermaid --- 24.[公共 API](https://github.com/public-apis/public-apis) - 20 多個類別的 1400 多個 API。 ------------------------------------------------------------------------------- ![公共API](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/sjapk9rwlzdl6bcyqdnl.png) 我們主要使用外部 API 來建立應用程式,在這裡您可以找到所有 API 的清單。網站連結在最後。 它在 GitHub 上擁有大約 279k+ 顆星。 ![公共API](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/rld5i88smezo1naawz7a.png) 從儲存庫取得網站連結非常困難。所以,我把它貼在這裡。 網址 - [Collective-api.vercel.app/](https://collective-api.vercel.app/) https://github.com/public-apis/public-apis --- 25. [Framer Motion](https://github.com/framer/motion) - 像魔法一樣的動畫。 ----------------------------------------------------------------- ![成幀器運動](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/hn4ecqkrhs8f4729bzps.png) 可用的最強大的動畫庫之一。 Framer 使用簡單的聲明性語法意味著您編寫的程式碼更少。更少的程式碼意味著您的程式碼庫更易於閱讀和維護。 您可以建立事件和手勢,並且使用 Framer 的社區很大,這意味著良好的支援。 開始使用以下 npm 指令。 ``` npm install framer-motion ``` 您可以這樣使用它。 ``` import { motion } from "framer-motion" <motion.div whileHover={{ scale: 1.2 }} whileTap={{ scale: 1.1 }} drag="x" dragConstraints={{ left: -100, right: 100 }} /> ``` 您可以閱讀[文件](https://www.framer.com/motion/introduction/)。 https://github.com/framer/motion --- 26.[順便說一句](https://github.com/btw-so/btw)- 在幾分鐘內建立您的個人部落格。 ---------------------------------------------------------- ![順便提一句](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/gnne3lrfpolotmxkdz2m.png) 順便說一句,您可以註冊並使用,而無需安裝任何東西。您也可以使用開源版本自行託管。 ![順便提一句](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/2rli7hpoccqwpvba29b4.png) 使用順便說一句建立的[範例部落格](https://www.siddg.com/about)。 https://github.com/btw-so/btw --- 27. [Formbricks](https://github.com/formbricks/formbricks) - 開源調查平台。 -------------------------------------------------------------------- ![成型磚](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/tp6ggyom33vdifd3m1vt.png) Formbricks 提供免費、開源的測量平台。透過精美的應用程式內、網站、連結和電子郵件調查收集用戶旅程中每個點的回饋。在 Formbricks 之上建置或利用預先建置的資料分析功能。 開始使用以下 npm 指令。 ``` npm install @formbricks/js ``` 這就是您開始使用 formbricks 的方法。 ``` import formbricks from "@formbricks/js"; if (typeof window !== "undefined") { formbricks.init({ environmentId: "claV2as2kKAqF28fJ8", apiHost: "https://app.formbricks.com", }); } ``` 您可以閱讀[文件](https://formbricks.com/docs/getting-started/quickstart-in-app-survey)。 https://github.com/formbricks/formbricks --- 28. [Stripe](https://github.com/stripe) - 支付基礎設施。 ------------------------------------------------- ![條紋](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/79yvcgsi4744cmryh15j.png) 數以百萬計的各種規模的公司在線上和親自使用 Stripe 來接受付款、發送付款、自動化財務流程並最終增加收入。 開始使用以下 npm 指令 (React.js)。 ``` npm install @stripe/react-stripe-js @stripe/stripe-js ``` 這就是使用鉤子的方法。 ``` import React, {useState} from 'react'; import ReactDOM from 'react-dom'; import {loadStripe} from '@stripe/stripe-js'; import { PaymentElement, Elements, useStripe, useElements, } from '@stripe/react-stripe-js'; const CheckoutForm = () => { const stripe = useStripe(); const elements = useElements(); const [errorMessage, setErrorMessage] = useState(null); const handleSubmit = async (event) => { event.preventDefault(); if (elements == null) { return; } // Trigger form validation and wallet collection const {error: submitError} = await elements.submit(); if (submitError) { // Show error to your customer setErrorMessage(submitError.message); return; } // Create the PaymentIntent and obtain clientSecret from your server endpoint const res = await fetch('/create-intent', { method: 'POST', }); const {client_secret: clientSecret} = await res.json(); const {error} = await stripe.confirmPayment({ //`Elements` instance that was used to create the Payment Element elements, clientSecret, confirmParams: { return_url: 'https://example.com/order/123/complete', }, }); if (error) { // This point will only be reached if there is an immediate error when // confirming the payment. Show error to your customer (for example, payment // details incomplete) setErrorMessage(error.message); } else { // Your customer will be redirected to your `return_url`. For some payment // methods like iDEAL, your customer will be redirected to an intermediate // site first to authorize the payment, then redirected to the `return_url`. } }; return ( <form onSubmit={handleSubmit}> <PaymentElement /> <button type="submit" disabled={!stripe || !elements}> Pay </button> {/* Show error message to your customers */} {errorMessage && <div>{errorMessage}</div>} </form> ); }; const stripePromise = loadStripe('pk_test_6pRNASCoBOKtIshFeQd4XMUh'); const options = { mode: 'payment', amount: 1099, currency: 'usd', // Fully customizable with appearance API. appearance: { /*...*/ }, }; const App = () => ( <Elements stripe={stripePromise} options={options}> <CheckoutForm /> </Elements> ); ReactDOM.render(<App />, document.body); ``` 您可以閱讀[文件](https://github.com/stripe/react-stripe-js?tab=readme-ov-file#minimal-example)。 您幾乎可以整合任何東西。它有一個巨大的選項清單。 ![整合](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/67f3pb2i8xolt635rp2p.png) https://github.com/stripe --- 29. [Upscayl](https://github.com/upscayl/upscayl) - 開源 AI 影像升級器。 ---------------------------------------------------------------- ![高級](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/2c1837rev5jb260ro2sd.png) 適用於 Linux、MacOS 和 Windows 的免費開源 AI Image Upscaler 採用 Linux 優先概念建構。 它可能與全端無關,但它對於升級圖像很有用。 ![高級](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/a4qq1wm3wey3vihn9al4.png) 透過最先進的人工智慧,Upscayl 可以幫助您將低解析度影像變成高解析度。清脆又鋒利! https://github.com/upscayl/upscayl --- 30.[重新發送](https://github.com/resend)- 為開發人員提供的電子郵件 API。 ------------------------------------------------------- ![重發](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/x3auhh3hbxjmmzehe5v0.png) 您可以使用 React 建立和傳送電子郵件。 2023 年最受炒作的產品之一。 開始使用以下 npm 指令。 ``` npm install @react-email/components -E ``` 這是將其與 next.js 專案整合的方法。 ``` import { EmailTemplate } from '@/components/email-template'; import { Resend } from 'resend'; const resend = new Resend(process.env.RESEND_API_KEY); export async function POST() { const { data, error } = await resend.emails.send({ from: '[email protected]', to: '[email protected]', subject: 'Hello world', react: EmailTemplate({ firstName: 'John' }), }); if (error) { return Response.json({ error }); } return Response.json(data); } ``` 您可以閱讀[文件](https://resend.com/docs/introduction)。 ![重發](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/rer9ym187e4i9l11afkg.png) 基本概念是一個簡單、優雅的介面,讓您可以在幾分鐘內開始發送電子郵件。它可以透過適用於您最喜歡的程式語言的 SDK 直接融入您的程式碼中。 https://github.com/resend --- 哇!如此長的專案清單。 我知道您有更多想法,分享它們,讓我們一起建造:D 如今建立全端應用程式並不難,但每個應用程式都可以透過有效地使用優秀的開源專案來解決任何問題來增加這一獨特因素。 例如,您可以建立一些提供通知或建立 UI 流來抓取資料的東西。 我希望其中一些內容對您的開發之旅有用。他們擁有一流的開發人員經驗;你可以依賴他們。 由於您將要建造東西,因此您可以在這裡找到一些[瘋狂的想法](https://github.com/florinpop17/app-ideas)。 祝你有美好的一天!直到下一次。 --- 原文出處:https://dev.to/copilotkit/im-building-a-full-stack-app-here-are-the-libraries-im-going-to-use-51nk

我如何建立 NotesGPT – 一個全端人工智慧語音筆記應用程式

上週,我推出了[notesGPT](https://usenotesgpt.com/) ,這是一款免費開源語音記事應用程式,上週迄今為止已有[35,000 名訪客](https://twitter.com/nutlope/status/1760053364791050285)、7,000 名用戶和超過 1,000 名 GitHub star。它允許您錄製語音筆記,使用[Whisper](https://github.com/openai/whisper)進行轉錄,並透過[Together](https://together.ai/)使用 Mixtral 來提取操作項並將其顯示在操作項視圖中。它也是[完全開源的](https://github.com/nutlope/notesgpt),配備了身份驗證、儲存、向量搜尋、操作項,並且在行動裝置上完全響應,易於使用。 我將向您詳細介紹我是如何建造它的。 架構和技術堆疊 ------- 這是架構的快速圖表。我們將更深入地討論每個部分,並同時展示程式碼範例。 ![架構圖](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/sjl3i4bu23fn0pabldsw.png) 這是我使用的整體技術堆疊: - 資料庫和雲端函數的[convex](https://convex.dev/) - Next.js [App Router](https://nextjs.org/docs/app)框架 - [複製](https://replicate.com/)Whisper 轉錄 - LLM 與[JSON 模式](https://docs.together.ai/docs/json-mode)的[Mixtral](https://mistral.ai/news/mixtral-of-experts/) - [Together.ai](http://Together.ai)用於推理和嵌入 - 用於儲存語音註釋的[凸檔存儲](https://docs.convex.dev/file-storage) - [凸向量搜尋](https://docs.convex.dev/vector-search)用於向量搜尋 - 負責使用者身份驗證的[職員](https://clerk.dev/) - [Tailwind CSS](https://tailwindcss.com/)樣式 登陸頁面 ---- 該應用程式的第一部分是您導航到notesGPT 時看到的登入頁面。 ![NotesGPT 的登陸頁面](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/0hfscmudh4l33oab3azw.png) 用戶首先看到的是這個登陸頁面,它與應用程式的其餘部分一起使用 Next.js 和 Tailwind CSS 進行樣式建立。我喜歡使用 Next.js,因為它可以輕鬆啟動 Web 應用程式並編寫 React 程式碼。 Tailwind CSS 也很棒,因為它允許您在網頁上快速迭代,同時與 JSX 保持在同一檔案中。 與 Clerk 和 Convex 進行身份驗證 ----------------------- 當使用者點擊主頁上的任一按鈕時,他們將被導向到登入畫面。這是由 Clerk 提供支援的,這是一個與 Convex 很好整合的簡單身份驗證解決方案,我們將在整個後端使用它,包括雲端功能、資料庫、儲存和向量搜尋。 ![認證頁面](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/02khgd6f2jfew1w7dufn.png) Clerk 和 Convex 都很容易設定。您只需在這兩個服務上建立一個帳戶,安裝它們的 npm 庫,執行`npx convex dev`來設定您的凸資料夾,然後建立一個如下所示的`ConvexProvider.ts`檔案來包裝您的應用程式。 ``` 'use client'; import { ReactNode } from 'react'; import { ConvexReactClient } from 'convex/react'; import { ConvexProviderWithClerk } from 'convex/react-clerk'; import { ClerkProvider, useAuth } from '@clerk/nextjs'; const convex = new ConvexReactClient(process.env.NEXT_PUBLIC_CONVEX_URL!); export default function ConvexClientProvider({ children, }: { children: ReactNode; }) { return ( <ClerkProvider publishableKey={process.env.NEXT_PUBLIC_CLERK_PUBLISHABLE_KEY!} > <ConvexProviderWithClerk client={convex} useAuth={useAuth}> {children} </ConvexProviderWithClerk> </ClerkProvider> ); } ``` 請查看[Convex Quickstart](https://docs.convex.dev/quickstart/nextjs)和[Convex Clerk](https://docs.convex.dev/auth/clerk) auth 部分以了解更多詳細資訊。 設定我們的架構 ------- 您可以在有或沒有模式的情況下使用 Convex。就我而言,我知道資料的結構並想要定義它,所以我在下面這樣做了。這也為您提供了一個非常好的類型安全 API,可以在與資料庫互動時使用。我們定義兩個表格-一個用於儲存所有語音註解資訊的`notes`表和用於提取的操作專案的`actionItems`表。我們還將定義索引,以便能夠透過`userId`和`noteId`快速查詢資料。 ``` import { defineSchema, defineTable } from 'convex/server'; import { v } from 'convex/values'; export default defineSchema({ notes: defineTable({ userId: v.string(), audioFileId: v.string(), audioFileUrl: v.string(), title: v.optional(v.string()), transcription: v.optional(v.string()), summary: v.optional(v.string()), embedding: v.optional(v.array(v.float64())), generatingTranscript: v.boolean(), generatingTitle: v.boolean(), generatingActionItems: v.boolean(), }) .index('by_userId', ['userId']) .vectorIndex('by_embedding', { vectorField: 'embedding', dimensions: 768, filterFields: ['userId'], }), actionItems: defineTable({ noteId: v.id('notes'), userId: v.string(), task: v.string(), }) .index('by_noteId', ['noteId']) .index('by_userId', ['userId']), }); ``` 儀表板 --- 現在我們已經有了後端和身份驗證設定以及模式,我們可以看看如何獲取資料。登入應用程式後,用戶可以查看其儀表板,其中列出了他們錄製的所有語音筆記。 ![儀表板](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/6u9f1b60kgfp4txbszur.png) 為此,我們首先在凸資料夾中定義一個查詢,該查詢使用 auth 接收`userId` ,驗證其有效,並傳回與使用者的`userId`相符的所有註解。 ``` export const getNotes = queryWithUser({ args: {}, handler: async (ctx, args) => { const userId = ctx.userId; if (userId === undefined) { return null; } const notes = await ctx.db .query('notes') .withIndex('by_userId', (q) => q.eq('userId', userId)) .collect(); const results = Promise.all( notes.map(async (note) => { const count = ( await ctx.db .query('actionItems') .withIndex('by_noteId', (q) => q.eq('noteId', note._id)) .collect() ).length; return { count, ...note, }; }), ); return results; }, }); ``` 之後,我們可以透過凸提供的函數使用使用者的驗證令牌來呼叫此`getNotes`查詢,以在儀表板中顯示所有使用者的註解。我們使用伺服器端渲染在伺服器上取得此資料,然後將其傳遞到`<DashboardHomePage />`客戶端元件。這也確保了客戶端上的資料也保持最新。 ``` import { api } from '@/convex/_generated/api'; import { preloadQuery } from 'convex/nextjs'; import DashboardHomePage from './dashboard'; import { getAuthToken } from '../auth'; const ServerDashboardHomePage = async () => { const token = await getAuthToken(); const preloadedNotes = await preloadQuery(api.notes.getNotes, {}, { token }); return <DashboardHomePage preloadedNotes={preloadedNotes} />; }; export default ServerDashboardHomePage; ``` 錄製語音筆記 ------ 最初,使用者的儀表板上不會有任何語音註釋,因此他們可以點擊「錄製新語音註釋」按鈕來錄製。他們將看到以下螢幕,允許他們進行錄製。 ![錄製語音筆記頁面](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/e3lm22akd3zanf3ar0za.png) 這將使用本機瀏覽器 API 錄製語音筆記,將檔案保存在 Convex 檔案儲存中,然後透過 Replicate 將其傳送至 Whisper 進行轉錄。我們要做的第一件事是在凸資料夾中定義一個`createNote`突變,它將接收此記錄,在凸資料庫中保存一些訊息,然後呼叫耳語操作。 ``` export const createNote = mutationWithUser({ args: { storageId: v.id('_storage'), }, handler: async (ctx, { storageId }) => { const userId = ctx.userId; let fileUrl = (await ctx.storage.getUrl(storageId)) as string; const noteId = await ctx.db.insert('notes', { userId, audioFileId: storageId, audioFileUrl: fileUrl, generatingTranscript: true, generatingTitle: true, generatingActionItems: true, }); await ctx.scheduler.runAfter(0, internal.whisper.chat, { fileUrl, id: noteId, }); return noteId; }, }); ``` 耳語動作如下圖所示。它使用 Replicate 作為 Whisper 的託管提供者。 ``` export const chat = internalAction({ args: { fileUrl: v.string(), id: v.id('notes'), }, handler: async (ctx, args) => { const replicateOutput = (await replicate.run( 'openai/whisper:4d50797290df275329f202e48c76360b3f22b08d28c196cbc54600319435f8d2', { input: { audio: args.fileUrl, model: 'large-v3', translate: false, temperature: 0, transcription: 'plain text', suppress_tokens: '-1', logprob_threshold: -1, no_speech_threshold: 0.6, condition_on_previous_text: true, compression_ratio_threshold: 2.4, temperature_increment_on_fallback: 0.2, }, }, )) as whisperOutput; const transcript = replicateOutput.transcription || 'error'; await ctx.runMutation(internal.whisper.saveTranscript, { id: args.id, transcript, }); }, }); ``` 此外,所有這些檔案都可以在 Convex 儀表板的「檔案」下看到。 ![凸形儀表板](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/mz51ysreunwsk52tqjr9.png) 生成行動專案 ------ 使用者完成語音記錄並透過耳語進行轉錄後,輸出將傳遞到 Together AI 中。我們同時顯示此加載畫面。 ![頁面載入](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/1rcr80meap2xql9nrzlf.png) 我們首先定義一個我們希望輸出所在的模式。然後,我們將此模式傳遞到 Together.ai 上託管的 Mixtral 模型中,並提示辨識語音註釋的摘要、文字記錄,並根據成績單。然後我們將所有這些資訊保存到 Convex 資料庫中。為此,我們在凸資料夾中建立一個凸動作。 ``` // convex/together.ts const NoteSchema = z.object({ title: z .string() .describe('Short descriptive title of what the voice message is about'), summary: z .string() .describe( 'A short summary in the first person point of view of the person recording the voice message', ) .max(500), actionItems: z .array(z.string()) .describe( 'A list of action items from the voice note, short and to the point. Make sure all action item lists are fully resolved if they are nested', ), }); export const chat = internalAction({ args: { id: v.id('notes'), transcript: v.string(), }, handler: async (ctx, args) => { const { transcript } = args; const extract = await client.chat.completions.create({ messages: [ { role: 'system', content: 'The following is a transcript of a voice message. Extract a title, summary, and action items from it and answer in JSON in this format: {title: string, summary: string, actionItems: [string, string, ...]}', }, { role: 'user', content: transcript }, ], model: 'mistralai/Mixtral-8x7B-Instruct-v0.1', response_model: { schema: NoteSchema, name: 'SummarizeNotes' }, max_tokens: 1000, temperature: 0.6, max_retries: 3, }); const { title, summary, actionItems } = extract; await ctx.runMutation(internal.together.saveSummary, { id: args.id, summary, actionItems, title, }); }); ``` 當 Together.ai 做出回應時,我們會看到最終畫面,使用者可以在左側的記錄和摘要之間切換,並查看並勾選右側的操作專案。 ![完整語音筆記頁面](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/cnd6j68hgusa0aj2buhv.png) 向量搜尋 ---- 該應用程式的最後一部分是向量搜尋。我們使用 Together.ai 嵌入來嵌入文字記錄,並使人們可以根據文字記錄的語義在儀表板中進行搜尋。 我們透過在凸資料夾中建立一個`similarNotes`操作來實現此目的,該操作接受使用者的搜尋查詢,為其產生嵌入,並找到要在頁面上顯示的最相似的註釋。 ``` export const similarNotes = actionWithUser({ args: { searchQuery: v.string(), }, handler: async (ctx, args): Promise<SearchResult[]> => { // 1. Create the embedding const getEmbedding = await togetherai.embeddings.create({ input: [args.searchQuery.replace('/n', ' ')], model: 'togethercomputer/m2-bert-80M-32k-retrieval', }); const embedding = getEmbedding.data[0].embedding; // 2. Then search for similar notes const results = await ctx.vectorSearch('notes', 'by_embedding', { vector: embedding, limit: 16, filter: (q) => q.eq('userId', ctx.userId), // Only search my notes. }); return results.map((r) => ({ id: r._id, score: r._score, })); }, }); ``` 結論 -- 就像這樣,我們建立了一個可投入生產的全端人工智慧應用程式,配備身份驗證、資料庫、儲存和 API。請隨意查看[notesGPT,](https://usenotesgpt.com/)以從您的筆記或[GitHub 儲存庫](https://github.com/nutlope/notesGPT)產生操作專案以供參考。如果您有任何疑問,[請私訊我](twitter.com/nutlope),我將非常樂意回答! --- 原文出處:https://dev.to/nutlope/how-i-built-notesgpt-a-full-stack-ai-voice-note-app-265o

為初學者到專家提供的 101 個 Bash 指令和提示

> **2019 年 9 月 25 日更新:**感謝[ラナ・kuaru](https://twitter.com/rana_kualu)的辛勤工作,本文現已提供日文版。請點擊下面的連結查看他們的工作。如果您知道本文被翻譯成其他語言,請告訴我,我會將其發佈在這裡。 [🇯🇵 閱讀日語](https://qiita.com/rana_kualu/items/7b62898d373901466f5c) > **2019 年 7 月 8 日更新:**我最近發現大約兩年前發佈在法語留言板上的[這篇非常相似的文章](https://bookmarks.ecyseo.net/?EAWvDw)。如果您有興趣學習一些 shell 命令——並且您*會說 français* ,那麼它是對我下面的文章的一個很好的補充。 直到大約一年前,我幾乎只在 macOS 和 Ubuntu 作業系統中工作。在這兩個作業系統上, `bash`是我的預設 shell。在過去的六、七年裡,我對`bash`工作原理有了大致的了解,並想為那些剛入門的人概述一些更常見/有用的命令。如果您認為您了解有關`bash`所有訊息,請無論如何看看下面的內容 - 我已經提供了一些提示和您可能忘記的標誌的提醒,這可以讓您的工作更輕鬆一些。 下面的命令或多或少以敘述風格排列,因此如果您剛開始使用`bash` ,您可以從頭到尾完成操作。事情到最後通常會變得不那麼常見並且變得更加困難。 <a name="toc"></a> 目錄 -- - [基礎](#the-basics) ``` - [First Commands, Navigating the Filesystem](#first-commands) ``` ``` - [`pwd / ls / cd`](#pwd-ls-cd) ``` ``` - [`; / && / &`](#semicolon-andand-and) ``` ``` - [Getting Help](#getting-help) ``` ``` - [`-h`](#minus-h) ``` ``` - [`man`](#man) ``` ``` - [Viewing and Editing Files](#viewing-and-editing-files) ``` ``` - [`head / tail / cat / less`](#head-tail-cat-less) ``` ``` - [`nano / nedit`](#nano-nedit) ``` ``` - [Creating and Deleting Files and Directories](#creating-and-deleting-files) ``` ``` - [`touch`](#touch) ``` ``` - [`mkdir / rm / rmdir`](#mkdir-rm-rmdir) ``` ``` - [Moving and Copying Files, Making Links, Command History](#moving-and-copying-files) ``` ``` - [`mv / cp / ln`](#mv-cp-ln) ``` ``` - [Command History](#command-history) ``` ``` - [Directory Trees, Disk Usage, and Processes](#directory-trees-disk-usage-processes) ``` ``` - [`mkdir –p / tree`](#mkdir--p-tree) ``` ``` - [`df / du / ps`](#df-du-ps) ``` ``` - [Miscellaneous](#basic-misc) ``` ``` - [`passwd / logout / exit`](#passwd-logout-exit) ``` ``` - [`clear / *`](#clear-glob) ``` - [中間的](#intermediate) ``` - [Disk, Memory, and Processor Usage](#disk-memory-processor) ``` ``` - [`ncdu`](#ncdu) ``` ``` - [`top / htop`](#top-htop) ``` ``` - [REPLs and Software Versions](#REPLs-software-versions) ``` ``` - [REPLs](#REPLs) ``` ``` - [`-version / --version / -v`](#version) ``` ``` - [Environment Variables and Aliases](#env-vars-aliases) ``` ``` - [Environment Variables](#env-vars) ``` ``` - [Aliases](#aliases) ``` ``` - [Basic `bash` Scripting](#basic-bash-scripting) ``` ``` - [`bash` Scripts](#bash-scripts) ``` ``` - [Custom Prompt and `ls`](#custom-prompt-ls) ``` ``` - [Config Files](#config-files) ``` ``` - [Config Files / `.bashrc`](#config-bashrc) ``` ``` - [Types of Shells](#types-of-shells) ``` ``` - [Finding Things](#finding-things) ``` ``` - [`whereis / which / whatis`](#whereis-which-whatis) ``` ``` - [`locate / find`](#locate-find) ``` ``` - [Downloading Things](#downloading-things) ``` ``` - [`ping / wget / curl`](#ping-wget-curl) ``` ``` - [`apt / gunzip / tar / gzip`](#apt-gunzip-tar-gzip) ``` ``` - [Redirecting Input and Output](#redirecting-io) ``` ``` - [`| / > / < / echo / printf`](#pipe-gt-lt-echo-printf) ``` ``` - [`0 / 1 / 2 / tee`](#std-tee) ``` - [先進的](#advanced) ``` - [Superuser](#superuser) ``` ``` - [`sudo / su`](#sudo-su) ``` ``` - [`!!`](#click-click) ``` ``` - [File Permissions](#file-permissions) ``` ``` - [File Permissions](#file-permissions-sub) ``` ``` - [`chmod / chown`](#chmod-chown) ``` ``` - [User and Group Management](#users-groups) ``` ``` - [Users](#users) ``` ``` - [Groups](#groups) ``` ``` - [Text Processing](#text-processing) ``` ``` - [`uniq / sort / diff / cmp`](#uniq-sort-diff-cmp) ``` ``` - [`cut / sed`](#cut-sed) ``` ``` - [Pattern Matching](#pattern-matching) ``` ``` - [`grep`](#grep) ``` ``` - [`awk`](#awk) ``` ``` - [Copying Files Over `ssh`](#ssh) ``` ``` - [`ssh / scp`](#ssh-scp) ``` ``` - [`rsync`](#rsync) ``` ``` - [Long-Running Processes](#long-running-processes) ``` ``` - [`yes / nohup / ps / kill`](#yes-nohup-ps-kill) ``` ``` - [`cron / crontab / >>`](#cron) ``` ``` - [Miscellaneous](#advanced-misc) ``` ``` - [`pushd / popd`](#pushd-popd) ``` ``` - [`xdg-open`](#xdg-open) ``` ``` - [`xargs`](#xargs) ``` - [獎勵:有趣但大多無用的東西](#bonus) ``` - [`w / write / wall / lynx`](#w-write-wall-lynx) ``` ``` - [`nautilus / date / cal / bc`](#nautilus-date-cal-bc) ``` --- <a name="the-basics"></a> 基礎 == <a name="first-commands"></a> 第一個指令,瀏覽檔案系統 ------------ 現代檔案系統具有目錄(資料夾)樹,其中目錄要么是*根目錄*(沒有父目錄),要么是*子目錄*(包含在單一其他目錄中,我們稱之為“父目錄”)。向後遍歷檔案樹(從子目錄到父目錄)將始終到達根目錄。有些檔案系統有多個根目錄(如 Windows 的磁碟機: `C:\` 、 `A:\`等),但 Unix 和類別 Unix 系統只有一個名為`\`的根目錄。 <a name="pwd-ls-cd"></a> ### `pwd / ls / cd` [\[ 返回目錄 \]](#toc) 在檔案系統中工作時,使用者始終*在*某個目錄中工作,我們稱之為當前目錄或*工作目錄*。使用`pwd`列印使用者的工作目錄: ``` [ andrew@pc01 ~ ]$ pwd /home/andrew ``` 使用`ls`列出該目錄的內容(檔案和/或子目錄等): ``` [ andrew@pc01 ~ ]$ ls Git TEST jdoc test test.file ``` > **獎金:** > > 使用`ls -a`顯示隱藏(“點”)文件 > > 使用`ls -l`顯示文件詳細訊息 > > 組合多個標誌,如`ls -l -a` > > 有時您可以連結諸如`ls -la`之類的標誌,而不是`ls -l -a` 使用`cd`更改到不同的目錄(更改目錄): ``` [ andrew@pc01 ~ ]$ cd TEST/ [ andrew@pc01 TEST ]$ pwd /home/andrew/TEST [ andrew@pc01 TEST ]$ cd A [ andrew@pc01 A ]$ pwd /home/andrew/TEST/A ``` `cd ..`是「 `cd`到父目錄」的簡寫: ``` [ andrew@pc01 A ]$ cd .. [ andrew@pc01 TEST ]$ pwd /home/andrew/TEST ``` `cd ~`或只是`cd`是「 `cd`到我的主目錄」的簡寫(通常`/home/username`或類似的東西): ``` [ andrew@pc01 TEST ]$ cd [ andrew@pc01 ~ ]$ pwd /home/andrew ``` > **獎金:** > > `cd ~user`表示「 `cd`到`user`的主目錄 > > 您可以使用`cd ../..`等跳轉多個目錄等級。 > > 使用`cd -`返回到最近的目錄 > > `.`是「此目錄」的簡寫,因此`cd .`不會做太多事情 <a name="semicolon-andand-and"></a> ### `; / && / &` [\[ 返回目錄 \]](#toc) 我們在命令列中輸入的內容稱為*命令*,它們總是執行儲存在電腦上某處的一些機器碼。有時這個機器碼是一個內建的Linux命令,有時它是一個應用程式,有時它是你自己寫的一些程式碼。有時,我們會想依序執行一個指令。為此,我們可以使用`;` (分號): ``` [ andrew@pc01 ~ ]$ ls; pwd Git TEST jdoc test test.file /home/andrew ``` 上面的分號表示我首先 ( `ls` ) 列出工作目錄的內容,然後 ( `pwd` ) 列印其位置。連結命令的另一個有用工具是`&&` 。使用`&&`時,如果左側命令失敗,則右側命令將不會執行。 `;`和`&&`都可以在同一行中多次使用: ``` # whoops! I made a typo here! [ andrew@pc01 ~ ]$ cd /Giit/Parser && pwd && ls && cd -bash: cd: /Giit/Parser: No such file or directory # the first command passes now, so the following commands are run [ andrew@pc01 ~ ]$ cd Git/Parser/ && pwd && ls && cd /home/andrew/Git/Parser README.md doc.sh pom.xml resource run.sh shell.sh source src target ``` ....但是與`;` ,即使第一個命令失敗,第二個命令也會執行: ``` # pwd and ls still run, even though the cd command failed [ andrew@pc01 ~ ]$ cd /Giit/Parser ; pwd ; ls -bash: cd: /Giit/Parser: No such file or directory /home/andrew Git TEST jdoc test test.file ``` `&`看起來與`&&`類似,但實際上實現了完全不同的功能。通常,當您執行長時間執行的命令時,命令列將等待該命令完成,然後才允許您輸入另一個命令。在命令後面加上`&`可以防止這種情況發生,並允許您在舊命令仍在執行時執行新命令: ``` [ andrew@pc01 ~ ]$ cd Git/Parser && mvn package & cd [1] 9263 ``` > **額外的好處:**當我們在命令後使用`&`來「隱藏」它時,我們說該作業(或「進程」;這些術語或多或少可以互換)是「後台的」。若要查看目前正在執行的背景作業,請使用`jobs`指令: > ````bash \[ andrew@pc01 ~ \]$ 職位 \[1\]+ 執行 cd Git/Parser/ && mvn package & ``` <a name="getting-help"></a> ## Getting Help <a name="minus-h"></a> ### `-h` [[ Back to Table of Contents ]](#toc) Type `-h` or `--help` after almost any command to bring up a help menu for that command: ``` \[ andrew@pc01 ~ \]$ du --help 用法:你\[選項\]...\[檔案\]... 或: du \[選項\]... --files0-from=F 對目錄遞歸地總結文件集的磁碟使用情況。 長期權的強制性參數對於短期權也是強制性的。 -0, --null 以 NUL 結束每個輸出行,而不是換行符 -a, --all 計算所有檔案的寫入計數,而不僅僅是目錄 ``` --apparent-size print apparent sizes, rather than disk usage; although ``` ``` the apparent size is usually smaller, it may be ``` ``` larger due to holes in ('sparse') files, internal ``` ``` fragmentation, indirect blocks, and the like ``` -B, --block-size=SIZE 在列印前按 SIZE 縮放大小;例如, ``` '-BM' prints sizes in units of 1,048,576 bytes; ``` ``` see SIZE format below ``` … ``` <a name="man"></a> ### `man` [[ Back to Table of Contents ]](#toc) Type `man` before almost any command to bring up a manual for that command (quit `man` with `q`): ``` LS(1) 使用者指令 LS(1) 姓名 ``` ls - list directory contents ``` 概要 ``` ls [OPTION]... [FILE]... ``` 描述 ``` List information about the FILEs (the current directory by default). ``` ``` Sort entries alphabetically if none of -cftuvSUX nor --sort is speci- ``` ``` fied. ``` ``` Mandatory arguments to long options are mandatory for short options ``` ``` too. ``` … ``` <a name="viewing-and-editing-files"></a> ## Viewing and Editing Files <a name="head-tail-cat-less"></a> ### `head / tail / cat / less` [[ Back to Table of Contents ]](#toc) `head` outputs the first few lines of a file. The `-n` flag specifies the number of lines to show (the default is 10): ``` 列印前三行 ===== \[ andrew@pc01 ~ \]$ 頭 -n 3 c 這 文件 有 ``` `tail` outputs the last few lines of a file. You can get the last `n` lines (like above), or you can get the end of the file beginning from the `N`-th line with `tail -n +N`: ``` 從第 4 行開始列印文件末尾 ============== \[ andrew@pc01 ~ \]$ tail -n +4 c 確切地 六 線 ``` `cat` concatenates a list of files and sends them to the standard output stream (usually the terminal). `cat` can be used with just a single file, or multiple files, and is often used to quickly view them. (**Be warned**: if you use `cat` in this way, you may be accused of a [_Useless Use of Cat_ (UUOC)](http://bit.ly/2SPHE4V), but it's not that big of a deal, so don't worry too much about it.) ``` \[ andrew@pc01 ~ \]$ 貓 a 歸檔一個 \[ andrew@pc01 ~ \]$ 貓 ab 歸檔一個 文件b ``` `less` is another tool for quickly viewing a file -- it opens up a `vim`-like read-only window. (Yes, there is a command called `more`, but `less` -- unintuitively -- offers a superset of the functionality of `more` and is recommended over it.) Learn more (or less?) about [less](http://man7.org/linux/man-pages/man1/less.1.html) and [more](http://man7.org/linux/man-pages/man1/more.1.html) at their `man` pages. <a name="nano-nedit"></a> ### `nano / nedit` [[ Back to Table of Contents ]](#toc) `nano` is a minimalistic command-line text editor. It's a great editor for beginners or people who don't want to learn a million shortcuts. It was more than sufficient for me for the first few years of my coding career (I'm only now starting to look into more powerful editors, mainly because defining your own syntax highlighting in `nano` can be a bit of a pain.) `nedit` is a small graphical editor, it opens up an X Window and allows point-and-click editing, drag-and-drop, syntax highlighting and more. I use `nedit` sometimes when I want to make small changes to a script and re-run it over and over. Other common CLI (command-line interface) / GUI (graphical user interface) editors include `emacs`, `vi`, `vim`, `gedit`, Notepad++, Atom, and lots more. Some cool ones that I've played around with (and can endorse) include Micro, Light Table, and VS Code. All modern editors offer basic conveniences like search and replace, syntax highlighting, and so on. `vi(m)` and `emacs` have more features than `nano` and `nedit`, but they have a much steeper learning curve. Try a few different editors out and find one that works for you! <a name="creating-and-deleting-files"></a> ## Creating and Deleting Files and Directories <a name="touch"></a> ### `touch` [[ Back to Table of Contents ]](#toc) `touch` was created to modify file timestamps, but it can also be used to quickly create an empty file. You can create a new file by opening it with a text editor, like `nano`: ``` \[ andrew@pc01 前 \]$ ls \[ andrew@pc01 ex \]$ 奈米 a ``` _...editing file..._ ``` \[ andrew@pc01 前 \]$ ls A ``` ...or by simply using `touch`: ``` \[ andrew@pc01 ex \]$ touch b && ls ab ``` > **Bonus**: > > Background a process with \^z (Ctrl+z) > > ```bash > [ andrew@pc01 ex ]$ nano a > ``` > > _...editing file, then hit \^z..._ > > ```bash > Use fg to return to nano > > [1]+ Stopped nano a > [ andrew@pc01 ex ]$ fg > ``` > > _...editing file again..._ --- > **Double Bonus:** > > Kill the current (foreground) process by pressing \^c (Ctrl+c) while it’s running > > Kill a background process with `kill %N` where `N` is the job index shown by the `jobs` command <a name="mkdir-rm-rmdir"></a> ### `mkdir / rm / rmdir` [[ Back to Table of Contents ]](#toc) `mkdir` is used to create new, empty directories: ``` \[ andrew@pc01 ex \]$ ls && mkdir c && ls ab ABC ``` You can remove any file with `rm` -- but be careful, this is non-recoverable! ``` \[ andrew@pc01 ex \]$ rm a && ls 西元前 ``` You can add an _"are you sure?"_ prompt with the `-i` flag: ``` \[ andrew@pc01 前 \]$ rm -ib rm:刪除常規空文件“b”? y ``` Remove an empty directory with `rmdir`. If you `ls -a` in an empty directory, you should only see a reference to the directory itself (`.`) and a reference to its parent directory (`..`): ``` \[ andrew@pc01 ex \]$ rmdir c && ls -a 。 .. ``` `rmdir` removes empty directories only: ``` \[ andrew@pc01 ex \]$ cd .. && ls 測試/ \*.txt 0.txt 1.txt a a.txt bc \[ andrew@pc01 ~ \]$ rmdir 測試/ rmdir:無法刪除“test/”:目錄不為空 ``` ...but you can remove a directory -- and all of its contents -- with `rm -rf` (`-r` = recursive, `-f` = force): ``` \[ andrew@pc01 ~ \]$ rm –rf 測試 ``` <a name="moving-and-copying-files"></a> ## Moving and Copying Files, Making Links, Command History <a name="mv-cp-ln"></a> ### `mv / cp / ln` [[ Back to Table of Contents ]](#toc) `mv` moves / renames a file. You can `mv` a file to a new directory and keep the same file name or `mv` a file to a "new file" (rename it): ``` \[ andrew@pc01 ex \]$ ls && mv ae && ls A B C D BCDE ``` `cp` copies a file: ``` \[ andrew@pc01 ex \]$ cp e e2 && ls BCDE E2 ``` `ln` creates a hard link to a file: ``` ln 的第一個參數是 TARGET,第二個參數是 NEW LINK ================================= \[ andrew@pc01 ex \]$ ln bf && ls bcde e2 f ``` `ln -s` creates a soft link to a file: ``` \[ andrew@pc01 ex \]$ ln -sbg && ls BCDE E2 FG ``` Hard links reference the same actual bytes in memory which contain a file, while soft links refer to the original file name, which itself points to those bytes. [You can read more about soft vs. hard links here.](http://bit.ly/2D0W8cN) <a name="command-history"></a> ### Command History [[ Back to Table of Contents ]](#toc) `bash` has two big features to help you complete and re-run commands, the first is _tab completion_. Simply type the first part of a command, hit the \<tab\> key, and let the terminal guess what you're trying to do: ``` \[ andrew@pc01 目錄 \]$ ls 另一個長檔名 這是一個長檔名 一個新檔名 \[ andrew@pc01 目錄 \]$ ls t ``` _...hit the TAB key after typing `ls t` and the command is completed..._ ``` \[ andrew@pc01 dir \]$ ls 這是檔名 這是長檔名 ``` You may have to hit \<TAB\> multiple times if there's an ambiguity: ``` \[ andrew@pc01 目錄 \]$ ls a \[ andrew@pc01 目錄 \]$ ls an 一個新檔名另一個長檔名 ``` `bash` keeps a short history of the commands you've typed previously and lets you search through those commands by typing \^r (Ctrl+r): ``` \[ andrew@pc01 目錄 \] ``` _...hit \^r (Ctrl+r) to search the command history..._ ``` (反向搜尋)``: ``` _...type 'anew' and the last command containing this is found..._ ``` (reverse-i-search)`anew': 觸碰新檔名 ``` <a name="directory-trees-disk-usage-processes"></a> ## Directory Trees, Disk Usage, and Processes <a name="mkdir--p-tree"></a> ### `mkdir –p / tree` [[ Back to Table of Contents ]](#toc) `mkdir`, by default, only makes a single directory. This means that if, for instance, directory `d/e` doesn't exist, then `d/e/f` can't be made with `mkdir` by itself: ``` \[ andrew@pc01 ex \]$ ls && mkdir d/e/f ABC mkdir:無法建立目錄「d/e/f」:沒有這樣的檔案或目錄 ``` But if we pass the `-p` flag to `mkdir`, it will make all directories in the path if they don't already exist: ``` \[ andrew@pc01 ex \]$ mkdir -pd/e/f && ls A B C D ``` `tree` can help you better visualise a directory's structure by printing a nicely-formatted directory tree. By default, it prints the entire tree structure (beginning with the specified directory), but you can restrict it to a certain number of levels with the `-L` flag: ``` \[ andrew@pc01 前 \]$ 樹 -L 2 。 |-- 一個 |-- b |-- c `--d ``` `--e ``` 3個目錄,2個文件 ``` You can hide empty directories in `tree`'s output with `--prune`. Note that this also removes "recursively empty" directories, or directories which aren't empty _per se_, but which contain only other empty directories, or other recursively empty directories: ``` \[ andrew@pc01 ex \]$ 樹 --prune 。 |-- 一個 `--b ``` <a name="df-du-ps"></a> ### `df / du / ps` [[ Back to Table of Contents ]](#toc) `df` is used to show how much space is taken up by files for the disks or your system (hard drives, etc.). ``` \[ andrew@pc01 前 \]$ df -h 已使用的檔案系統大小 可用 使用% 安裝於 udev 126G 0 126G 0% /dev tmpfs 26G 2.0G 24G 8% /執行 /dev/mapper/ubuntu--vg-root 1.6T 1.3T 252G 84% / … ``` In the above command, `-h` doesn't mean "help", but "human-readable". Some commands use this convention to display file / disk sizes with `K` for kilobytes, `G` for gigabytes, and so on, instead of writing out a gigantic integer number of bytes. `du` shows file space usage for a particular directory and its subdirectories. If you want to know how much space is free on a given hard drive, use `df`; if you want to know how much space a directory is taking up, use `du`: ``` \[ andrew@pc01 ex \]$ 你 4 ./d/e/f 8./d/e 12 ./天 4./c 20 . ``` `du` takes a `--max-depth=N` flag, which only shows directories `N` levels down (or fewer) from the specified directory: ``` \[ andrew@pc01 ex \]$ du -h --max-深度=1 12K./天 4.0K./c 20K。 ``` `ps` shows all of the user's currently-running processes (aka. jobs): ``` \[ andrew@pc01 前 \]$ ps PID TTY 時間 CMD 16642 分/15 00:00:00 ps 25409 點/15 00:00:00 重擊 ``` <a name="basic-misc"></a> ## Miscellaneous <a name="passwd-logout-exit"></a> ### `passwd / logout / exit` [[ Back to Table of Contents ]](#toc) Change your account password with `passwd`. It will ask for your current password for verification, then ask you to enter the new password twice, so you don't make any typos: ``` \[ andrew@pc01 目錄 \]$ 密碼 更改安德魯的密碼。 (目前)UNIX 密碼: 輸入新的 UNIX 密碼: 重新輸入新的 UNIX 密碼: passwd:密碼更新成功 ``` `logout` exits a shell you’ve logged in to (where you have a user account): ``` \[ andrew@pc01 目錄 \]$ 註銷 ────────────────────────────────────────────────── ── ────────────────────────────── 會話已停止 ``` - Press <return> to exit tab ``` ``` - Press R to restart session ``` ``` - Press S to save terminal output to file ``` ``` `exit` exits any kind of shell: ``` \[ andrew@pc01 ~ \]$ 退出 登出 ────────────────────────────────────────────────── ── ────────────────────────────── 會話已停止 ``` - Press <return> to exit tab ``` ``` - Press R to restart session ``` ``` - Press S to save terminal output to file ``` ``` <a name="clear-glob"></a> ### `clear / *` [[ Back to Table of Contents ]](#toc) Run `clear` to move the current terminal line to the top of the screen. This command just adds blank lines below the current prompt line. It's good for clearing your workspace. Use the glob (`*`, aka. Kleene Star, aka. wildcard) when looking for files. Notice the difference between the following two commands: ``` \[ andrew@pc01 ~ \]$ ls Git/Parser/source/ PArrayUtils.java PFile.java PSQLFile.java PWatchman.java PDateTimeUtils.java PFixedWidthFile.java PStringUtils.java PXSVFile.java PDelimitedFile.java PNode.java PTextFile.java Parser.java \[ andrew@pc01 ~ \]$ ls Git/Parser/source/PD\* Git/Parser/source/PDateTimeUtils.java Git/Parser/source/PDelimitedFile.java ``` The glob can be used multiple times in a command and matches zero or more characers: ``` \[ andrew@pc01 ~ \]$ ls Git/Parser/source/P *D* m\* Git/Parser/source/PDateTimeUtils.java Git/Parser/source/PDelimitedFile.java ``` <a name="intermediate"></a> # Intermediate <a name="disk-memory-processor"></a> ## Disk, Memory, and Processor Usage <a name="ncdu"></a> ### `ncdu` [[ Back to Table of Contents ]](#toc) `ncdu` (NCurses Disk Usage) provides a navigable overview of file space usage, like an improved `du`. It opens a read-only `vim`-like window (press `q` to quit): ``` \[ andrew@pc01 ~ \]$ ncdu ncdu 1.11 ~ 使用箭頭鍵導航,按 ?求助 \--- /home/安德魯 ------------------------------------------- ------------------ 148.2 MiB \[##########\] /.m2 91.5 MiB \[######\] /.sbt 79.8 MiB \[######\] /.cache 64.9 MiB \[####\] /.ivy2 40.6 MiB \[##\] /.sdkman 30.2 MiB \[##\] /.local 27.4 MiB \[#\] /.mozilla 24.4 MiB \[#\] /.nanobackups 10.2 MiB \[ \] .confout3.txt ``` 8.4 MiB [ ] /.config ``` ``` 5.9 MiB [ ] /.nbi ``` ``` 5.8 MiB [ ] /.oh-my-zsh ``` ``` 4.3 MiB [ ] /Git ``` ``` 3.7 MiB [ ] /.myshell ``` ``` 1.7 MiB [ ] /jdoc ``` ``` 1.5 MiB [ ] .confout2.txt ``` ``` 1.5 MiB [ ] /.netbeans ``` ``` 1.1 MiB [ ] /.jenv ``` 564.0 KiB \[ \] /.rstudio-desktop 磁碟使用總量:552.7 MiB 表觀大小:523.6 MiB 專案:14618 ``` <a name="top-htop"></a> ### `top / htop` [[ Back to Table of Contents ]](#toc) `top` displays all currently-running processes and their owners, memory usage, and more. `htop` is an improved, interactive `top`. (Note: you can pass the `-u username` flag to restrict the displayed processes to only those owner by `username`.) ``` \[ andrew@pc01 ~ \]$ htop 1 \[ 0.0%\] 9 \[ 0.0%\] 17 \[ 0.0%\] 25 \[ 0.0%\] 2 \[ 0.0%\] 10 \[ 0.0%\] 18 \[ 0.0%\] 26 \[ 0.0%\] 3 \[ 0.0%\] 11 \[ 0.0%\] 19 \[ 0.0%\] 27 \[ 0.0%\] 4 \[ 0.0%\] 12 \[ 0.0%\] 20 \[ 0.0%\] 28 \[ 0.0%\] 5 \[ 0.0%\] 13 \[ 0.0%\] 21 \[| 1.3%\] 29 \[ 0.0%\] 6 \[ 0.0%\] 14 \[ 0.0%\] 22 \[ 0.0%\] 30 \[| 0.6%\] 7 \[ 0.0%\] 15 \[ 0.0%\] 23 \[ 0.0%\] 31 \[ 0.0%\] 8 \[ 0.0%\] 16 \[ 0.0%\] 24 \[ 0.0%\] 32 \[ 0.0%\] Mem\[|||||||||||||||||||1.42G/252G\] 任務:188、366 個; 1 執行 交換電壓\[| 2.47G/256G\]平均負載:0.00 0.00 0.00 ``` Uptime: 432 days(!), 00:03:55 ``` PID USER PRI NI VIRT RES SHR S CPU% MEM% TIME+ 指令 9389 安德魯 20 0 23344 3848 2848 R 1.3 0.0 0:00.10 htop 10103 根 20 0 3216M 17896 2444 S 0.7 0.0 5h48:56 /usr/bin/dockerd ``` 1 root 20 0 181M 4604 2972 S 0.0 0.0 15:29.66 /lib/systemd/syst ``` 533 根 20 0 44676 6908 6716 S 0.0 0.0 11:19.77 /lib/systemd/syst 546 根 20 0 244M 0 0 S 0.0 0.0 0:01.39 /sbin/lvmetad -f 1526 根 20 0 329M 2252 1916 S 0.0 0.0 0:00.00 /usr/sbin/ModemMa 1544 根 20 0 329M 2252 1916 S 0.0 0.0 0:00.06 /usr/sbin/ModemMa F1Help F2Setup F3SearchF4FilterF5Tree F6SortByF7Nice -F8Nice +F9Kill F10Quit ``` <a name="REPLs-software-versions"></a> ## REPLs and Software Versions <a name="REPLs"></a> ### REPLs [[ Back to Table of Contents ]](#toc) A **REPL** is a Read-Evaluate-Print Loop, similar to the command line, but usually used for particular programming languages. You can open the Python REPL with the `python` command (and quit with the `quit()` function): ``` \[ andrew@pc01 ~ \]$ python Python 3.5.2(默認,2018 年 11 月 12 日,13:43:14)... > > > 辭職() ``` Open the R REPL with the `R` command (and quit with the `q()` function): ``` \[ andrew@pc01 ~ \]$ R R版3.5.2(2018-12-20)--「蛋殼冰屋」... > q() 儲存工作區影像? \[是/否/c\]: 否 ``` Open the Scala REPL with the `scala` command (and quit with the `:quit` command): ``` \[ andrew@pc01 ~ \]$ scala 歡迎使用 Scala 2.11.12 ... 斯卡拉>:退出 ``` Open the Java REPL with the `jshell` command (and quit with the `/exit` command): ``` \[ andrew@pc01 ~ \]$ jshell |歡迎使用 JShell——版本 11.0.1 ... jshell> /退出 ``` Alternatively, you can exit any of these REPLs with \^d (Ctrl+d). \^d is the EOF (end of file) marker on Unix and signifies the end of input. <a name="version"></a> ### `-version / --version / -v` [[ Back to Table of Contents ]](#toc) Most commands and programs have a `-version` or `--version` flag which gives the software version of that command or program. Most applications make this information easily available: ``` \[ andrew@pc01 ~ \]$ ls --version ls (GNU coreutils) 8.25 ... \[ andrew@pc01 ~ \]$ ncdu -版本 NCDU 1.11 \[ andrew@pc01 ~ \]$ python --version Python 3.5.2 ``` ...but some are less intuitive: ``` \[ andrew@pc01 ~ \]$ sbt scalaVersion … \[資訊\]2.12.4 ``` Note that some programs use `-v` as a version flag, while others use `-v` to mean "verbose", which will run the application while printing lots of diagnostic or debugging information: ``` SCP(1) BSD 通用指令手冊 SCP(1) 姓名 ``` scp -- secure copy (remote file copy program) ``` … -v 詳細模式。導致 scp 和 ssh(1) 列印偵錯訊息 ``` about their progress. This is helpful in debugging connection, ``` ``` authentication, and configuration problems. ``` … ``` <a name="env-vars-aliases"></a> ## Environment Variables and Aliases <a name="env-vars"></a> ### Environment Variables [[ Back to Table of Contents ]](#toc) **Environment variables** (sometimes shortened to "env vars") are persistent variables that can be created and used within your `bash` shell. They are defined with an equals sign (`=`) and used with a dollar sign (`$`). You can see all currently-defined env vars with `printenv`: ``` \[ andrew@pc01 ~ \]$ printenv SPARK\_HOME=/usr/local/spark 術語=xterm … ``` Set a new environment variable with an `=` sign (don't put any spaces before or after the `=`, though!): ``` \[ andrew@pc01 ~ \]$ myvar=你好 ``` Print a specific env var to the terminal with `echo` and a preceding `$` sign: ``` \[ andrew@pc01 ~ \]$ echo $myvar 你好 ``` Environment variables which contain spaces or other whitespace should be surrounded by quotes (`"..."`). Note that reassigning a value to an env var overwrites it without warning: ``` \[ andrew@pc01 ~ \]$ myvar="你好,世界!" && 回顯 $myvar 你好世界! ``` Env vars can also be defined using the `export` command. When defined this way, they will also be available to sub-processes (commands called from this shell): ``` \[ andrew@pc01 ~ \]$ export myvar="另一" && echo $myvar 另一個 ``` You can unset an environment variable by leaving the right-hand side of the `=` blank or by using the `unset` command: ``` \[ andrew@pc01 ~ \]$ 取消設定 mynewvar \[ andrew@pc01 ~ \]$ echo $mynewvar ``` <a name="aliases"></a> ### Aliases [[ Back to Table of Contents ]](#toc) **Aliases** are similar to environment variables but are usually used in a different way -- to replace long commands with shorter ones: ``` \[ andrew@pc01 apidocs \]$ ls -l -a -h -t 總計 220K drwxr-xr-x 5 安德魯 安德魯 4.0K 12 月 21 日 12:37 。 -rw-r--r-- 1 安德魯 安德魯 9.9K 十二月 21 12:37 help-doc.html -rw-r--r-- 1 安德魯 安德魯 4.5K 12 月 21 日 12:37 script.js … \[ andrew@pc01 apidocs \]$ 別名 lc="ls -l -a -h -t" \[ andrew@pc01 apidocs \]$ lc 總計 220K drwxr-xr-x 5 安德魯 安德魯 4.0K 12 月 21 日 12:37 。 -rw-r--r-- 1 安德魯 安德魯 9.9K 十二月 21 12:37 help-doc.html -rw-r--r-- 1 安德魯 安德魯 4.5K 12 月 21 日 12:37 script.js … ``` You can remove an alias with `unalias`: ``` \[ andrew@pc01 apidocs \]$ unalias lc \[ andrew@pc01 apidocs \]$ lc 目前未安裝程式“lc”。 … ``` > **Bonus:** > > [Read about the subtle differences between environment variables and aliases here.](http://bit.ly/2TDG8Tx) > > [Some programs, like **git**, allow you to define aliases specifically for that software.](http://bit.ly/2TG8X1A) <a name="basic-bash-scripting"></a> ## Basic `bash` Scripting <a name="bash-scripts"></a> ### `bash` Scripts [[ Back to Table of Contents ]](#toc) `bash` scripts (usually ending in `.sh`) allow you to automate complicated processes, packaging them into reusable functions. A `bash` script can contain any number of normal shell commands: ``` \[ andrew@pc01 ~ \]$ echo "ls && touch file && ls" > ex.sh ``` A shell script can be executed with the `source` command or the `sh` command: ``` \[ andrew@pc01 ~ \]$ 源 ex.sh 桌面 Git TEST c ex.sh 專案測試 桌面 Git TEST c ex.sh 檔案專案測試 ``` Shell scripts can be made executable with the `chmod` command (more on this later): ``` \[ andrew@pc01 ~ \]$ echo "ls && touch file2 && ls" > ex2.sh \[ andrew@pc01 ~ \]$ chmod +x ex2.sh ``` An executable shell script can be run by preceding it with `./`: ``` \[ andrew@pc01 ~ \]$ ./ex2.sh 桌面 Git TEST c ex.sh ex2.sh 檔案專案測試 桌面 Git TEST c ex.sh ex2.sh 檔案 file2 專案測試 ``` Long lines of code can be split by ending a command with `\`: ``` \[ andrew@pc01 ~ \]$ echo "for i in {1..3}; do echo \\ > \\"歡迎\\$i次\\";完成” > ex3.sh ``` Bash scripts can contain loops, functions, and more! ``` \[ andrew@pc01 ~ \]$ 源 ex3.sh 歡迎1次 歡迎2次 歡迎3次 ``` <a name="custom-prompt-ls"></a> ### Custom Prompt and `ls` [[ Back to Table of Contents ]](#toc) Bash scripting can make your life a whole lot easier and more colourful. [Check out this great bash scripting cheat sheet.](https://devhints.io/bash) `$PS1` (Prompt String 1) is the environment variable that defines your main shell prompt ([learn about the other prompts here](http://bit.ly/2SPgsmT)): ``` \[ andrew@pc01 ~ \]$ printf "%q" $PS1 $'\\n\\\[\\E\[1m\\\]\\\[\\E\[30m\\\]\\A'$'\\\[\\E\[37m\\\]|\\\[\\E\[36m\\\]\\u\\\[\\E\[37m \\\]@\\\[\\E\[34m\\\]\\h'$'\\\[\\E\[32m\\\]\\W\\\[\\E\[37m\\\]|'$'\\\[\\E(B\\E\[m\\\] ' ``` You can change your default prompt with the `export` command: ``` \[ andrew@pc01 ~ \]$ export PS1="\\n此處指令> " 此處指令> echo $PS1 \\n此處指令> ``` ...[you can add colours, too!](http://bit.ly/2TMbEit): ``` 此處指令> export PS1="\\e\[1;31m\\n程式碼: \\e\[39m" (這應該是紅色的,但在 Markdown 中可能不會這樣顯示) =============================== 程式碼:回顯$PS1 \\e\[1;31m\\n程式碼: \\e\[39m ``` You can also change the colours shown by `ls` by editing the `$LS_COLORS` environment variable: ``` (同樣,這些顏色可能不會出現在 Markdown 中) =========================== 程式碼:ls 桌面 Git TEST c ex.sh ex2.sh ex3.sh 檔案 file2 專案測試 程式碼:匯出 LS\_COLORS='di=31:fi=0:ln=96:or=31:mi=31:ex=92' 程式碼:ls 桌面 Git TEST c ex.sh ex2.sh ex3.sh 檔案 file2 專案測試 ``` <a name="config-files"></a> ## Config Files <a name="config-bashrc"></a> ### Config Files / `.bashrc` [[ Back to Table of Contents ]](#toc) If you tried the commands in the last section and logged out and back in, you may have noticed that your changes disappeared. _config_ (configuration) files let you maintain settings for your shell or for a particular program every time you log in (or run that program). The main configuration file for a `bash` shell is the `~/.bashrc` file. Aliases, environment variables, and functions added to `~/.bashrc` will be available every time you log in. Commands in `~/.bashrc` will be run every time you log in. If you edit your `~/.bashrc` file, you can reload it without logging out by using the `source` command: ``` \[ andrew@pc01 ~ \]$ nano ~/.bashrc ``` _...add the line `echo “~/.bashrc loaded!”` to the top of the file_... ``` \[ andrew@pc01 ~ \]$ 源 ~/.bashrc ~/.bashrc 已載入! ``` _...log out and log back in..._ ``` 最後登入:2019 年 1 月 11 日星期五 10:29:07 從 111.11.11.111 ~/.bashrc 已加載! \[ 安德魯@pc01 ~ \] ``` <a name="types-of-shells"></a> ### Types of Shells [[ Back to Table of Contents ]](#toc) _Login_ shells are shells you log in to (where you have a username). _Interactive_ shells are shells which accept commands. Shells can be login and interactive, non-login and non-interactive, or any other combination. In addition to `~/.bashrc`, there are a few other scripts which are `sourced` by the shell automatically when you log in or log out. These are: - `/etc/profile` - `~/.bash_profile` - `~/.bash_login` - `~/.profile` - `~/.bash_logout` - `/etc/bash.bash_logout` Which of these scripts are sourced, and the order in which they're sourced, depend on the type of shell opened. See [the bash man page](https://linux.die.net/man/1/bash) and [these](http://bit.ly/2TGCwA8) Stack Overflow [posts](http://bit.ly/2TFHFsf) for more information. Note that `bash` scripts can `source` other scripts. For instance, in your `~/.bashrc`, you could include the line: ``` 來源~/.bashrc\_addl ``` ...which would also `source` that `.bashrc_addl` script. This file can contain its own aliases, functions, environment variables, and so on. It could, in turn, `source` other scripts, as well. (Be careful to avoid infinite loops of script-sourcing!) It may be helpful to split commands into different shell scripts based on functionality or machine type (Ubuntu vs. Red Hat vs. macOS), for example: - `~/.bash_ubuntu` -- configuration specific to Ubuntu-based machines - `~/.bashrc_styles` -- aesthetic settings, like `PS1` and `LS_COLORS` - `~/.bash_java` -- configuration specific to the Java language I try to keep separate `bash` files for aesthetic configurations and OS- or machine-specific code, and then I have one big `bash` file containing shortcuts, etc. that I use on every machine and every OS. Note that there are also _different shells_. `bash` is just one kind of shell (the "Bourne Again Shell"). Other common ones include `zsh`, `csh`, `fish`, and more. Play around with different shells and find one that's right for you, but be aware that this tutorial contains `bash` shell commands only and not everything listed here (maybe none of it) will be applicable to shells other than `bash`. <a name="finding-things"></a> ## Finding Things <a name="whereis-which-whatis"></a> ### `whereis / which / whatis` [[ Back to Table of Contents ]](#toc) `whereis` searches for "possibly useful" files related to a particular command. It will attempt to return the location of the binary (executable machine code), source (code source files), and `man` page for that command: ``` \[ andrew@pc01 ~ \]$ whereis ls ls: /bin/ls /usr/share/man/man1/ls.1.gz ``` `which` will only return the location of the binary (the command itself): ``` \[ andrew@pc01 ~ \]$ 其中 ls /bin/ls ``` `whatis` prints out the one-line description of a command from its `man` page: ``` \[ andrew@pc01 ~ \]$ 什麼是哪裡是哪個什麼是 whereis (1) - 尋找指令的二進位、原始檔和手冊頁文件 which (1) - 定位指令 Whatis (1) - 顯示一行手冊頁描述 ``` `which` is useful for finding the "original version" of a command which may be hidden by an alias: ``` \[ andrew@pc01 ~ \]$ 別名 ls="ls -l" “original” ls 已被上面定義的別名“隱藏” =========================== \[ andrew@pc01 ~ \]$ ls 總計 36 drwxr-xr-x 2 安德魯 andrew 4096 Jan 9 14:47 桌面 drwxr-xr-x 4 安德魯 安德魯 4096 十二月 6 10:43 Git … 但我們仍然可以使用返回的位置來呼叫「原始」ls ======================= \[ andrew@pc01 ~ \]$ /bin/ls 桌面 Git TEST c ex.sh ex2.sh ex3.sh 檔案 file2 專案測試 ``` <a name="locate-find"></a> ### `locate / find` [[ Back to Table of Contents ]](#toc) `locate` finds a file anywhere on the system by referring to a semi-regularly-updated cached list of files: ``` \[ andrew@pc01 ~ \]$ 找到 README.md /home/andrew/.config/micro/plugins/gotham-colors/README.md /home/andrew/.jenv/README.md /home/andrew/.myshell/README.md … ``` Because it's just searching a list, `locate` is usually faster than the alternative, `find`. `find` iterates through the file system to find the file you're looking for. Because it's actually looking at the files which _currently_ exist on the system, though, it will always return an up-to-date list of files, which is not necessarily true with `locate`. ``` \[ andrew@pc01 ~ \]$ find ~/ -iname "README.md" /home/andrew/.jenv/README.md /home/andrew/.config/micro/plugins/gotham-colors/README.md /home/andrew/.oh-my-zsh/plugins/ant/README.md … ``` `find` was written for the very first version of Unix in 1971, and is therefore much more widely available than `locate`, which was added to GNU in 1994. `find` has many more features than `locate`, and can search by file age, size, ownership, type, timestamp, permissions, depth within the file system; `find` can search using regular expressions, execute commands on files it finds, and more. When you need a fast (but possibly outdated) list of files, or you’re not sure what directory a particular file is in, use `locate`. When you need an accurate file list, maybe based on something other than the files’ names, and you need to do something with those files, use `find`. <a name="downloading-things"></a> ## Downloading Things <a name="ping-wget-curl"></a> ### `ping / wget / curl` [[ Back to Table of Contents ]](#toc) `ping` attempts to open a line of communication with a network host. Mainly, it's used to check whether or not your Internet connection is down: ``` \[ andrew@pc01 ~ \]$ ping google.com PING google.com (74.125.193.100) 56(84) 位元組資料。 使用 32 位元組資料 Ping 74.125.193.100: 來自 74.125.193.100 的回覆:位元組=32 時間<1ms TTL=64 … ``` `wget` is used to easily download a file from the Internet: ``` \[ andrew@pc01 ~ \]$ wget \\ > http://releases.ubuntu.com/18.10/ubuntu-18.10-desktop-amd64.iso ``` `curl` can be used just like `wget` (don’t forget the `--output` flag): ``` \[ andrew@pc01 ~ \]$ 捲曲 \\ > http://releases.ubuntu.com/18.10/ubuntu-18.10-desktop-amd64.iso \\ > \--輸出ubuntu.iso ``` `curl` and `wget` have their own strengths and weaknesses. `curl` supports many more protocols and is more widely available than `wget`; `curl` can also send data, while `wget` can only receive data. `wget` can download files recursively, while `curl` cannot. In general, I use `wget` when I need to download things from the Internet. I don’t often need to send data using `curl`, but it’s good to be aware of it for the rare occasion that you do. <a name="apt-gunzip-tar-gzip"></a> ### `apt / gunzip / tar / gzip` [[ Back to Table of Contents ]](#toc) Debian-descended Linux distributions have a fantastic package management tool called `apt`. It can be used to install, upgrade, or delete software on your machine. To search `apt` for a particular piece of software, use `apt search`, and install it with `apt install`: ``` \[ andrew@pc01 ~ \]$ apt 搜尋漂白位 ...bleachbit/bionic、bionic 2.0-2 全部 從系統中刪除不需要的文件 您需要“sudo”來安裝軟體 ============== \[ andrew@pc01 ~ \]$ sudo apt installbleachbit ``` Linux software often comes packaged in `.tar.gz` ("tarball") files: ``` \[ andrew@pc01 ~ \]$ wget \\ > https://github.com/atom/atom/releases/download/v1.35.0-beta0/atom-amd64.tar.gz ``` ...these types of files can be unzipped with `gunzip`: ``` \[ andrew@pc01 ~ \]$gunzipatom-amd64.tar.gz && ls 原子 amd64.tar ``` A `.tar.gz` file will be `gunzip`-ped to a `.tar` file, which can be extracted to a directory of files using `tar -xf` (`-x` for "extract", `-f` to specify the file to "untar"): ``` \[ andrew@pc01 ~ \]$ tar -xfatom-amd64.tar && mv \\ 原子-beta-1.35.0-beta0-amd64 原子 && ls 原子atom-amd64.tar ``` To go in the reverse direction, you can create (`-c`) a tar file from a directory and zip it (or unzip it, as appropriate) with `-z`: ``` \[ andrew@pc01 ~ \]$ tar -zcf 壓縮.tar.gz 原子 && ls 原子atom-amd64.tar壓縮.tar.gz ``` `.tar` files can also be zipped with `gzip`: ``` \[ andrew@pc01 ~ \]$ gzipatom-amd64.tar && ls 原子 原子-amd64.tar.gz 壓縮.tar.gz ``` <a name="redirecting-io"></a> ## Redirecting Input and Output <a name="pipe-gt-lt-echo-printf"></a> ### `| / > / < / echo / printf` [[ Back to Table of Contents ]](#toc) By default, shell commands read their input from the standard input stream (aka. stdin or 0) and write to the standard output stream (aka. stdout or 1), unless there’s an error, which is written to the standard error stream (aka. stderr or 2). `echo` writes text to stdout by default, which in most cases will simply print it to the terminal: ``` \[ andrew@pc01 ~ \]$ 回顯“你好” 你好 ``` The pipe operator, `|`, redirects the output of the first command to the input of the second command: ``` 'wc'(字數)傳回檔案中的行數、字數、位元組數 ======================== \[ andrew@pc01 ~ \]$ echo "範例文件" |廁所 ``` 1 2 17 ``` ``` `>` redirects output from stdout to a particular location ``` \[ andrew@pc01 ~ \]$ echo "test" > 文件 && 頭文件 測試 ``` `printf` is an improved `echo`, allowing formatting and escape sequences: ``` \[ andrew@pc01 ~ \]$ printf "1\\n3\\n2" 1 3 2 ``` `<` gets input from a particular location, rather than stdin: ``` 'sort' 依字母/數字順序對檔案的行進行排序 ======================== \[ andrew@pc01 ~ \]$ sort <(printf "1\\n3\\n2") 1 2 3 ``` Rather than a [UUOC](#viewing-and-editing-files), the recommended way to send the contents of a file to a command is to use `<`. Note that this causes data to "flow" right-to-left on the command line, rather than (the perhaps more natural, for English-speakers) left-to-right: ``` \[ andrew@pc01 ~ \]$ printf "1\\n3\\n2" > 文件 && 排序 < 文件 1 2 3 ``` <a name="std-tee"></a> ### `0 / 1 / 2 / tee` [[ Back to Table of Contents ]](#toc) 0, 1, and 2 are the standard input, output, and error streams, respectively. Input and output streams can be redirected with the `|`, `>`, and `<` operators mentioned previously, but stdin, stdout, and stderr can also be manipulated directly using their numeric identifiers: Write to stdout or stderr with `>&1` or `>&2`: ``` \[ andrew@pc01 ~ \]$ 貓測試 回顯“標準輸出”>&1 回顯“標準錯誤”>&2 ``` By default, stdout and stderr both print output to the terminal: ``` \[ andrew@pc01 ~ \]$ ./測試 標準錯誤 標準輸出 ``` Redirect stdout to `/dev/null` (only print output sent to stderr): ``` \[ andrew@pc01 ~ \]$ ./test 1>/dev/null 標準錯誤 ``` Redirect stderr to `/dev/null` (only print output sent to stdout): ``` \[ andrew@pc01 ~ \]$ ./test 2>/dev/null 標準輸出 ``` Redirect all output to `/dev/null` (print nothing): ``` \[ andrew@pc01 ~ \]$ ./test &>/dev/null ``` Send output to stdout and any number of additional locations with `tee`: ``` \[ andrew@pc01 ~ \]$ ls && echo "測試" | tee 文件1 文件2 文件3 && ls 文件0 測試 文件0 文件1 文件2 文件3 ``` <a name="advanced"></a> # Advanced <a name="superuser"></a> ## Superuser <a name="sudo-su"></a> ### `sudo / su` [[ Back to Table of Contents ]](#toc) You can check what your username is with `whoami`: ``` \[ andrew@pc01 abc \]$ whoami 安德魯 ``` ...and run a command as another user with `sudo -u username` (you will need that user's password): ``` \[ andrew@pc01 abc \]$ sudo -u 測試觸摸 def && ls -l 總計 0 -rw-r--r-- 1 次測試 0 Jan 11 20:05 def ``` If `–u` is not provided, the default user is the superuser (usually called "root"), with unlimited permissions: ``` \[ andrew@pc01 abc \]$ sudo touch ghi && ls -l 總計 0 -rw-r--r-- 1 次測試 0 Jan 11 20:05 def -rw-r--r-- 1 root root 0 Jan 11 20:14 ghi ``` Use `su` to become another user temporarily (and `exit` to switch back): ``` \[ andrew@pc01 abc \]$ su 測試 密碼: test@pc01:/home/andrew/abc$ whoami 測試 test@pc01:/home/andrew/abc$ 退出 出口 \[ andrew@pc01 abc \]$ whoami 安德魯 ``` [Learn more about the differences between `sudo` and `su` here.](http://bit.ly/2SKQH77) <a name="click-click"></a> ### `!!` [[ Back to Table of Contents ]](#toc) The superuser (usually "root") is the only person who can install software, create users, and so on. Sometimes it's easy to forget that, and you may get an error: ``` \[ andrew@pc01 ~ \]$ apt 安裝 ruby E:無法開啟鎖定檔案 /var/lib/dpkg/lock-frontend - 開啟(13:權限被拒絕) E: 無法取得 dpkg 前端鎖定 (/var/lib/dpkg/lock-frontend),您是 root 嗎? ``` You could retype the command and add `sudo` at the front of it (run it as the superuser): ``` \[ andrew@pc01 ~ \]$ sudo apt install ruby 正在閱讀包裝清單... ``` Or, you could use the `!!` shortcut, which retains the previous command: ``` \[ andrew@pc01 ~ \]$ apt 安裝 ruby E:無法開啟鎖定檔案 /var/lib/dpkg/lock-frontend - 開啟(13:權限被拒絕) E: 無法取得 dpkg 前端鎖定 (/var/lib/dpkg/lock-frontend),您是 root 嗎? \[ andrew@pc01 ~ \]$ sudo !! sudo apt 安裝 ruby 正在閱讀包裝清單... ``` By default, running a command with `sudo` (and correctly entering the password) allows the user to run superuser commands for the next 15 minutes. Once those 15 minutes are up, the user will again be prompted to enter the superuser password if they try to run a restricted command. <a name="file-permissions"></a> ## File Permissions <a name="file-permissions-sub"></a> ### File Permissions [[ Back to Table of Contents ]](#toc) Files may be able to be read (`r`), written to (`w`), and/or executed (`x`) by different users or groups of users, or not at all. File permissions can be seen with the `ls -l` command and are represented by 10 characters: ``` \[ andrew@pc01 ~ \]$ ls -lh 總計 8 drwxr-xr-x 4 安德魯 安德魯 4.0K 1 月 4 日 19:37 品嚐 -rwxr-xr-x 1 安德魯 安德魯 40 Jan 11 16:16 測試 -rw-r--r-- 1 安德魯 安德魯 0 一月 11 16:34 tist ``` The first character of each line represents the type of file, (`d` = directory, `l` = link, `-` = regular file, and so on); then there are three groups of three characters which represent the permissions held by the user (u) who owns the file, the permissions held by the group (g) which owns the file, and the permissions held any other (o) users. (The number which follows this string of characters is the number of links in the file system to that file (4 or 1 above).) `r` means that person / those people have read permission, `w` is write permission, `x` is execute permission. If a directory is “executable”, that means it can be opened and its contents can be listed. These three permissions are often represented with a single three-digit number, where, if `x` is enabled, the number is incremented by 1, if `w` is enabled, the number is incremented by 2, and if `r` is enabled, the number is incremented by 4. Note that these are equivalent to binary digits (`r-x` -> `101` -> `5`, for example). So the above three files have permissions of 755, 755, and 644, respectively. The next two strings in each list are the name of the owner (`andrew`, in this case) and the group of the owner (also `andrew`, in this case). Then comes the size of the file, its most recent modification time, and its name. The `–h` flag makes the output human readable (i.e. printing `4.0K` instead of `4096` bytes). <a name="chmod-chown"></a> ### `chmod / chown` [[ Back to Table of Contents ]](#toc) File permissions can be modified with `chmod` by setting the access bits: ``` \[ andrew@pc01 ~ \]$ chmod 777 測試 && chmod 000 tit && ls -lh 總計 8.0K drwxr-xr-x 4 安德魯 安德魯 4.0K 1 月 4 日 19:37 品嚐 -rwxrwxrwx 1 安德魯 安德魯 40 Jan 11 16:16 測試 \---------- 1 安德魯 安德魯 0 一月 11 16:34 tist ``` ...or by adding (`+`) or removing (`-`) `r`, `w`, and `x` permissions with flags: ``` \[ andrew@pc01 ~ \]$ chmod +rwx Tist && chmod -w 測試 && ls -lh chmod:測試:新權限是 r-xrwxrwx,而不是 r-xr-xr-x 總計 8.0K drwxr-xr-x 4 安德魯 安德魯 4.0K 1 月 4 日 19:37 品嚐 -r-xrwxrwx 1 安德魯 安德魯 40 Jan 11 16:16 測試 -rwxr-xr-x 1 安德魯 安德魯 0 一月 11 16:34 tist ``` The user who owns a file can be changed with `chown`: ``` \[ andrew@pc01 ~ \]$ sudo chown 碼頭測試 ``` The group which owns a file can be changed with `chgrp`: ``` \[ andrew@pc01 ~ \]$ sudo chgrp hadoop tit && ls -lh 總計 8.0K drwxr-xr-x 4 安德魯 安德魯 4.0K 1 月 4 日 19:37 品嚐 \-----w--w- 1 瑪麗娜·安德魯 2011 年 1 月 40 日 16:16 測試 -rwxr-xr-x 1 安德魯 hadoop 0 一月 11 16:34 tist ``` <a name="users-groups"></a> ## User and Group Management <a name="users"></a> ### Users [[ Back to Table of Contents ]](#toc) `users` shows all users currently logged in. Note that a user can be logged in multiple times if -- for instance -- they're connected via multiple `ssh` sessions. ``` \[ andrew@pc01 ~ \]$ 用戶 安德魯·科林·科林·科林·科林·科林·克里希納·克里希納 ``` To see all users (even those not logged in), check `/etc/passwd`. (**WARNING**: do not modify this file! You can corrupt your user accounts and make it impossible to log in to your system.) ``` \[ andrew@pc01 ~ \]$ alias au="cut -d: -f1 /etc/passwd \\ > |排序| uniq”&& au \_易於 一個廣告 安德魯... ``` Add a user with `useradd`: ``` \[ andrew@pc01 ~ \]$ sudo useradd aardvark && au \_易於 土豚 一個廣告... ``` Delete a user with `userdel`: ``` \[ andrew@pc01 ~ \]$ sudo userdel aardvark && au \_易於 一個廣告 安德魯... ``` [Change a user’s default shell, username, password, or group membership with `usermod`.](http://bit.ly/2D4upIg) <a name="groups"></a> ### Groups [[ Back to Table of Contents ]](#toc) `groups` shows all of the groups of which the current user is a member: ``` \[ andrew@pc01 ~ \]$ 組 andrew adm cdrom sudo dial plugdev lpadmin sambashare hadoop ``` To see all groups on the system, check `/etc/group`. (**DO NOT MODIFY** this file unless you know what you are doing.) ``` \[ andrew@pc01 ~ \]$ alias ag=“cut -d: -f1 /etc/group \\ > |排序”&& ag 管理員 一個廣告 安德魯... ``` Add a group with `groupadd`: ``` \[ andrew@pc01 ~ \]$ sudo groupadd aardvark && ag 土豚 管理員 一個廣告... ``` Delete a group with `groupdel`: ``` \[ andrew@pc01 ~ \]$ sudo groupdel aardvark && ag 管理員 一個廣告 安德魯... ``` [Change a group’s name, ID number, or password with `groupmod`.](https://linux.die.net/man/8/groupmod) <a name="text-processing"></a> ## Text Processing <a name="uniq-sort-diff-cmp"></a> ### `uniq / sort / diff / cmp` [[ Back to Table of Contents ]](#toc) `uniq` can print unique lines (default) or repeated lines: ``` \[ andrew@pc01 man \]$ printf "1\\n2\\n2" > a && \\ > printf "1\\n3\\n2" > b \[ andrew@pc01 人 \]$ uniq a 1 2 ``` `sort` will sort lines alphabetically / numerically: ``` \[ andrew@pc01 man \]$ 排序 b 1 2 3 ``` `diff` will report which lines differ between two files: ``` \[ andrew@pc01 人 \]$ diff ab 2c2 < 2 --- > 3 ``` `cmp` reports which bytes differ between two files: ``` \[andrew@pc01 人\]$ cmp ab ab 不同:字元 3,第 2 行 ``` <a name="cut-sed"></a> ### `cut / sed` [[ Back to Table of Contents ]](#toc) `cut` is usually used to cut a line into sections on some delimiter (good for CSV processing). `-d` specifies the delimiter and `-f` specifies the field index to print (starting with 1 for the first field): ``` \[ andrew@pc01 人 \]$ printf "137.99.234.23" > c \[ andrew@pc01 man \]$ cut -d'.' c-f1 137 ``` `sed` is commonly used to replace a string with another string in a file: ``` \[ andrew@pc01 man \]$ echo "舊" | sed s/舊/新/ 新的 ``` ...but `sed` is an extremely powerful utility, and cannot be properly summarised here. It’s actually Turing-complete, so it can do anything that any other programming language can do. `sed` can find and replace based on regular expressions, selectively print lines of a file which match or contain a certain pattern, edit text files in-place and non-interactively, and much more. A few good tutorials on `sed` include: - [https://www.tutorialspoint.com/sed/](https://www.tutorialspoint.com/sed/) - [http://www.grymoire.com/Unix/Sed.html](http://www.grymoire.com/Unix/Sed.html) - [https://www.computerhope.com/unix/used.htm](https://www.computerhope.com/unix/used.htm) <a name="pattern-matching"></a> ## Pattern Matching <a name="grep"></a> ### `grep` [[ Back to Table of Contents ]](#toc) The name `grep` comes from `g`/`re`/`p` (search `g`lobally for a `r`egular `e`xpression and `p`rint it); it’s used for finding text in files. `grep` is used to find lines of a file which match some pattern: ``` \[ andrew@pc01 ~ \]$ grep -e " *.fi.* " /etc/profile /etc/profile:Bourne shell 的系統範圍 .profile 檔案 (sh(1)) =================================================== ``` # The file bash.bashrc already sets the default PS1. ``` ``` fi ``` ``` fi ``` … ``` ...or contain some word: ``` \[ andrew@pc01 ~ \]$ grep "andrew" /etc/passwd 安德魯:x:1000:1000:安德魯,,,:/home/andrew:/bin/bash ``` `grep` is usually the go-to choice for simply finding matching lines in a file, if you’re planning on allowing some other program to handle those lines (or if you just want to view them). `grep` allows for (`-E`) use of extended regular expressions, (`-F`) matching any one of multiple strings at once, and (`-r`) recursively searching files within a directory. These flags used to be implemented as separate commands (`egrep`, `fgrep`, and `rgrep`, respectively), but those commands are now deprecated. > **Bonus**: [see the origins of the names of a few famous `bash` commands](https://kb.iu.edu/d/abnd) <a name="awk"></a> ### `awk` [[ Back to Table of Contents ]](#toc) `awk` is a pattern-matching language built around reading and manipulating delimited data files, like CSV files. As a rule of thumb, `grep` is good for finding strings and patterns in files, `sed` is good for one-to-one replacement of strings in files, and `awk` is good for extracting strings and patterns from files and analysing them. As an example of what `awk` can do, here’s a file containing two columns of data: ``` \[ andrew@pc01 ~ \]$ printf "A 10\\nB 20\\nC 60" > 文件 ``` Loop over the lines, add the number to sum, increment count, print the average: ``` \[ andrew@pc01 ~ \]$ awk 'BEGIN {sum=0;計數=0; OFS=“”} {sum+=$2; count++} END {print "平均值:", sum/count}' 文件 平均:30 ``` `sed` and `awk` are both Turing-complete languages. There have been multiple books written about each of them. They can be extremely useful with pattern matching and text processing. I really don’t have enough space here to do either of them justice. Go read more about them! > **Bonus**: [learn about some of the differences between `sed`, `grep`, and `awk`](http://bit.ly/2AI3IaN) <a name="ssh"></a> ## Copying Files Over `ssh` <a name="ssh-scp"></a> ### `ssh / scp` [[ Back to Table of Contents ]](#toc) `ssh` is how Unix-based machines connect to each other over a network: ``` \[ andrew@pc01 ~ \]$ ssh –p安德魯@137.xxx.xxx.89 上次登入:2019 年 1 月 11 日星期五 12:30:52,來自 137.xxx.xxx.199 ``` Notice that my prompt has changed as I’m now on a different machine: ``` \[ andrew@pc02 ~ \]$ 退出 登出 與 137.xxx.xxx.89 的連線已關閉。 ``` Create a file on machine 1: ``` \[ andrew@pc01 ~ \]$ echo "你好" > 你好 ``` Copy it to machine 2 using `scp` (secure copy; note that `scp` uses `–P` for a port #, `ssh` uses `–p`) ``` \[ andrew@pc01 ~ \]$ scp –P你好安德魯@137.xxx.xxx.89:~ 你好 100% 0 0.0KB/秒 00:00 ``` `ssh` into machine 2: ``` \[ andrew@pc02 ~ \]$ ssh –p安德魯@137.xxx.xxx.89 上次登入:2019 年 1 月 11 日星期五 22:47:37,來自 137.xxx.xxx.79 ``` The file’s there! ``` \[ andrew@pc02 ~ \]$ ls 你好多xargs \[ andrew@pc02 ~ \]$ 貓你好 你好 ``` <a name="rsync"></a> ### `rsync` [[ Back to Table of Contents ]](#toc) `rsync` is a file-copying tool which minimises the amount of data copied by looking for deltas (changes) between files. Suppose we have two directories: `d`, with one file, and `s`, with two files: ``` \[ andrew@pc01 d \]$ ls && ls ../s f0 f0 f1 ``` Sync the directories (copying only missing data) with `rsync`: ``` \[ andrew@pc01 d \]$ rsync -off ../s/\* . 正在發送增量文件列表... ``` `d` now contains all files that `s` contains: ``` \[ andrew@pc01 d \]$ ls f0 f1 ``` `rsync` can be performed over `ssh` as well: ``` \[ andrew@pc02 r \]$ ls \[ andrew@pc02 r \]$ rsync -avz -e "ssh -p “ [email protected]:~/s/\* 。 接收增量檔案列表 f0 f1 發送 62 位元組 接收 150 位元組 141.33 位元組/秒 總大小為 0 加速率為 0.00 \[ andrew@pc02 r \]$ ls f0 f1 ``` <a name="long-running-processes"></a> ## Long-Running Processes <a name="yes-nohup-ps-kill"></a> ### `yes / nohup / ps / kill` [[ Back to Table of Contents ]](#toc) Sometimes, `ssh` connections can disconnect due to network or hardware problems. Any processes initialized through that connection will be “hung up” and terminate. Running a command with `nohup` insures that the command will not be hung up if the shell is closed or if the network connection fails. Run `yes` (continually outputs "y" until it’s killed) with `nohup`: ``` \[ andrew@pc01 ~ \]$ nohup 是 & \[1\]13173 ``` `ps` shows a list of the current user’s processes (note PID number 13173): ``` \[ andrew@pc01 ~ \]$ ps | sed -n '/是/p' 13173 分/10 00:00:12 是 ``` _...log out and log back into this shell..._ The process has disappeared from `ps`! ``` \[ andrew@pc01 ~ \]$ ps | sed -n '/是/p' ``` But it still appears in `top` and `htop` output: ``` \[ andrew@pc01 ~ \]$ 頂部 -bn 1 | sed -n '/是/p' 13173 安德魯 20 0 4372 704 636 D 25.0 0.0 0:35.99 是 ``` Kill this process with `-9` followed by its process ID (PID) number: ``` \[ andrew@pc01 ~ \]$ 殺死 -9 13173 ``` It no longer appears in `top`, because it’s been killed: ``` \[ andrew@pc01 ~ \]$ 頂部 -bn 1 | sed -n '/是/p' ``` <a name="cron"></a> ### `cron / crontab / >>` [[ Back to Table of Contents ]](#toc) `cron` provides an easy way of automating regular, scheduled tasks. You can edit your `cron` jobs with `crontab –e` (opens a text editor). Append the line: ``` - - - - - 日期 >> ~/datefile.txt ``` This will run the `date` command every minute, appending (with the `>>` operator) the output to a file: ``` \[ andrew@pc02 ~ \]$ head ~/datefile.txt 2019 年 1 月 12 日星期六 14:37:01 GMT 2019 年 1 月 12 日星期六

🌌 31 個開源庫 + Good First Issues(開始你的旅程)⛰️

為優秀的開源庫做出貢獻是建立您的作品集並加入令人驚嘆的社群的最佳方式。 我編譯了 31 個開源程式庫和一些好的第一期,以幫助推動您的旅程。 不要忘記加星號並支持這些🌟 ![圖片描述](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/637xnt75fuwgfeaasdke.gif) --- #AI最愛🦾: ### 1. [CopilotKit](https://github.com/CopilotKit/CopilotKit) - 應用內 AI 聊天機器人與 AI 文字區域 ![圖片描述](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/ox3mv8nmqzot6m4kvkdh.png) 開源平台,用於使用兩個 React 元件將關鍵 AI 功能整合到 React 應用程式中。 CopilotPortal:應用程式內人工智慧聊天機器人,可以「查看」當前應用程式狀態並採取行動。 CopilotTextarea:AI 驅動的 <textarea /'> 替換。具有自動完成、插入和生成功能。 ###[好第一期:](https://github.com/CopilotKit/CopilotKit/issues/62) ``` Gracefully fail if CopilotProvider is omitted The bug: Virtually every CopilotKit functionality depends on a CopilotContext provided by the CopilotProvider. e.g. CopilotTextarea autocompletions, chatbot, etc. However when a CopilotProvider does not wrap the component, functionality fails silently. To Reproduce 1. Omit <CopilotProvider>...</CopilotProvider> 2. trigger useMakeCopilotReadable, useMakeCopilotActionable, CopilotTextarea, CopilotSidebarUIProvider 3. See how functionality does not work, but no error is emitted Expected behavior An error is emitted, with clear description of the likely core issue and how to resolve it (namely, wrap the app in a CopilotProvider). Point to docs. ``` {% cta https://github.com/CopilotKit/CopilotKit %} Star CopilotKit ⭐️ {% endcta %} --- ###2.[PortKeyAI](https://github.com/Portkey-AI/gateway){% embed https://github.com/Portkey-AI/gateway no-readme %} ###3.[Pezzo.ai](https://github.com/pezzolabs/pezzo){% 嵌入 https://github.com/pezzolabs/pezzo no-readme %} ###4.[OpenVoice](https://github.com/myshell-ai/OpenVoice){% 嵌入 https://github.com/myshell-ai/OpenVoice no-readme %} ###5.[LLMCourse](https://github.com/mlabonne/llm-course){% 嵌入 https://github.com/mlabonne/llm-course no-readme %} --- &nbsp; #雲端和資料庫☁️ ### 6. [Winglang](https://github.com/winglang/wing) - 雲端導向的程式語言 ![圖片描述](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/gvfykepsj1tszs8260wj.png) Wing 是一種用於雲端應用程式的程式語言。 它結合了雲端基礎設施和應用程式的程式碼,使雲端服務開發變得更加容易。 Wing 獨特的執行模型和測試模擬器有助於高效建置和部署雲端應用程式。 ###[第一期好:](https://github.com/winglang/wing/issues/4998) ``` Support Array.sort() method Feature Spec: let arr: MutArray<num>=[2, 1, 3, 9, 6, 4]; arr.sort(); log("${arr}"); // it should print sorted array in ascending order, eg: [1, 2, 3, 4, 6, 9] Component: Wing SDK Community Notes: If you are interested to work on this issue, please leave a comment. If this issue is labeled needs-discussion, it means the spec has not been finalized yet. Please reach out on the #dev channel in the Wing Slack. ``` {% cta https://github.com/winglang/wing %} 星翼朗 ⭐️ {% endcta %} --- ### 7. [StackQL](https://github.com/stackql/stackql) - 以 SQL 為基礎的雲端資源管理 ![圖片描述](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/sdtf51ekap09idn80xnh.png) StackQL 提供了一個獨特的 SQL 為基礎的框架來管理和查詢跨不同提供者(例如 Google、AWS、Azure 等)的雲端資源和 API。 它允許使用類似 SQL 的命令來配置和操作雲端服務,從而簡化了雲端操作。 這使得 StackQL 成為雲端資源管理和互動的多功能工具,特別是對於熟悉 SQL 的人來說。 ###[好第一期:](https://github.com/stackql/stackql/issues/280) ``` Add unit testing to package writer Add unit testing for internal/stackql/writer . Description: add implementation for testing sql_writer.go modify sql_writer.go by adding function for dependency injection add implementation for testing generic.go modify generic.go by adding variable for patching GetDB function modify entryutil.go to adjust sql_writer.go ``` {% cta https://github.com/stackql/stackql/ %} Star StackSQL ⭐️ {% endcta %} --- ###8.[Appwrite](https://github.com/appwrite/appwrite){% 嵌入 https://github.com/appwrite/appwrite no-readme %} ###9.[Supabase](https://github.com/supabase/supabase){% 嵌入 https://github.com/supabase/supabase no-readme %} ###10.[SuperDuperDB](https://github.com/SuperDuperDB/superduperdb){% 嵌入 https://github.com/SuperDuperDB/superduperdb no-readme %} --- &nbsp; #開發實用程式🛠️ ### 11. [Firecamp](https://github.com/firecamp-dev/firecamp) - 多協定 API 協作工具 ![圖片描述](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/adt6n8uv5dseylmemng0.png) Firecamp 是一款多功能 API 開發工具,支援 Rest、GraphQL 和 WebSockets 等協定。 它簡化了 API 的設計、測試和記錄,並增強了 API 專案的團隊協作。 ###[好第一期:](https://github.com/firecamp-dev/firecamp/issues/137) ``` Help out with Manual Testing of Firecamp Responsibilities: Executing test cases and reporting results Logging bugs and issues in the Github issue tracker Providing feedback on usability and the testing process Suggesting improvements to tests and expanding test coverage. Benefits Benefits for your testing profile and career: Experience testing a real-world open source application Each release will include your name with bugs fixes. Exposure to different types of testing such as UI, API, integration, etc Opportunity to have your contributions and feedback incorporated into the product Collaborating with an open source community Having your testing work visible to potential employers Firecamp Swags (T-shirts and stickers) Community shoutout and promotion ``` {% cta https://github.com/firecamp-dev/firecamp %} 星際火營 ⭐️ {% endcta %} --- ###12.[Odigos](https://github.com/keyval-dev/odigos){% 嵌入 https://github.com/keyval-dev/odigos no-readme %} ###13.[Digger](https://github.com/diggerhq/digger){% 嵌入 https://github.com/diggerhq/digger no-readme %} ###14.[鏡像](https://github.com/metalbear-co/mirrord){% 嵌入 https://github.com/metalbear-co/mirrord no-readme %} --- &nbsp; #後端⚙️ ### 15. [Cerbos](https://github.com/cerbos/cerbos) - 可擴充、與語言無關的授權 ![圖片描述](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/cljttnnxua54lyg4w65x.png) Cerbos 提供獨特、可擴展的解決方案,用於在應用程式中實施特定於上下文的使用者權限。 其靈活的、與語言無關的方法可以輕鬆整合和管理複雜的授權結構。 與眾不同的是,Cerbos 簡化了存取控制策略的開發,使其更能適應各種應用需求。 ###[第一期好:](https://github.com/cerbos/cerbos/issues/1920) ``` Produce output when the rule condition is not satisfied Currently the output block is only evaluated if the rule is actually activated (action, roles and conditions are satisfied). In certain situations, it's desirable to produce output when the rule is nearly activated (action and roles match but the condition is not satisfied). In order to maintain backward compatibility, reduce noise, and to keep policy execution as fast as possible (outputs incur a tiny overhead), the proposal is to let users add an optional when section to the output block to opt into this behaviour. - actions: ['view'] effect: EFFECT_ALLOW roles: ['user'] condition: match: expr: timestamp(R.expiry_date) > now() output: expr: > format("%d hours until expiry", (timestamp(R.expiry_date) - now()).getHours()) when: cond_fail: > format("expired on %s", R.expiry_date) When evaluating the above rule, if the action, roles and condition match, output will be the result of evaluating output.expr If the condition is not satisfied, output will be the result of evaluating output. when.cond_fail if it exists. Otherwise no output will be produced. ``` {% cta https://github.com/cerbos/cerbos %} 明星 Cerbos ⭐️ {% endcta %} --- ###16.[Novu](https://github.com/novuhq/novu){% 嵌入 https://github.com/novuhq/novu no-readme %} ###17.[Trigger.dev](https://github.com/triggerdotdev/trigger.dev){% 嵌入 https://github.com/triggerdotdev/trigger.dev no-readme %} ###18.[SuperTokens](https://github.com/supertokens/supertokens-core){% 嵌入 https://github.com/supertokens/supertokens-core no-readme %} ###19.[Wazuh](https://github.com/wazuh/wazuh){% 嵌入 https://github.com/wazuh/wazuh no-readme %} --- &nbsp; #UI/UX🦋: ### 20. [Flowbite](https://github.com/themesberg/flowbite) - 頂級 CSS 元件庫 ![圖片描述](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/98dwqyrhf1pbiqkpko8g.png) 最好、最受尊敬的 UI 元件庫之一。 基於實用優先的 CSS 框架。 易於使用,充滿重要的支援和模板。 {% cta https://github.com/themesberg/flowbite %} 明星 Flowbite ⭐️ {% endcta %} &nbsp; ###21.[MaterialUI](https://github.com/mui/material-ui) - 使用 Google 的 Material Design 實現的基礎 React 元件 {% 嵌入 https://github.com/mui/material-ui no-readme %} &nbsp; ###22。 [SwiperUI](https://github.com/nolimits4web/swiper) - 用於實現行動滑動 UI 的受人尊敬的庫 {% 嵌入 https://github.com/nolimits4web/swiper no-readme %} &nbsp; ###23.[ReactSpring](https://github.com/pmndrs/react-spring) - 在 React 中實現具有真實物理效果的動畫 {% 嵌入 https://github.com/pmndrs/react-spring no-readme %} --- &nbsp; #雜項🎨 ### 24. [SwirlSearch](https://github.com/swirlai/swirl-search) - 多源人工智慧資料搜尋器 ![圖片描述](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/thplxod3d4vh1qq5hhpa.jpeg) Swirl 是一款由人工智慧驅動的搜尋工具,可同時查詢多個資料來源,包括資料庫和公用資料服務。 它使用人工智慧對結果進行排名並產生見解,從而可以跨不同的資料儲存庫進行全面搜尋。 Swirl 一次簡化了對各種來源的資料的搜尋和分析,使其成為資料驅動洞察的獨特工具。 ###[第一期好:](https://github.com/swirlai/swirl-search/issues/789) ``` Add a Connector: Yahoo search It would help to search anything with Swirl on Yahoo effectively. Locate and read a bit in their search API first. You might just need to make a new SearchProvider configration vs. a new Connector. Their docs should help guide you a bit in which way you might need to go. ``` {% cta https://github.com/swirlai/swirl-search/ %} Star SwirlSearch ⭐️ {% endcta %} --- ### 25. [Wasp](https://github.com/wasp-lang/wasp) - 使用 React 和 Node.js 開發全端 Web 應用程式 ![圖片描述](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/54jp6j6r8ils6we97i0f.png) 使用 React 和 Node.js 進行快速全端 Web 應用程式開發。 Wasp 提供了一種建立現代 Web 應用程式的簡化方法,將前端的 React 和後端的 Node.js 結合在一個緊密結合的框架中。 ###[好第一期:](https://github.com/wasp-lang/wasp/issues/874) ``` Add images (or link to the example app) of auth UI helpers Wasp provides At this point in docs (also in the tutorial if we're using it), it would be nice to add an image of UI helpers for Auth (login/signup form, Google/GitHub button, ...) so developers can immediately see what they are getting and how nice it looks. ``` {% cta https://github.com/wasp-lang/wasp %} 星黃蜂 ⭐️ {% endcta %} ###26.[Logstash](https://github.com/elastic/logstash) {% 嵌入 https://github.com/elastic/logstash 無自述文件 %} ###27.[Snapify](https://github.com/MarconLP/snapify) {% 嵌入 https://github.com/MarconLP/snapify 無自述文件 %} --- &nbsp; #為了好玩🎭 ###28.[Twitter 的演算法](https://github.com/twitter/the-algorithm){% embed https://github.com/twitter/the-algorithm no-readme %} ###29.[十億行挑戰](https://github.com/gunnarmorling/1brc){% embed https://github.com/gunnarmorling/1brc no-readme %} ###30.【秘密知識之書】(https://github.com/trimstray/the-book-of-secret-knowledge){% embed https://github.com/trimstray/the-book-of -秘密知識無自述文件%} ###31.[GenAI 初學者](https://github.com/microsoft/generative-ai-for-beginners){% 嵌入 https://github.com/microsoft/generative-ai-for-beginners no -自述文件%} --- 原文出處:https://dev.to/copilotkit/31-open-source-libraries-to-kickstart-your-journey-4hhd

Supabase Auth:身分連結、Hooks 和 HaveIBeenPwned 集成

我們很高興地宣布 Supabase Auth 的四項新功能: 1. 身份連結 2. 會話控制 3. 密碼外洩保護 4. 帶有 Postgres 函數的 Auth Hooks {% 嵌入 https://youtu.be/LF8GABnAFyE %} ## 身份連結 當使用者登入時,系統會使用身份驗證方法和登入提供者建立身分。從歷史上看,如果身分與使用者共享相同的經過驗證的電子郵件,[Supabase Auth](https://supabase.com/docs/guides/auth) 會自動將身分連結到使用者。這可以方便地刪除重複的用戶帳戶。然而,一些開發人員還需要靈活地連結不共享相同電子郵件的帳戶。 今天,我們推出身份連結,開發人員可以使用它手動連結兩個單獨的身份。我們為開發人員新增了兩個新端點來管理身分連結流程: 使用者登入後,使用「linkIdentity()」[連結 OAuth 身分:](https://supabase.com/docs/reference/javascript/auth-linkidentity) ``` const { data, error } = await supabase.auth.linkIdentity({ provider: 'google', }) ``` 使用 `unlinkIdentity()` 來[取消連結身分](https://supabase.com/docs/reference/javascript/auth-unlinkidentity): ``` // retrieve all identities linked to a user const { data: { identities }, } = await supabase.auth.getUserIdentities() // find the google identity linked to the user const googleIdentity = identities.find(({ provider }) => provider === 'google') // unlink the google identity from the user const { data, error } = await supabase.auth.unlinkIdentity(googleIdentity) ``` 目前,這些方法支援連結 OAuth 身分。要將電子郵件或電話身分連結到用戶,您可以使用 [updateUser()](https://supabase.com/blog/supabase-auth-identity-linking-hooks#:~:text=can%20use% 20the -,updateUser(),-method.)方法。 預設情況下禁用手動連結。您可以在[儀表板驗證設定](https://supabase.com/dashboard/project/_/settings/auth) 中為您的專案啟用它。 ![如何啟用手動連結](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/kmsrmiw0ue3q5rshji7v.jpg) > 有關更多訊息,請參閱[身份連結文件](https://supabase.com/docs/guides/auth/auth-identity-linking)。 ## 會話控制 Supabase Auth 從使用者登入應用程式那一刻起管理整個會話生命週期。這涉及以下步驟: 1. 為使用者建立會話。 2. 刷新會話以使其保持活動狀態。 3. 過期或登出時撤銷會話。 對於想要更好地控制使用者會話的開發人員,我們公開了 3 個新設定: - **時間盒使用者會話:** 強制使用者在一段時間間隔後再次登入。 - **不活動逾時:** 如果使用者在一段時間內不活動,則強制使用者重新登入。 - **每個使用者單一會話:** 將使用者限制為單一會話。保留最近的活動會話,並終止所有其他會話。 這些會話控制設定在專業版及以上版本中可用。 ![如何強制每個使用者單一會話](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/aov0nw5xch0m4hw97vsg.jpg) > 有關更多訊息,請參閱[會話管理文件](https://supabase.com/docs/guides/auth/sessions)。 ## 密碼外洩保護 由於常見的使用者行為(例如選擇可猜測的密碼或在不同平台上重複使用密碼),密碼本質上可能是不安全的。 儘管 OAuth 和 magiclinks 更安全,但我們認識到密碼將繼續存在。我們希望讓用戶不易陷入潛在的陷阱。為了實現這一目標,我們在 Supabase Auth 中整合了 [HaveIBeenPwned.org](https://haveibeenpwned.com/) _Pwned Passwords API_,以防止使用者使用洩漏的密碼。 > **去圖書館** ℹ️ 我們開源了一個 Go 函式庫,用於與我們在身分驗證伺服器中使用的 [HaveIBeenPwned.org](http://haveibeenpwned.org/) Pwned 密碼 API 互動。查看 [存儲庫](https://github.com/supabase/hibp) 並隨時貢獻! 作為附加步驟,我們新增了為您的使用者指定密碼要求的功能。這可以透過[儀表板:](https://supabase.com/dashboard/project/_/settings/auth) 中專案的身份驗證設定進行配置 ![新增密碼要求](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/3o9ax3tudp7v8tba67hd.jpg) > 請參閱[密碼文件](https://supabase.com/docs/guides/auth/passwords) 以了解更多資訊。 ## 驗證掛鉤 我們收到了大量回饋,詢問如何自訂 Auth,例如: - 將自訂聲明新增至存取權杖 JWT - 多次嘗試 MFA 驗證失敗後註銷用戶 - 對密碼驗證嘗試套用自訂規則 我們的目標是保持簡單、無縫的 Supabase Auth 體驗。對於大多數開發人員來說,它應該可以輕鬆工作,而無需自訂。但是,認識到應用程式的多樣性,您現在可以透過 Auth Hook 擴展標準 Auth 功能。 Auth Hooks 只是 Postgres 函數,它們在 Auth 生命週期的關鍵點同步執行,以更改操作的結果。 例如,要使用 Auth Hooks 自訂 JWT 聲明,您可以建立一個 Postgres 函數,該函數接受第一個參數中的 JWT 聲明並傳回您希望 Supabase Auth 使用的 JWT。 假設您正在建立一個遊戲化應用程式,並且希望將用戶的層級作為自訂聲明附加到 JWT: ``` create function custom_access_token_hook(event jsonb) returns jsonb language plpgsql as $$ declare user_level jsonb; begin -- fetch the current user's level select to_jsonb(level) into user_level from profiles where user_id = event->>'user_id'::uuid; -- change the event.claims.level return jsonb_set( event, '{claims,level}', user_level); end; $$ ``` 在資料庫中建立函數後,您只需使用 Supabase Auth 註冊它: ![Auth Hooks](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/unkgs56o62l8c4kfjzjg.jpg) 目前,您可以為流程中的以下點註冊 Auth Hook: - **自訂存取權杖:** 每次產生新的 JWT 時都會呼叫。 - **MFA 驗證嘗試:** 每次驗證 MFA 因素時都會呼叫,從而可以更好地控制檢測和阻止嘗試。 - **密碼驗證嘗試:** 每次使用密碼登入使用者時都會呼叫,從而可以更好地控制使用者帳戶的安全性。 如果編寫 PL/pgSQL 函數不是您的強項,您始終可以使用 [pg_net](https://supabase.com/docs/guides/database/extensions/pg_net) 向後端 API 發送請求,或使用[plv8]( https://supabase.com/docs/guides/database/extensions/plv8) 透過用JavaScript 編寫函數來更輕鬆地操作JSON。 Auth Hooks 今天可供自架,並將於下個月推出到該平台。如果您需要盡快存取,請透過[支援](https://supabase.help/)與我們聯繫! 那不是全部! Postgres 函數並不是寫鉤子的唯一方法。 Supabase 是 [Standard Webhooks](https://www.standardwebhooks.com/) 的創始貢獻者,這是一組關於輕鬆、安全、可靠地發送和接收 Webhook 的開源工具和指南。當然,Auth Hooks 將在 2024 年第一季支援 Webhooks。 ## 還有一件事… 如果您從一開始就關注我們(https://supabase.com/blog/supabase-auth),您就會知道Supabase Auth 是透過分叉[Netlify 的GoTrue 伺服器](https://github.com)開始的/netlify/gotrue)。從那時起,發生了很多變化,我們已經偏離了上游儲存庫。在這個階段,將專案重新命名為其他名稱是有意義的(提示鼓聲)-Auth。 這僅僅意味著儲存庫將從使用“gotrue”重新命名為“auth”。但別擔心! Docker 映像和庫(如“@supabase/gotrue-js”)將繼續發布,只要當前 v2 版本受支持,您就可以互換使用“@supabase/auth-js”。所有類別和方法都保持不變。這裡沒有重大變化! ## 結論 感謝您閱讀到最後!我們希望您喜歡第 X 週發布的 Supabase Auth 更新:身分連結、會話控制、洩露密碼保護和帶有 Postgres 功能的 Auth Hooks。 我們期待看到您使用這些新功能建立的內容,當然還有您的回饋意見,以使它們變得更好。 ## 更多發布第 X 週 - [第 1 天 - Supabase Studio 更新:AI 助理與使用者模擬](https://supabase.com/blog/studio-introducing-assistant) - [第 2 天 - Edge Functions:節點和本機 npm 相容性 ](https://supabase.com/blog/edge-functions-node-npm) -[第 3 天 - 介紹 Supabase Branching,這是一個針對每個拉取請求的 Postgres 資料庫](https://supabase.com/blog/supabase-branching) - [Postgres語言伺服器:實作解析器](https://supabase.com/blog/postgres-language-server-implementing-parser) - [Supabase 專輯](https://www.youtube.com/watch?v=r1POD-IdG-I) - [Supabase 啟動週 X 黑客松](https://supabase.com/blog/supabase-hackathon-lwx) - [啟動週 X 社群聚會](https://supabase.com/blog/community-meetups-lwx) --- 原文出處:https://dev.to/supabase/supabase-auth-identity-linking-hooks-and-haveibeenpwned-integration-19e1

🧙‍♂️ 使用 ChatGPT 助理產生部落格 🪄 ✨

# 長話短說;博士 我們都已經看到了 ChatGPT 的功能(這對任何人來說都不陌生)。 很多文章都是使用 ChatGPT 一遍又一遍地寫的。 **實際上**,DEV 上的文章有一半是用 ChatGPT 寫的。 你可以使用一些[AI內容偵測器](https://copyleaks.com/ai-content- detector)來檢視。 問題是,ChatGPT 永遠不會產生一些非凡的內容,除了它內部已經有(經過訓練/微調)的內容。 但有一種方法可以超越目前使用 RAG(OpenAI 助理)訓練的內容。 [上一篇](https://dev.to/triggerdotdev/train-chatgpt-on-your-documentation-1a9g),我們討論了在您的文件上「訓練」ChatGPT;今天,讓我們看看如何從中製作出很多內容。我們將: - 使用 Docusaurus 建立新的部落格系統。 - 詢問 ChatGPT,為我們寫一篇與文件相關的部落格文章。 ![部落格](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/ms26qb0uahpi898s0qun.gif) --- ## 你的後台工作平台🔌 [Trigger.dev](https://trigger.dev/) 是一個開源程式庫,可讓您使用 NextJS、Remix、Astro 等為您的應用程式建立和監控長時間執行的作業! &nbsp; [![GiveUsStars](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/bm9mrmovmn26izyik95z.gif)](https://github.com/triggerdotdev/trigger.dev) 請幫我們一顆星🥹。 這將幫助我們建立更多這樣的文章💖 {% cta https://github.com/triggerdotdev/trigger.dev %} 為 Trigger.dev 儲存庫加註星標 ⭐️ {% endcta %} --- ## 上次回顧 ⏰ - 我們建立了一個作業來取得文件 XML 並提取所有 URL。 - 我們抓取了每個網站的 URL 並提取了標題和內容。 - 我們將所有內容儲存到文件中並將其發送給 ChatGPT 助手。 - 我們建立了一個 ChatBot 畫面來詢問 ChatGPT 有關文件的資訊。 您可以在此處找到上一個[教學]的完整原始程式碼(https://github.com/triggerdotdev/blog/tree/main/openai-assistant)。 --- ![工具](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/i4adju83b5s1k0qozh3x.png) ## 稍作修改⚙️ 上次,我們建立了一個文件助理。我們寫: ``` You are a documentation assistant, loaded with documentation from ' + payload.url + ', return everything in an MD format. ``` 讓我們將其更改為部落格作者,請轉到“jobs/process.documentation.ts”第 92 行,並將其替換為以下內容: ``` You are a content writer assistant. You have been loaded with documentation from ${payload.url}, you write blog posts based on the documentation and return everything in the following MD format: --- slug: [post-slug] title: [post-title] --- [post-content] ``` 使用“slug”和“title”非常重要,因為這是 Docusaurus 的格式 - 我們的部落格系統可以接受(當然,我們也以 MD 格式發送所有輸出) --- ![Docusaurus](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/gu8wlh7qk8e3rh6mz35v.png) ## 多庫龍🦖 您可以使用多種類型的部落格系統! 對於我們的用例,我們將使用 Docusaurus,它可以讀取基於 MD 的格式(我們從 ChatGPT 請求的輸出)。 **我們可以透過執行來安裝 Docusaurus:** ``` npx create-docusaurus@latest blog classic --typescript ``` 接下來,我們可以進入已建立的目錄並執行以下命令: ``` npm run start ``` 這將啟動 Docusaurus。你可以關註一下。還有一個名為“blog”的附加目錄,其中包含所有部落格文章;這是我們保存 ChatGPT 產生的部落格文章的地方。 ![範例](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/pgo25rlkw85nfvbh0y4s.png) --- ![部落格](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/v3oxjtli1dn9i9klnj5t.png) ## 產生部落格 📨 我們需要創造一個就業機會 - 取得部落格標題 - 使用 ChatGPT 產生完整的部落格文章 - 將其保存到我們部落格上的 MD 文件中 我們可以輕鬆地使用 ChatGPT 來實現這一點! 前往“jobs”資料夾並新增一個名為“process.blog.ts”的新檔案。新增以下程式碼: ``` import { eventTrigger } from "@trigger.dev/sdk"; import { client } from "@openai-assistant/trigger"; import {object, string} from "zod"; import {openai} from "@openai-assistant/helper/open.ai"; import {writeFileSync} from "fs"; import slugify from "slugify"; client.defineJob({ // This is the unique identifier for your Job, it must be unique across all Jobs in your project. id: "process-blog", name: "Process Blog", version: "0.0.1", // This is triggered by an event using eventTrigger. You can also trigger Jobs with webhooks, on schedules, and more: https://trigger.dev/docs/documentation/concepts/triggers/introduction trigger: eventTrigger({ name: "process.blog.event", schema: object({ title: string(), aId: string(), }) }), integrations: { openai }, run: async (payload, io, ctx) => { const {title, aId} = payload; const thread = await io.openai.beta.threads.create('create-thread'); await io.openai.beta.threads.messages.create('create-message', thread.id, { content: ` title: ${title} `, role: 'user', }); const run = await io.openai.beta.threads.runs.createAndWaitForCompletion('run-thread', thread.id, { model: 'gpt-4-1106-preview', assistant_id: payload.aId, }); if (run.status !== "completed") { console.log('not completed'); throw new Error(`Run finished with status ${run.status}: ${JSON.stringify(run.last_error)}`); } const messages = await io.openai.beta.threads.messages.list("list-messages", run.thread_id, { query: { limit: "1" } }); return io.runTask('save-blog', async () => { const content = messages[0].content[0]; if (content.type === 'text') { const fileName = slugify(title, {lower: true, strict: true, trim: true}); writeFileSync(`./blog/blog/${fileName}.md`, content.text.value) return {fileName}; } }); }, }); ``` - 我們加入了一些必要的變數: - `title` 部落格文章標題 - `aId` 上一篇文章中新增的助手 ID。 - 我們為助手建立了一個新線程(`io.openai.beta.threads.create`) - 我們無法在沒有任何線程的情況下質疑它。與之前的教程不同,在這裡,我們對每個請求建立一個新線程。我們不需要對話中最後一條訊息的上下文。 - 然後,我們使用部落格標題為線程(`io.openai.beta.threads.messages.create`)新增訊息。我們不需要提供額外的說明 - 我們已經在第一部分完成了該部分😀 - 我們執行 `io.openai.beta.threads.runs.createAndWaitForCompletion` 來啟動進程 - 通常,您需要某種每分鐘執行一次的遞歸來檢查作業是否完成,但是 [Trigger.dev]( http://Trigger .dev)已經加入了一種執行進程並同時等待它的方法🥳 - 我們在查詢正文中執行帶有“limit: 1”的“io.openai.beta.threads.messages.list”,以從對話中獲取第一則訊息(在ChatGPT 結果中,第一則訊息是最後一條訊息) 。 - 然後,我們使用「writeFileSync」從 ChatGPT 取得的值來儲存新建立的部落格 - 確保您擁有正確的部落格路徑。 轉到“jobs/index.ts”並加入以下行: ``` export * from "./process.blog"; ``` 現在,讓我們建立一個新的路由來觸發該作業。 前往“app/api”,建立一個名為“blog”的新資料夾,並在一個名為“route.tsx”的新檔案中 新增以下程式碼: ``` import {client} from "@openai-assistant/trigger"; export async function POST(request: Request) { const payload = await request.json(); if (!payload.title || !payload.aId) { return new Response(JSON.stringify({error: 'Missing parameters'}), {status: 400}); } // We send an event to the trigger to process the documentation const {id: eventId} = await client.sendEvent({ name: "process.blog.event", payload }); return new Response(JSON.stringify({eventId}), {status: 200}); } ``` - 我們檢查標題和助理 ID 是否存在。 - 我們在 [Trigger.dev](http://Trigger.dev) 中觸發事件並發送訊息。 - 我們將事件 ID 傳送回客戶端,以便我們可以追蹤作業的進度。 --- ![前端](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/kgh52s7mxd20w91kr3c9.png) ## 前端🎩 沒什麼好做的! 在我們的「components」目錄中,建立一個名為「blog.component.tsx」的新檔案和以下程式碼: ``` "use client"; import {FC, useCallback, useEffect, useState} from "react"; import {ExtendedAssistant} from "@openai-assistant/components/main"; import {SubmitHandler, useForm} from "react-hook-form"; import {useEventRunDetails} from "@trigger.dev/react"; interface Blog { title: string, aId: string; } export const BlogComponent: FC<{list: ExtendedAssistant[]}> = (props) => { const {list} = props; const {register, formState, handleSubmit} = useForm<Blog>(); const [event, setEvent] = useState<string | undefined>(undefined); const addBlog: SubmitHandler<Blog> = useCallback(async (param) => { const {eventId} = await (await fetch('/api/blog', { method: 'POST', headers: { 'Content-Type': 'application/json' }, body: JSON.stringify(param) })).json(); setEvent(eventId); }, []); return ( <> <form className="flex flex-col gap-3 mt-5" onSubmit={handleSubmit(addBlog)}> <div className="flex flex-col gap-1"> <div className="font-bold">Assistant</div> <select className="border border-gray-200 rounded-xl py-2 px-3" {...register('aId', {required: true})}> {list.map(val => ( <option key={val.id} value={val.aId}>{val.url}</option> ))} </select> </div> <div className="flex flex-col gap-1"> <div className="font-bold">Title</div> <input className="border border-gray-200 rounded-xl py-2 px-3" placeholder="Blog title" {...register('title', {required: true})} /> </div> <button className="border border-gray-200 rounded-xl py-2 px-3 bg-gray-100 hover:bg-gray-200" disabled={formState.isSubmitting}>Create blog</button> </form> {!!event && ( <Blog eventId={event} /> )} </> ) } export const Blog: FC<{eventId: string}> = (props) => { const {eventId} = props; const { data, error } = useEventRunDetails(eventId); if (data?.status !== 'SUCCESS') { return <div className="pointer bg-yellow-300 border-yellow-500 p-1 px-3 text-yellow-950 border rounded-2xl">Loading</div> } return ( <div> <a href={`http://localhost:3000/blog/${data.output.fileName}`}>Check blog post</a> </div> ) }; ``` - 我們使用「react-hook-form」來輕鬆控制我們的輸入。 - 我們讓使用者選擇他們想要使用的助手。 - 我們建立一個包含文章標題的新輸入。 - 我們將所有內容傳送到先前建立的路由並傳回作業的「eventId」。 - 我們建立一個新的「<Blog />」元件,該元件顯示載入直到事件完成,並使用新建立的教程新增指向我們部落格的連結。 將元件加入我們的“components/main.tsx”檔案中: ``` {assistantState.filter(f => !f.pending).length > 0 && <BlogComponent list={assistantState} />} ``` 我們完成了! ![完成](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/fkm37v5idrxexjje2u3o.png) 現在,讓我們新增部落格標題並點擊「生成」。 ![部落格](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/gosm1f1ttz3q1m0atu7s.png) --- ![圖片](https://res.cloudinary.com/practicaldev/image/fetch/s--uTFwMeAp--/c_limit%2Cf_auto%2Cfl_progressive%2Cq_auto%2Cw_800/https://dev-to-uploads.s3。 amazonaws.com/uploads/articles/0half2g6r5zfn7asq084.png) ## 讓我們聯絡吧! 🔌 作為開源開發者,您可以加入我們的[社群](https://discord.gg/nkqV9xBYWy) 做出貢獻並與維護者互動。請隨時造訪我們的 [GitHub 儲存庫](https://github.com/triggerdotdev/trigger.dev),貢獻並建立與 Trigger.dev 相關的問題。 本教學的源程式碼可在此處取得: https://github.com/triggerdotdev/blog/tree/main/openai-blog-writer 感謝您的閱讀! --- 原文出處:https://dev.to/triggerdotdev/generate-blogs-with-chatgpt-assistant-1894

✨ 用您的文件訓練 ChatGPT 🪄 ✨

# 簡介 ChatGPT 訓練至 2022 年。 但是,如果您希望它專門為您提供有關您網站的資訊怎麼辦?最有可能的是,這是不可能的,**但不再是了!** OpenAI 推出了他們的新功能 - [助手](https://platform.openai.com/docs/assistants/how-it-works)。 現在您可以輕鬆地為您的網站建立索引,然後向 ChatGPT 詢問有關該網站的問題。在本教程中,我們將建立一個系統來索引您的網站並讓您查詢它。我們將: - 抓取文件網站地圖。 - 從網站上的所有頁面中提取資訊。 - 使用新資訊建立新助理。 - 建立一個簡單的ChatGPT前端介面並查詢助手。 ![助手](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/ekre38der95twom33tqb.gif) --- ## 你的後台工作平台🔌 [Trigger.dev](https://trigger.dev/) 是一個開源程式庫,可讓您使用 NextJS、Remix、Astro 等為您的應用程式建立和監控長時間執行的作業!   [![GiveUsStars](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/bm9mrmovmn26izyik95z.gif)](https://github.com/triggerdotdev/trigger.dev) 請幫我們一顆星🥹。 這將幫助我們建立更多這樣的文章💖 --- ## 讓我們開始吧🔥 讓我們建立一個新的 NextJS 專案。 ``` npx create-next-app@latest ``` >💡 我們使用 NextJS 新的應用程式路由器。安裝專案之前請確保您的節點版本為 18+ 讓我們建立一個新的資料庫來保存助手和抓取的頁面。 對於我們的範例,我們將使用 [Prisma](https://www.prisma.io/) 和 SQLite。 安裝非常簡單,只需執行: ``` npm install prisma @prisma/client --save ``` 然後加入架構和資料庫 ``` npx prisma init --datasource-provider sqlite ``` 轉到“prisma/schema.prisma”並將其替換為以下架構: ``` // This is your Prisma schema file, // learn more about it in the docs: https://pris.ly/d/prisma-schema generator client { provider = "prisma-client-js" } datasource db { provider = "sqlite" url = env("DATABASE_URL") } model Docs { id Int @id @default(autoincrement()) content String url String @unique identifier String @@index([identifier]) } model Assistant { id Int @id @default(autoincrement()) aId String url String @unique } ``` 然後執行 ``` npx prisma db push ``` 這將建立一個新的 SQLite 資料庫(本機檔案),其中包含兩個主表:“Docs”和“Assistant” - 「Docs」包含所有抓取的頁面 - `Assistant` 包含文件的 URL 和內部 ChatGPT 助理 ID。 讓我們新增 Prisma 客戶端。 建立一個名為「helper」的新資料夾,並新增一個名為「prisma.ts」的新文件,並在其中新增以下程式碼: ``` import {PrismaClient} from '@prisma/client'; export const prisma = new PrismaClient(); ``` 我們稍後可以使用“prisma”變數來查詢我們的資料庫。 --- ![ScrapeAndIndex](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/fc05wtlc4peosr62ydnx.png) ## 刮擦和索引 ### 建立 Trigger.dev 帳戶 抓取頁面並為其建立索引是一項長期執行的任務。 **我們需要:** - 抓取網站地圖的主網站元 URL。 - 擷取網站地圖內的所有頁面。 - 前往每個頁面並提取內容。 - 將所有內容儲存到 ChatGPT 助手中。 為此,我們使用 Trigger.dev! 註冊 [Trigger.dev 帳號](https://trigger.dev/)。 註冊後,建立一個組織並為您的工作選擇一個專案名稱。 ![pic1](https://res.cloudinary.com/practicaldev/image/fetch/s--B2jtIoA6--/c_limit%2Cf_auto%2Cfl_progressive%2Cq_auto%2Cw_800/https://dev-to-uploads.s3.amazonaws.com/uploads/articles/bdnxq8o7el7t4utvgf1u.jpeg) 選擇 Next.js 作為您的框架,並按照將 Trigger.dev 新增至現有 Next.js 專案的流程進行操作。 ![pic2](https://res.cloudinary.com/practicaldev/image/fetch/s--K4k6T6mi--/c_limit%2Cf_auto%2Cfl_progressive%2Cq_auto%2Cw_800/https://dev-to-uploads.s3.amazonaws.com/uploads/articles/e4kt7e5r1mwg60atqfka.jpeg) 否則,請點選專案儀表板側邊欄選單上的「環境和 API 金鑰」。 ![pic3](https://res.cloudinary.com/practicaldev/image/fetch/s--Ysm1Dd0r--/c_limit%2Cf_auto%2Cfl_progressive%2Cq_auto%2Cw_800/https://dev-to-uploads.s3.amazonaws.com/uploads/articles/ser7a2j5qft9vw8rfk0m.png) 複製您的 DEV 伺服器 API 金鑰並執行下面的程式碼片段來安裝 Trigger.dev。 仔細按照說明進行操作。 ``` npx @trigger.dev/cli@latest init ``` 在另一個終端中執行以下程式碼片段,在 Trigger.dev 和您的 Next.js 專案之間建立隧道。 ``` npx @trigger.dev/cli@latest dev ``` ### 安裝 ChatGPT (OpenAI) 我們將使用OpenAI助手,因此我們必須將其安裝到我們的專案中。 [建立新的 OpenAI 帳戶](https://platform.openai.com/) 並產生 API 金鑰。 ![pic4](https://res.cloudinary.com/practicaldev/image/fetch/s--uV1LwOH---/c_limit%2Cf_auto%2Cfl_progressive%2Cq_auto%2Cw_800/https://dev-to-uploads.s3.amazonaws.com/uploads/articles/ashau6i2sxcpd0qcxuwq.png) 點擊下拉清單中的「檢視 API 金鑰」以建立 API 金鑰。 ![pic5](https://res.cloudinary.com/practicaldev/image/fetch/s--Tp8aLqSa--/c_limit%2Cf_auto%2Cfl_progressive%2Cq_auto%2Cw_800/https://dev-to-uploads.s3.amazonaws.com/uploads/articles/4bzc6e7f7avemeuuaygr.png) 接下來,透過執行下面的程式碼片段來安裝 OpenAI 套件。 ``` npm install @trigger.dev/openai ``` 將您的 OpenAI API 金鑰新增至「.env.local」檔案。 ``` OPENAI_API_KEY=<your_api_key> ``` 建立一個新目錄“helper”並新增一個新檔案“open.ai.tsx”,其中包含以下內容: ``` import {OpenAI} from "@trigger.dev/openai"; export const openai = new OpenAI({ id: "openai", apiKey: process.env.OPENAI_API_KEY!, }); ``` 這是我們透過 Trigger.dev 整合封裝的 OpenAI 用戶端。 ### 建立後台作業 讓我們繼續建立一個新的後台作業! 前往“jobs”並建立一個名為“process.documentation.ts”的新檔案。 **新增以下程式碼:** ``` import { eventTrigger } from "@trigger.dev/sdk"; import { client } from "@openai-assistant/trigger"; import {object, string} from "zod"; import {JSDOM} from "jsdom"; import {openai} from "@openai-assistant/helper/open.ai"; client.defineJob({ // This is the unique identifier for your Job; it must be unique across all Jobs in your project. id: "process-documentation", name: "Process Documentation", version: "0.0.1", // This is triggered by an event using eventTrigger. You can also trigger Jobs with webhooks, on schedules, and more: https://trigger.dev/docs/documentation/concepts/triggers/introduction trigger: eventTrigger({ name: "process.documentation.event", schema: object({ url: string(), }) }), integrations: { openai }, run: async (payload, io, ctx) => { } }); ``` 我們定義了一個名為「process.documentation.event」的新作業,並新增了一個名為 URL 的必要參數 - 這是我們稍後要傳送的文件 URL。 正如您所看到的,該作業是空的,所以讓我們向其中加入第一個任務。 我們需要獲取網站網站地圖並將其返回。 抓取網站將返回我們需要解析的 HTML。 為此,我們需要安裝 JSDOM。 ``` npm install jsdom --save ``` 並將其導入到我們文件的頂部: ``` import {JSDOM} from "jsdom"; ``` 現在,我們可以新增第一個任務。 用「runTask」包裝我們的程式碼很重要,這可以讓 Trigger.dev 將其與其他任務分開。觸發特殊架構將任務拆分為不同的進程,因此 Vercel 無伺服器逾時不會影響它們。 **這是第一個任務的程式碼:** ``` const getSiteMap = await io.runTask("grab-sitemap", async () => { const data = await (await fetch(payload.url)).text(); const dom = new JSDOM(data); const sitemap = dom.window.document.querySelector('[rel="sitemap"]')?.getAttribute('href'); return new URL(sitemap!, payload.url).toString(); }); ``` - 我們透過 HTTP 請求從 URL 取得整個 HTML。 - 我們將其轉換為 JS 物件。 - 我們找到網站地圖 URL。 - 我們解析它並返回它。 接下來,我們需要抓取網站地圖,提取所有 URL 並返回它們。 讓我們安裝“Lodash”——陣列結構的特殊函數。 ``` npm install lodash @types/lodash --save ``` 這是任務的程式碼: ``` export const makeId = (length: number) => { let text = ''; const possible = 'ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789'; for (let i = 0; i < length; i += 1) { text += possible.charAt(Math.floor(Math.random() * possible.length)); } return text; }; const {identifier, list} = await io.runTask("load-and-parse-sitemap", async () => { const urls = /(http|ftp|https):\/\/([\w_-]+(?:(?:\.[\w_-]+)+))([\w.,@?^=%&:\/~+#-]*[\w@?^=%&\/~+#-])/g; const identifier = makeId(5); const data = await (await fetch(getSiteMap)).text(); // @ts-ignore return {identifier, list: chunk(([...new Set(data.match(urls))] as string[]).filter(f => f.includes(payload.url)).map(p => ({identifier, url: p})), 25)}; }); ``` - 我們建立一個名為 makeId 的新函數來為所有頁面產生隨機辨識碼。 - 我們建立一個新任務並加入正規表示式來提取每個可能的 URL - 我們發送一個 HTTP 請求來載入網站地圖並提取其所有 URL。 - 我們將 URL「分塊」為 25 個元素的陣列(如果有 100 個元素,則會有四個 25 個元素的陣列) 接下來,讓我們建立一個新作業來處理每個 URL。 **這是完整的程式碼:** ``` function getElementsBetween(startElement: Element, endElement: Element) { let currentElement = startElement; const elements = []; // Traverse the DOM until the endElement is reached while (currentElement && currentElement !== endElement) { currentElement = currentElement.nextElementSibling!; // If there's no next sibling, go up a level and continue if (!currentElement) { // @ts-ignore currentElement = startElement.parentNode!; startElement = currentElement; if (currentElement === endElement) break; continue; } // Add the current element to the list if (currentElement && currentElement !== endElement) { elements.push(currentElement); } } return elements; } const processContent = client.defineJob({ // This is the unique identifier for your Job; it must be unique across all Jobs in your project. id: "process-content", name: "Process Content", version: "0.0.1", // This is triggered by an event using eventTrigger. You can also trigger Jobs with webhooks, on schedules, and more: https://trigger.dev/docs/documentation/concepts/triggers/introduction trigger: eventTrigger({ name: "process.content.event", schema: object({ url: string(), identifier: string(), }) }), run: async (payload, io, ctx) => { return io.runTask('grab-content', async () => { // We first grab a raw html of the content from the website const data = await (await fetch(payload.url)).text(); // We load it with JSDOM so we can manipulate it const dom = new JSDOM(data); // We remove all the scripts and styles from the page dom.window.document.querySelectorAll('script, style').forEach((el) => el.remove()); // We grab all the titles from the page const content = Array.from(dom.window.document.querySelectorAll('h1, h2, h3, h4, h5, h6')); // We grab the last element so we can get the content between the last element and the next element const lastElement = content[content.length - 1]?.parentElement?.nextElementSibling!; const elements = []; // We loop through all the elements and grab the content between each title for (let i = 0; i < content.length; i++) { const element = content[i]; const nextElement = content?.[i + 1] || lastElement; const elementsBetween = getElementsBetween(element, nextElement); elements.push({ title: element.textContent, content: elementsBetween.map((el) => el.textContent).join('\n') }); } // We create a raw text format of all the content const page = ` ---------------------------------- url: ${payload.url}\n ${elements.map((el) => `${el.title}\n${el.content}`).join('\n')} ---------------------------------- `; // We save it to our database await prisma.docs.upsert({ where: { url: payload.url }, update: { content: page, identifier: payload.identifier }, create: { url: payload.url, content: page, identifier: payload.identifier } }); }); }, }); ``` - 我們從 URL 中獲取內容(之前從網站地圖中提取) - 我們用`JSDOM`解析它 - 我們刪除頁面上存在的所有可能的“<script>”或“<style>”。 - 我們抓取頁面上的所有標題(`h1`、`h2`、`h3`、`h4`、`h5`、`h6`) - 我們迭代標題並獲取它們之間的內容。我們不想取得整個頁面內容,因為它可能包含不相關的內容。 - 我們建立頁面原始文字的版本並將其保存到我們的資料庫中。 現在,讓我們為每個網站地圖 URL 執行此任務。 觸發器引入了名為“batchInvokeAndWaitForCompletion”的東西。 它允許我們批量發送 25 個專案進行處理,並且它將同時處理所有這些專案。下面是接下來的幾行程式碼: ``` let i = 0; for (const item of list) { await processContent.batchInvokeAndWaitForCompletion( 'process-list-' + i, item.map( payload => ({ payload, }), 86_400), ); i++; } ``` 我們以 25 個為一組[手動觸發](https://trigger.dev/docs/documentation/concepts/triggers/invoke)之前建立的作業。 完成後,讓我們將保存到資料庫的所有內容並連接它: ``` const data = await io.runTask("get-extracted-data", async () => { return (await prisma.docs.findMany({ where: { identifier }, select: { content: true } })).map((d) => d.content).join('\n\n'); }); ``` 我們使用之前指定的標識符。 現在,讓我們在 ChatGPT 中使用新資料建立一個新檔案: ``` const file = await io.openai.files.createAndWaitForProcessing("upload-file", { purpose: "assistants", file: data }); ``` `createAndWaitForProcessing` 是 Trigger.dev 建立的任務,用於將檔案上傳到助手。如果您在沒有整合的情況下手動使用“openai”,則必須串流傳輸檔案。 現在讓我們建立或更新我們的助手: ``` const assistant = await io.openai.runTask("create-or-update-assistant", async (openai) => { const currentAssistant = await prisma.assistant.findFirst({ where: { url: payload.url } }); if (currentAssistant) { return openai.beta.assistants.update(currentAssistant.aId, { file_ids: [file.id] }); } return openai.beta.assistants.create({ name: identifier, description: 'Documentation', instructions: 'You are a documentation assistant, you have been loaded with documentation from ' + payload.url + ', return everything in an MD format.', model: 'gpt-4-1106-preview', tools: [{ type: "code_interpreter" }, {type: 'retrieval'}], file_ids: [file.id], }); }); ``` - 我們首先檢查是否有針對該特定 URL 的助手。 - 如果我們有的話,讓我們用新文件更新助手。 - 如果沒有,讓我們建立一個新的助手。 - 我們傳遞「你是文件助理」的指令,需要注意的是,我們希望最終輸出為「MD」格式,以便稍後更好地顯示。 對於拼圖的最後一塊,讓我們將新助手儲存到我們的資料庫中。 **這是程式碼:** ``` await io.runTask("save-assistant", async () => { await prisma.assistant.upsert({ where: { url: payload.url }, update: { aId: assistant.id, }, create: { aId: assistant.id, url: payload.url, } }); }); ``` 如果該 URL 已經存在,我們可以嘗試使用新的助手 ID 來更新它。 這是該頁面的完整程式碼: ``` import { eventTrigger } from "@trigger.dev/sdk"; import { client } from "@openai-assistant/trigger"; import {object, string} from "zod"; import {JSDOM} from "jsdom"; import {chunk} from "lodash"; import {prisma} from "@openai-assistant/helper/prisma.client"; import {openai} from "@openai-assistant/helper/open.ai"; const makeId = (length: number) => { let text = ''; const possible = 'ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789'; for (let i = 0; i < length; i += 1) { text += possible.charAt(Math.floor(Math.random() * possible.length)); } return text; }; client.defineJob({ // This is the unique identifier for your Job; it must be unique across all Jobs in your project. id: "process-documentation", name: "Process Documentation", version: "0.0.1", // This is triggered by an event using eventTrigger. You can also trigger Jobs with webhooks, on schedules, and more: https://trigger.dev/docs/documentation/concepts/triggers/introduction trigger: eventTrigger({ name: "process.documentation.event", schema: object({ url: string(), }) }), integrations: { openai }, run: async (payload, io, ctx) => { // The first task to get the sitemap URL from the website const getSiteMap = await io.runTask("grab-sitemap", async () => { const data = await (await fetch(payload.url)).text(); const dom = new JSDOM(data); const sitemap = dom.window.document.querySelector('[rel="sitemap"]')?.getAttribute('href'); return new URL(sitemap!, payload.url).toString(); }); // We parse the sitemap; instead of using some XML parser, we just use regex to get the URLs and we return it in chunks of 25 const {identifier, list} = await io.runTask("load-and-parse-sitemap", async () => { const urls = /(http|ftp|https):\/\/([\w_-]+(?:(?:\.[\w_-]+)+))([\w.,@?^=%&:\/~+#-]*[\w@?^=%&\/~+#-])/g; const identifier = makeId(5); const data = await (await fetch(getSiteMap)).text(); // @ts-ignore return {identifier, list: chunk(([...new Set(data.match(urls))] as string[]).filter(f => f.includes(payload.url)).map(p => ({identifier, url: p})), 25)}; }); // We go into each page and grab the content; we do this in batches of 25 and save it to the DB let i = 0; for (const item of list) { await processContent.batchInvokeAndWaitForCompletion( 'process-list-' + i, item.map( payload => ({ payload, }), 86_400), ); i++; } // We get the data that we saved in batches from the DB const data = await io.runTask("get-extracted-data", async () => { return (await prisma.docs.findMany({ where: { identifier }, select: { content: true } })).map((d) => d.content).join('\n\n'); }); // We upload the data to OpenAI with all the content const file = await io.openai.files.createAndWaitForProcessing("upload-file", { purpose: "assistants", file: data }); // We create a new assistant or update the old one with the new file const assistant = await io.openai.runTask("create-or-update-assistant", async (openai) => { const currentAssistant = await prisma.assistant.findFirst({ where: { url: payload.url } }); if (currentAssistant) { return openai.beta.assistants.update(currentAssistant.aId, { file_ids: [file.id] }); } return openai.beta.assistants.create({ name: identifier, description: 'Documentation', instructions: 'You are a documentation assistant, you have been loaded with documentation from ' + payload.url + ', return everything in an MD format.', model: 'gpt-4-1106-preview', tools: [{ type: "code_interpreter" }, {type: 'retrieval'}], file_ids: [file.id], }); }); // We update our internal database with the assistant await io.runTask("save-assistant", async () => { await prisma.assistant.upsert({ where: { url: payload.url }, update: { aId: assistant.id, }, create: { aId: assistant.id, url: payload.url, } }); }); }, }); export function getElementsBetween(startElement: Element, endElement: Element) { let currentElement = startElement; const elements = []; // Traverse the DOM until the endElement is reached while (currentElement && currentElement !== endElement) { currentElement = currentElement.nextElementSibling!; // If there's no next sibling, go up a level and continue if (!currentElement) { // @ts-ignore currentElement = startElement.parentNode!; startElement = currentElement; if (currentElement === endElement) break; continue; } // Add the current element to the list if (currentElement && currentElement !== endElement) { elements.push(currentElement); } } return elements; } // This job will grab the content from the website const processContent = client.defineJob({ // This is the unique identifier for your Job; it must be unique across all Jobs in your project. id: "process-content", name: "Process Content", version: "0.0.1", // This is triggered by an event using eventTrigger. You can also trigger Jobs with webhooks, on schedules, and more: https://trigger.dev/docs/documentation/concepts/triggers/introduction trigger: eventTrigger({ name: "process.content.event", schema: object({ url: string(), identifier: string(), }) }), run: async (payload, io, ctx) => { return io.runTask('grab-content', async () => { try { // We first grab a raw HTML of the content from the website const data = await (await fetch(payload.url)).text(); // We load it with JSDOM so we can manipulate it const dom = new JSDOM(data); // We remove all the scripts and styles from the page dom.window.document.querySelectorAll('script, style').forEach((el) => el.remove()); // We grab all the titles from the page const content = Array.from(dom.window.document.querySelectorAll('h1, h2, h3, h4, h5, h6')); // We grab the last element so we can get the content between the last element and the next element const lastElement = content[content.length - 1]?.parentElement?.nextElementSibling!; const elements = []; // We loop through all the elements and grab the content between each title for (let i = 0; i < content.length; i++) { const element = content[i]; const nextElement = content?.[i + 1] || lastElement; const elementsBetween = getElementsBetween(element, nextElement); elements.push({ title: element.textContent, content: elementsBetween.map((el) => el.textContent).join('\n') }); } // We create a raw text format of all the content const page = ` ---------------------------------- url: ${payload.url}\n ${elements.map((el) => `${el.title}\n${el.content}`).join('\n')} ---------------------------------- `; // We save it to our database await prisma.docs.upsert({ where: { url: payload.url }, update: { content: page, identifier: payload.identifier }, create: { url: payload.url, content: page, identifier: payload.identifier } }); } catch (e) { console.log(e); } }); }, }); ``` 我們已經完成建立後台作業來抓取和索引文件🎉 ### 詢問助理 現在,讓我們建立一個任務來詢問我們的助手。 前往“jobs”並建立一個新檔案“question.assistant.ts”。 **新增以下程式碼:** ``` import {eventTrigger} from "@trigger.dev/sdk"; import {client} from "@openai-assistant/trigger"; import {object, string} from "zod"; import {openai} from "@openai-assistant/helper/open.ai"; client.defineJob({ // This is the unique identifier for your Job; it must be unique across all Jobs in your project. id: "question-assistant", name: "Question Assistant", version: "0.0.1", // This is triggered by an event using eventTrigger. You can also trigger Jobs with webhooks, on schedules, and more: https://trigger.dev/docs/documentation/concepts/triggers/introduction trigger: eventTrigger({ name: "question.assistant.event", schema: object({ content: string(), aId: string(), threadId: string().optional(), }) }), integrations: { openai }, run: async (payload, io, ctx) => { // Create or use an existing thread const thread = payload.threadId ? await io.openai.beta.threads.retrieve('get-thread', payload.threadId) : await io.openai.beta.threads.create('create-thread'); // Create a message in the thread await io.openai.beta.threads.messages.create('create-message', thread.id, { content: payload.content, role: 'user', }); // Run the thread const run = await io.openai.beta.threads.runs.createAndWaitForCompletion('run-thread', thread.id, { model: 'gpt-4-1106-preview', assistant_id: payload.aId, }); // Check the status of the thread if (run.status !== "completed") { console.log('not completed'); throw new Error(`Run finished with status ${run.status}: ${JSON.stringify(run.last_error)}`); } // Get the messages from the thread const messages = await io.openai.beta.threads.messages.list("list-messages", run.thread_id, { query: { limit: "1" } }); const content = messages[0].content[0]; if (content.type === 'text') { return {content: content.text.value, threadId: thread.id}; } } }); ``` - 該事件需要三個參數 - `content` - 我們想要傳送給助理的訊息。 - `aId` - 我們先前建立的助手的內部 ID。 - `threadId` - 對話的執行緒 ID。正如您所看到的,這是一個可選參數,因為在第一個訊息中,我們還沒有線程 ID。 - 然後,我們建立或取得前一個執行緒的執行緒。 - 我們在助理提出的問題的線索中加入一條新訊息。 - 我們執行線程並等待它完成。 - 我們取得訊息清單(並將其限制為 1),因為第一則訊息是對話中的最後一則訊息。 - 我們返回訊息內容和我們剛剛建立的線程ID。 ### 新增路由 我們需要為我們的應用程式建立 3 個 API 路由: 1、派新助理進行處理。 2. 透過URL獲取特定助手。 3. 新增訊息給助手。 在「app/api」中建立一個名為assistant的新資料夾,並在其中建立一個名為「route.ts」的新檔案。裡面加入如下程式碼: ``` import {client} from "@openai-assistant/trigger"; import {prisma} from "@openai-assistant/helper/prisma.client"; export async function POST(request: Request) { const body = await request.json(); if (!body.url) { return new Response(JSON.stringify({error: 'URL is required'}), {status: 400}); } // We send an event to the trigger to process the documentation const {id: eventId} = await client.sendEvent({ name: "process.documentation.event", payload: {url: body.url}, }); return new Response(JSON.stringify({eventId}), {status: 200}); } export async function GET(request: Request) { const url = new URL(request.url).searchParams.get('url'); if (!url) { return new Response(JSON.stringify({error: 'URL is required'}), {status: 400}); } const assistant = await prisma.assistant.findFirst({ where: { url: url } }); return new Response(JSON.stringify(assistant), {status: 200}); } ``` 第一個「POST」方法取得一個 URL,並使用用戶端傳送的 URL 觸發「process.documentation.event」作業。 第二個「GET」方法從我們的資料庫中透過客戶端發送的 URL 取得助手。 現在,讓我們建立向助手新增訊息的路由。 在「app/api」內部建立一個新資料夾「message」並新增一個名為「route.ts」的新文件,然後新增以下程式碼: ``` import {prisma} from "@openai-assistant/helper/prisma.client"; import {client} from "@openai-assistant/trigger"; export async function POST(request: Request) { const body = await request.json(); // Check that we have the assistant id and the message if (!body.id || !body.message) { return new Response(JSON.stringify({error: 'Id and Message are required'}), {status: 400}); } // get the assistant id in OpenAI from the id in the database const assistant = await prisma.assistant.findUnique({ where: { id: +body.id } }); // We send an event to the trigger to process the documentation const {id: eventId} = await client.sendEvent({ name: "question.assistant.event", payload: { content: body.message, aId: assistant?.aId, threadId: body.threadId }, }); return new Response(JSON.stringify({eventId}), {status: 200}); } ``` 這是一個非常基本的程式碼。我們從客戶端獲取訊息、助手 ID 和線程 ID,並將其發送到我們之前建立的「question.assistant.event」。 最後要做的事情是建立一個函數來獲取我們所有的助手。 在「helpers」內部建立一個名為「get.list.ts」的新函數並新增以下程式碼: ``` import {prisma} from "@openai-assistant/helper/prisma.client"; // Get the list of all the available assistants export const getList = () => { return prisma.assistant.findMany({ }); } ``` 非常簡單的程式碼即可獲得所有助手。 我們已經完成了後端🥳 讓我們轉到前面。 --- ![前端](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/k3s5gks1j0ojoz11b93i.png) ## 建立前端 我們將建立一個基本介面來新增 URL 並顯示已新增 URL 的清單: ![ss1](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/ihvx4yn6uee6gritr9nh.png) ### 首頁 將 `app/page.tsx` 的內容替換為以下程式碼: ``` import {getList} from "@openai-assistant/helper/get.list"; import Main from "@openai-assistant/components/main"; export default async function Home() { const list = await getList(); return ( <Main list={list} /> ) } ``` 這是一個簡單的程式碼,它從資料庫中取得清單並將其傳遞給我們的 Main 元件。 接下來,讓我們建立“Main”元件。 在「app」內建立一個新資料夾「components」並新增一個名為「main.tsx」的新檔案。 **新增以下程式碼:** ``` "use client"; import {Assistant} from '@prisma/client'; import {useCallback, useState} from "react"; import {FieldValues, SubmitHandler, useForm} from "react-hook-form"; import {ChatgptComponent} from "@openai-assistant/components/chatgpt.component"; import {AssistantList} from "@openai-assistant/components/assistant.list"; import {TriggerProvider} from "@trigger.dev/react"; export interface ExtendedAssistant extends Assistant { pending?: boolean; eventId?: string; } export default function Main({list}: {list: ExtendedAssistant[]}) { const [assistantState, setAssistantState] = useState(list); const {register, handleSubmit} = useForm(); const submit: SubmitHandler<FieldValues> = useCallback(async (data) => { const assistantResponse = await (await fetch('/api/assistant', { body: JSON.stringify({url: data.url}), method: 'POST', headers: { 'Content-Type': 'application/json' } })).json(); setAssistantState([...assistantState, {...assistantResponse, url: data.url, pending: true}]); }, [assistantState]) const changeStatus = useCallback((val: ExtendedAssistant) => async () => { const assistantResponse = await (await fetch(`/api/assistant?url=${val.url}`, { method: 'GET', headers: { 'Content-Type': 'application/json' } })).json(); setAssistantState([...assistantState.filter((v) => v.id), assistantResponse]); }, [assistantState]) return ( <TriggerProvider publicApiKey={process.env.NEXT_PUBLIC_TRIGGER_PUBLIC_API_KEY!}> <div className="w-full max-w-2xl mx-auto p-6 flex flex-col gap-4"> <form className="flex items-center space-x-4" onSubmit={handleSubmit(submit)}> <input className="flex-grow p-3 border border-black/20 rounded-xl" placeholder="Add documentation link" type="text" {...register('url', {required: 'true'})} /> <button className="flex-shrink p-3 border border-black/20 rounded-xl" type="submit"> Add </button> </form> <div className="divide-y-2 divide-gray-300 flex gap-2 flex-wrap"> {assistantState.map(val => ( <AssistantList key={val.url} val={val} onFinish={changeStatus(val)} /> ))} </div> {assistantState.filter(f => !f.pending).length > 0 && <ChatgptComponent list={assistantState} />} </div> </TriggerProvider> ) } ``` 讓我們看看這裡發生了什麼: - 我們建立了一個名為「ExtendedAssistant」的新接口,其中包含兩個參數「pending」和「eventId」。當我們建立一個新的助理時,我們沒有最終的值,我們將只儲存`eventId`並監聽作業處理直到完成。 - 我們從伺服器元件取得清單並將其設定為新狀態(以便我們稍後可以修改它) - 我們新增了「TriggerProvider」來幫助我們監聽事件完成並用資料更新它。 - 我們使用「react-hook-form」建立一個新表單來新增助手。 - 我們新增了一個帶有一個輸入「URL」的表單來提交新的助理進行處理。 - 我們迭代並顯示所有現有的助手。 - 在提交表單時,我們將資訊傳送到先前建立的「路由」以新增助理。 - 事件完成後,我們觸發「changeStatus」以從資料庫載入助手。 - 最後,我們有了 ChatGPT 元件,只有在沒有等待處理的助手時才會顯示(`!f.pending`) 讓我們建立 `AssistantList` 元件。 在「components」內,建立一個新檔案「assistant.list.tsx」並在其中加入以下內容: ``` "use client"; import {FC, useEffect} from "react"; import {ExtendedAssistant} from "@openai-assistant/components/main"; import {useEventRunDetails} from "@trigger.dev/react"; export const Loading: FC<{eventId: string, onFinish: () => void}> = (props) => { const {eventId} = props; const { data, error } = useEventRunDetails(eventId); useEffect(() => { if (!data || error) { return ; } if (data.status === 'SUCCESS') { props.onFinish(); } }, [data]); return <div className="pointer bg-yellow-300 border-yellow-500 p-1 px-3 text-yellow-950 border rounded-2xl">Loading</div> }; export const AssistantList: FC<{val: ExtendedAssistant, onFinish: () => void}> = (props) => { const {val, onFinish} = props; if (val.pending) { return <Loading eventId={val.eventId!} onFinish={onFinish} /> } return ( <div key={val.url} className="pointer relative bg-green-300 border-green-500 p-1 px-3 text-green-950 border rounded-2xl hover:bg-red-300 hover:border-red-500 hover:text-red-950 before:content-[attr(data-content)]" data-content={val.url} /> ) } ``` 我們迭代我們建立的所有助手。如果助手已經建立,我們只顯示名稱。如果沒有,我們渲染`<Loading />`元件。 載入元件在螢幕上顯示“正在載入”,並長時間輪詢伺服器直到事件完成。 我們使用 Trigger.dev 建立的 useEventRunDetails 函數來了解事件何時完成。 事件完成後,它會觸發「onFinish」函數,用新建立的助手更新我們的客戶端。 ### 聊天介面 ![聊天介面](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/0u7db3qwz03d6jkk965a.png) 現在,讓我們加入 ChatGPT 元件並向我們的助手提問! - 選擇我們想要使用的助手 - 顯示訊息列表 - 新增我們要傳送的訊息的輸入和提交按鈕。 在「components」內部新增一個名為「chatgpt.component.tsx」的新文件 讓我們繪製 ChatGPT 聊天框: ``` "use client"; import {FC, useCallback, useEffect, useRef, useState} from "react"; import {ExtendedAssistant} from "@openai-assistant/components/main"; import Markdown from 'react-markdown' import {useEventRunDetails} from "@trigger.dev/react"; interface Messages { message?: string eventId?: string } export const ChatgptComponent = ({list}: {list: ExtendedAssistant[]}) => { const url = useRef<HTMLSelectElement>(null); const [message, setMessage] = useState(''); const [messagesList, setMessagesList] = useState([] as Messages[]); const [threadId, setThreadId] = useState<string>('' as string); const submitForm = useCallback(async (e: any) => { e.preventDefault(); setMessagesList((messages) => [...messages, {message: `**[ME]** ${message}`}]); setMessage(''); const messageResponse = await (await fetch('/api/message', { method: 'POST', body: JSON.stringify({message, id: url.current?.value, threadId}), })).json(); if (!threadId) { setThreadId(messageResponse.threadId); } setMessagesList((messages) => [...messages, {eventId: messageResponse.eventId}]); }, [message, messagesList, url, threadId]); return ( <div className="border border-black/50 rounded-2xl flex flex-col"> <div className="border-b border-b-black/50 h-[60px] gap-3 px-3 flex items-center"> <div>Assistant:</div> <div> <select ref={url} className="border border-black/20 rounded-xl p-2"> {list.filter(f => !f.pending).map(val => ( <option key={val.id} value={val.id}>{val.url}</option> ))} </select> </div> </div> <div className="flex-1 flex flex-col gap-3 py-3 w-full min-h-[500px] max-h-[1000px] overflow-y-auto overflow-x-hidden messages-list"> {messagesList.map((val, index) => ( <div key={index} className={`flex border-b border-b-black/20 pb-3 px-3`}> <div className="w-full"> {val.message ? <Markdown>{val.message}</Markdown> : <MessageComponent eventId={val.eventId!} onFinish={setThreadId} />} </div> </div> ))} </div> <form onSubmit={submitForm}> <div className="border-t border-t-black/50 h-[60px] gap-3 px-3 flex items-center"> <div className="flex-1"> <input value={message} onChange={(e) => setMessage(e.target.value)} className="read-only:opacity-20 outline-none border border-black/20 rounded-xl p-2 w-full" placeholder="Type your message here" /> </div> <div> <button className="border border-black/20 rounded-xl p-2 disabled:opacity-20" disabled={message.length < 3}>Send</button> </div> </div> </form> </div> ) } export const MessageComponent: FC<{eventId: string, onFinish: (threadId: string) => void}> = (props) => { const {eventId} = props; const { data, error } = useEventRunDetails(eventId); useEffect(() => { if (!data || error) { return ; } if (data.status === 'SUCCESS') { props.onFinish(data.output.threadId); } }, [data]); if (!data || error || data.status !== 'SUCCESS') { return ( <div className="flex justify-end items-center pb-3 px-3"> <div className="animate-spin rounded-full h-3 w-3 border-t-2 border-b-2 border-blue-500" /> </div> } return <Markdown>{data.output.content}</Markdown>; }; ``` 這裡正在發生一些令人興奮的事情: - 當我們建立新訊息時,我們會自動將其呈現在螢幕上作為「我們的」訊息,但是當我們將其發送到伺服器時,我們需要推送事件 ID,因為我們還沒有訊息。這就是我們使用 `{val.message ? <Markdown>{val.message}</Markdown> : <MessageComponent eventId={val.eventId!} onFinish={setThreadId} />}` - 我們用「Markdown」元件包裝訊息。如果您還記得,我們在前面的步驟中告訴 ChatGPT 以 MD 格式輸出所有內容,以便我們可以正確渲染它。 - 事件處理完成後,我們會更新線程 ID,以便我們從以下訊息中獲得相同對話的上下文。 我們就完成了🎉 --- ![完成](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/0half2g6r5zfn7asq084.png) ## 讓我們聯絡吧! 🔌 作為開源開發者,您可以加入我們的[社群](https://discord.gg/nkqV9xBYWy) 做出貢獻並與維護者互動。請隨時造訪我們的 [GitHub 儲存庫](https://github.com/triggerdotdev/trigger.dev),貢獻並建立與 Trigger.dev 相關的問題。 本教學的源程式碼可在此處取得: [https://github.com/triggerdotdev/blog/tree/main/openai-assistant](https://github.com/triggerdotdev/blog/tree/main/openai-assistant) 感謝您的閱讀! --- 原文出處:https://dev.to/triggerdotdev/train-chatgpt-on-your-documentation-1a9g

使用 Markdown 與 React 在 Github 文件中顯示淺色模式、深色模式

警告:所表達的觀點可能不適合所有受眾! 😂 ## 簡介 在本文結束時,您將了解並能夠根據使用者偏好 - **深色**或**淺色**模式展示您的 Markdown 影像。 1. 我將介紹如何在 GitHub README.md 中加入兩個圖像 - 根據所選的“主題”,您的圖像將正確回應。 2. 我將引導您在 Markdown 中合併影像的過程,並示範如何使用 React 使它們回應。 😎 ___ ## 你使用淺色還是深色? 我不了解你的情況,但無論平台如何,如果他們可以選擇在淺色和深色模式之間切換,那就沒有競爭了。 淺色主題正在切換為深色,事實上,當然在我寫這篇文章的時候! ![深色主題](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/s0yfjc2yv5pfyacu74go.png) 話雖如此,在軟體開發的快速發展中,創造無縫的使用者體驗至關重要。 這種體驗的一部分涉及適應使用者偏好,例如淺色和深色模式。 我還記得幾年前,Github 宣布了用戶可以切換到「深色模式」的選項,這是一件非常大的事情。 ![GitHub 深色主題](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/qi10urco3o6fdojm6ipf.png) 【Github揭曉黑暗主題的重要時刻】(https://t.co/HEotvXVJ7R) 🤩 2020 年 12 月 8 日🎆 近年來,使用者介面中深色和淺色模式選項的出現已成為一種流行趨勢。 我絕對不是唯一一個喜歡使用深色主題選項的人,根據 Android 用戶的說法,[91.8% 的用戶更喜歡深色模式](https://www.androidauthority.com/dark-mode-poll-results-1090716/) 所以我們可以猜測這個數字在所有作業系統中都相當高。 這當然可能會引起激烈的爭論,所以我會盡力將自己的觀點降到最低。 ![輕模式迷因](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/m3yiepj8a46rwhu69fgw.png) ## 改善使用者體驗 主要目標是透過在應用程式中提供選項來改善用戶體驗。 有多種方法可以建立每個圖像的多個版本,在本教程中我們不會深入討論細節。 只要確保您的圖像在兩個主題中脫穎而出並具有透明背景,您就會獲得成功。 **_讓我們開始派對吧!_** ## GitHub 自述文件中的響應式圖像 您有一個專案並想讓您的 GitHub 專案 README.md 真正流行嗎? 無論使用者使用什麼淺色主題,我們都需要一種方法來指定圖像應在 Markdown 中顯示哪種主題(淺色或深色)。 當您想要根據使用者選擇的配色方案優化圖片的顯示時,這特別有用,並且它涉及將 **HTML `<picture>`** 元素與 `prefers-color-scheme` 媒體功能結合使用如下所示。 繼續將圖片檔案直接拖曳到 GitHub 中並放在“srcset=”後面。 ``` <picture> <source media="(prefers-color-scheme: dark)" srcset="https://github.com/boxyhq/.github/assets/66887028/df1c9904-df2f-4515-b403-58b14a0e9093"> <source media="(prefers-color-scheme: light)" srcset="https://github.com/boxyhq/.github/assets/66887028/e093a466-72ea-41c6-a292-4c39a150facd"> <img alt="BoxyHQ Banner" src="https://github.com/boxyhq/jackson/assets/66887028/b40520b7-dbce-400b-88d3-400d1c215ea1"> </picture> ``` 瞧! ![SAML Jackson 暗模式](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/q51g41fjfqnposn50una.png) ![SAML Jackson 燈光模式](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/d0xzs88txjylnixilaqu.png) 太好了,你有 5 秒嗎? https://github.com/boxyhq/jackson --- ## 使用 React 在 Markdown 中回應影像 假設今天我將像平常一樣用 Markdown 編寫博客,並將其發佈到我的網站上。 我使用的圖像需要根據使用者偏好做出回應,但在 Markdown 中不可能偵聽本地儲存和設定狀態中的「主題」變更。 ![本機儲存](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/vrjz4to8x17h63dtybxn.png) 值得慶幸的是,如果我們將 React 匯入到 Markdown 檔案中,但先建立一個元件,就有一種方法可以解決這個困境。 ## 反應文件 ``` src/components/LightDarkToggle.js import React, { useEffect, useState } from 'react'; function ToggleImages() { // Define a state variable to track the user's login status const [currentTheme, setcurrentTheme] = useState(localStorage.getItem('theme')); // Add an event listener for the 'storage' event inside a useEffect useEffect(() => { const handleStorageChange = (event) => { console.log('Storage event detected:', event); // Check the changed key and update the state accordingly console.log("event", event.key) if (event.key === 'theme') { setcurrentTheme(event.newValue); } }; window.addEventListener('storage', handleStorageChange); // Clean up the event listener when the component unmounts return () => { window.removeEventListener('storage', handleStorageChange); }; }, []); // The empty dependency array ensures that this effect runs once when the component mounts return ( <div className="image-container"> {currentTheme == 'light'? ( <img id="light-mode-image" src="/img/blog/boxyhq-banner-light-bg.png" alt="Light Mode Image" ></img> ):( <img id="dark-mode-image" src="/img/blog/boxyhq-banner-dark-bg.png" alt="Dark Mode Image" ></img> )} </div> ); } export default ToggleImages; ``` 我在程式碼中加入了註釋和一些控制台日誌,以幫助了解正在發生的事情,但讓我們快速分解它。 - React useState 鉤子管理 `currentTheme` 的狀態,它代表使用者選擇的儲存在本機儲存中的主題。 - useEffect 掛鉤用於為「儲存」事件新增事件偵聽器。當儲存事件發生時(表示本機儲存發生變化),元件會檢查變更的鍵是否為“theme”,並相應地更新“currentTheme”狀態。 - 此元件根據使用者選擇的主題呈現不同的影像,如果主題是“淺色”,則顯示淺色模式影像;如果主題是其他主題,則顯示深色模式影像。 酷,讓我們繼續吧! ## 降價文件 讓我們為新部落格建立一個 .md 檔案。 ``` --- slug: light-and-dark-mode-responsive-images title: 'Light and Dark Mode Responsive Images' tags_disabled: [ developer, react, javascript, open-source, ] image: /img/blog/light-dark.png author: Nathan Tarbert author_title: Community Engineer @BoxyHQ author_url: https://github.com/NathanTarbert author_image_url: https://boxyhq.com/img/team/nathan.jpg --- import ToggleImages from '../src/components/LightDarkToggle.js'; ## 🤩 Let's start this blog off with a bang! Our business logo is now responsive with each user's preference, whether it's **light** or **dark** mode! <div> <ToggleImages /> </div> More blog words... ``` 此時,我們只需匯入 React 元件並將其呈現在 Markdown 檔案中。 由於這是一個 Next.js 應用程式,讓我們啟動伺服器“npm run dev”並查看結果。 ![貓鼓滾](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/lyjzjqgcwaubyj5ve1o3.gif) ![網站深色模式](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/qraltb34mrl9y8j9jppq.png) 並切換到淺色主題 ![網站燈光模式](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/u33jgzha5fbfy6tlb4hs.png) 讓我們打開控制台來查看我們的事件 ![console.log](https://dev-to-uploads.s3.amazonaws.com/uploads/articles/qpxz5gbhwt308vatsnkp.png) 你有它! 這些是在 Markdown 中展示響應式映像的幾種方法,其中一個範例使用 React 來幫助我們在本地儲存中設定狀態。 我希望您喜歡這篇文章,如果您喜歡開發,請在 [X (Twitter)](https://twitter.com/nathan_tarbert) 上關注我,我們下次再見! --- 原文出處:https://dev.to/nathan_tarbert/the-zebras-guide-to-showcase-your-images-in-light-dark-17f5

Javascript Proxy Magic:我如何建立一個零依賴的 2kB 狀態管理器(以及它如何為我提供了兩個不同的工作機會)

狀態管理器到底是什麼?狀態管理器是一個智慧模組,能夠保留(應用程式或 Web 應用程式的)會話資料並對資料的變更做出反應。 您是網頁開發人員嗎?使用過 Redux、Mobx 或 Zustand 等函式庫嗎?恭喜!您已經使用了狀態管理器。 我記得我第一天嘗試為 React 設定(舊的)Redux。只要想到所有不必要的複雜性——調度程序、減速器、中間件,我就會患上創傷後壓力症候群(PTSD)!我只是想聲明一些變數,_請讓它停止_。 ![](https://media.tenor.com/Fj8YV_9ut8UAAAC/makeitstop-i-just-want-it-to-stop.gif) 這是一個過度設計、臃腫的庫,每個人都在使用!由於某種瘋狂的、未知的原因,它成為了當時的行業標準。 ###一些背景故事 2021 年的一個晚上,當我無法入睡時,我漫無目的地打開 GitHub,注意到我以前的大學課程老師(我在 GH 上關注過他)為他現在的學生上傳了一份作業。該作業要求學生使用公共 Pokemon API 建立一個 Pokedex 網站。目標是用 Javascript 實現它(沒有框架或函式庫,因為他目前的學生是 Web 開發初學者,仍在學習 Javascript 和開發的基礎知識)。 作為一個笑話,主要是因為我睡不著,我開始在我的神奇寶貝網站上工作。最終,我能夠建立一些可行的東西,而無需使用任何外部庫。 ### 但一路走來,我很掙扎...... 你看,我已經習慣了擁有一個狀態管理器,以至於在不使用外部框架或庫的情況下建置一個簡單的兩頁應用程式的要求讓我開始思考 - _為什麼狀態管理器必須如此復雜?這只是變數和事件._ 長話短說,我發現自己在凌晨 2 點組裝了一個超級簡單的狀態管理器模組,只是為了管理我的 Pokemon Web 應用程式的狀態。我將我的網站部署到了 GitHub 頁面,然後就忘記了這一切。 幾個月過去了,但出於某種原因,我時不時地思考我的狀態管理解決方案...你看,它有其他庫沒有的東西 - _它太簡單了。_ _“嘿!”我心想,「我應該將它重寫為 NPM 套件」。_ 當天晚上,我就這麼做了——我把它寫成了一個獨立的 NPM 包。最後,它的重量為 2kB(相比之下 Redux 的 150kB),具有零依賴性,並且使用起來非常簡單,您只需 3 行程式碼即可完成設定。 ### 我稱之為 VSSM 代表**_非常小的狀態管理器_**。 您可以在[GitHub](https://github.com/lnahrf/Vssm)上查看原始程式碼。另外,請查看使用 React 和 VSSM 建立的[文件網站](https://lnahrf.github.io/Vssm-docs/)。 第二天,我發布了我的 NPM 包,然後又忘記了這件事。 同年晚些時候,我面試了兩家不同公司的全端開發人員職位。我在第一家公司的面試中取得了優異的成績,這是一家非常成熟的科技公司。作為面試過程的一部分,他們要求我告訴他們我是否在空閒時間編碼,或者是否有我貢獻過的任何開源專案等等。 當時我做的唯一很酷的事情就是 VSSM,所以我告訴了他們。他們對我自己建立一個「Redux 替代方案」的想法印象深刻。 另一方面,我在第二家公司的面試中慘敗。我的大腦一片空白,我很緊張,無法回答簡單的問題,例如 > “React 會在狀態變更時重新渲染整個應用程式,還是在使用 Redux 時僅更新受影響的元件及其子元件?” “每次狀態更新時,它都會重新渲染整個應用程式”,我說。 ![](https://media.tenor.com/ZFc20z8DItkAAAAd/facepalm-really.gif) 我很緊張,哈哈,顯然我知道正確的答案是「它只渲染註冊的元件以及可能受影響的子元件」。 直到今天我也不明白為什麼二號公司決定給我第二次機會。他們邀請我再次接受採訪(是的!)。 在我的第二次面試中,他們要求我告訴他們我是否在空閒時間編碼、開源貢獻,你知道該怎麼做。當我告訴面試官我的小副專案時,他看起來很高興,似乎他喜歡我只是因為我從頭開始編寫了一個狀態管理器。 我想情況確實如此,因為我第二次面試也失敗了(在程式設計挑戰期間耗盡了時間),但仍然得到了一份工作機會。 1 號公司打算向我發送報價,但我已經與 2 號公司簽署了報價。 我的底線是——我建立 VSSM 幫助我獲得了這兩個機會。 ![](https://media.tenor.com/BuoCYXAkk0AAAAAC/big-lebowski.gif) ### 我是怎麼做到的? 您是否知道 Javascript 內建了監視變數變更所需的所有功能? 它被稱為代理(它很神奇)。 Javascript 代理程式是程式碼和變數分配之間的附加邏輯層。 如果您要將物件包裝在代理程式中,您可以決定在每次更新時將其值記錄到控制台,除了為該物件指派新值之外,無需執行任何操作。 ``` const target = { v: "hello" } const proxyTarget = new Proxy(target, { set: (target, property, value) => { console.log(`${property} is now ${value}`); target[property] = value; return target[property]; } }); proxyTarget.v = "world!" // v is now world! ``` VSSM 是基於代理建置,它在變數賦值和其餘程式碼之間建立了一個層。使用代理,您可以設定 setter、getter,並在操作或請求目標值時實現任何類型的邏輯。 VSSM 不僅僅是一個代理,它是各種智慧代理,它們知道分配給變數的值是它的新值還是回調方法。 例如,使用 VSSM,您只需幾行程式碼即可設定狀態、監聽變更並發出事件。 ``` import { createVSSM, createState } from 'vssm'; import { getVSSM } from 'vssm'; // Create the initial state createVSSM({ user: createState('user', { address: '' }) }); // Get the user proxy reference const { user } = getVSSM(); // Listen to events on user.address user.address = () => { console.log(`Address updated! the new address is ${user.address}`); }; // Emit the mutation event user.address = 'P.Sherman 42 Wallaby Way, Sydney' ``` 正如您所看到的,我確保我的狀態管理器盡可能簡單。我的目標是擺脫僅僅為了分配一些變數而陷入減速器、中間件和極其複雜的配置的困境。 現在,一切都透過分配變數來進行!想要設定監聽器嗎?將回調函數指派給變數。想要編輯值並發出事件嗎?只需指派一個新值即可。 直到今天我仍然不明白為什麼流行的狀態管理器必須如此複雜,也許我永遠不會。 我鼓勵您繼續閱讀 [MDN Web Docs](https://developer.mozilla.org/en-US/docs/Web/JavaScript/Reference/Global_Objects/Proxy) 上有關 Javascript 代理的所有內容。 ### 這一切的結論是什麼? 我認為,對自己所做的事情充滿熱情是關鍵。 我建立 VSSM 只是為了突破自己的極限並發布合理的 NPM 包。它成功地給面試官和同事留下了深刻的印象,並讓我從那時起就進入了不同的職位。 沒有人會使用 VSSM,它不會流行。當我將其發佈到 NPM 時,我就意識到了這一事實。但我仍然選擇盡我所能,因為我熱衷於做一些我認為比行業標準更好的事情。我知道我可以做出一些必須更好的東西,即使這意味著它對我更好。 儘管 VSSM 已經死在 NPM 墓地裡,但它給我帶來了很多價值,並且因為這篇文章而繼續這樣做。 獲得開發工作的最佳方法是建立令人驚嘆的東西,即使您認為這一切以前都已經完成了 - 建置得更好。即使您認為沒有人會使用它,那又有什麼意義呢? - 現在建置,價值稍後顯現。 不要低估你的能力,如果你認為自己有不足,請知道你會進步。走出去,建構能夠帶來價值的專案,一次一小步。 祝您工程之旅順利。 --- 原文出處:https://dev.to/lnahrf/javascript-proxy-magic-how-i-built-a-2kb-state-manager-with-zero-dependencies-and-how-it-got-me-two-different-job-offers-2539

快速複習 React 基本觀念&實務範例:推薦新手參考

React 作為一個強大的前端工具,有很多需要熟悉的基本觀念&語法。 這篇文章做一個快速的全面整理,方便工作時,可以隨時翻閱,希望您喜歡。 - 原文出處:https://dev.to/reedbarger/the-react-cheatsheet-for-2020-real-world-examples-4hgg ## 核心概念 ### 元素和 JSX - React 元素的基本語法 ``` // In a nutshell, JSX allows us to write HTML in our JS // JSX can use any valid html tags (i.e. div/span, h1-h6, form/input, etc) <div>Hello React</div> ``` - JSX 元素就是一段表達式 ``` // as an expression, JSX can be assigned to variables... const greeting = <div>Hello React</div>; const isNewToReact = true; // ... or can be displayed conditionally function sayGreeting() { if (isNewToReact) { // ... or returned from functions, etc. return greeting; // displays: Hello React } else { return <div>Hi again, React</div>; } } ``` - JSX 允許我們嵌套表達式 ``` const year = 2020; // we can insert primitive JS values in curly braces: {} const greeting = <div>Hello React in {year}</div>; // trying to insert objects will result in an error ``` - JSX 允許我們嵌套元素 ``` // to write JSX on multiple lines, wrap in parentheses: () const greeting = ( // div is the parent element <div> {/* h1 and p are child elements */} <h1>Hello!</h1> <p>Welcome to React</p> </div> ); // 'parents' and 'children' are how we describe JSX elements in relation // to one another, like we would talk about HTML elements ``` - HTML 和 JSX 的語法略有不同 ``` // Empty div is not <div></div> (HTML), but <div/> (JSX) <div/> // A single tag element like input is not <input> (HTML), but <input/> (JSX) <input name="email" /> // Attributes are written in camelcase for JSX (like JS variables <button className="submit-button">Submit</button> // not 'class' (HTML) ``` - 最基本的 React 應用程式需要三樣東西: - 1. ReactDOM.render() 渲染我們的應用程序 - 2. 一個 JSX 元素(在此情況下,稱為根節點) - 3. 一個用於掛載應用程式的 DOM 元素(通常是 index.html 文件中 id 為 root 的 div) ``` // imports needed if using NPM package; not if from CDN links import React from "react"; import ReactDOM from "react-dom"; const greeting = <h1>Hello React</h1>; // ReactDOM.render(root node, mounting point) ReactDOM.render(greeting, document.getElementById("root")); ``` ### 元件和 Props - 基本 React 元件的語法 ``` import React from "react"; // 1st component type: function component function Header() { // function components must be capitalized unlike normal JS functions // note the capitalized name here: 'Header' return <h1>Hello React</h1>; } // function components with arrow functions are also valid const Header = () => <h1>Hello React</h1>; // 2nd component type: class component // (classes are another type of function) class Header extends React.Component { // class components have more boilerplate (with extends and render method) render() { return <h1>Hello React</h1>; } } ``` - 如何使用元件 ``` // do we call these function components like normal functions? // No, to execute them and display the JSX they return... const Header = () => <h1>Hello React</h1>; // ...we use them as 'custom' JSX elements ReactDOM.render(<Header />, document.getElementById("root")); // renders: <h1>Hello React</h1> ``` - 元件可以在我們的應用程式中重複使用 ``` // for example, this Header component can be reused in any app page // this component shown for the '/' route function IndexPage() { return ( <div> <Header /> <Hero /> <Footer /> </div> ); } // shown for the '/about' route function AboutPage() { return ( <div> <Header /> <About /> <Testimonials /> <Footer /> </div> ); } ``` - 資料可以通過 props 動態傳遞給元件 ``` // What if we want to pass data to our component from a parent? // I.e. to pass a user's name to display in our Header? const username = "John"; // we add custom 'attributes' called props ReactDOM.render( <Header username={username} />, document.getElementById("root") ); // we called this prop 'username', but can use any valid JS identifier // props is the object that every component receives as an argument function Header(props) { // the props we make on the component (i.e. username) // become properties on the props object return <h1>Hello {props.username}</h1>; } ``` - Props 不可直接改變(mutate) ``` // Components must ideally be 'pure' functions. // That is, for every input, we be able to expect the same output // we cannot do the following with props: function Header(props) { // we cannot mutate the props object, we can only read from it props.username = "Doug"; return <h1>Hello {props.username}</h1>; } // But what if we want to modify a prop value that comes in? // That's where we would use state (see the useState section) ``` - 如果我們想將元素/元件作為 props 傳遞給其它元件,可以用 children props ``` // Can we accept React elements (or components) as props? // Yes, through a special property on the props object called 'children' function Layout(props) { return <div className="container">{props.children}</div>; } // The children prop is very useful for when you want the same // component (such as a Layout component) to wrap all other components: function IndexPage() { return ( <Layout> <Header /> <Hero /> <Footer /> </Layout> ); } // different page, but uses same Layout component (thanks to children prop) function AboutPage() { return ( <Layout> <About /> <Footer /> </Layout> ); } ``` - 可以用三元運算子來條件顯示元件 ``` // if-statements are fine to conditionally show , however... // ...only ternaries (seen below) allow us to insert these conditionals // in JSX, however function Header() { const isAuthenticated = checkAuth(); return ( <nav> <Logo /> {/* if isAuth is true, show AuthLinks. If false, Login */} {isAuthenticated ? <AuthLinks /> : <Login />} {/* if isAuth is true, show Greeting. If false, nothing. */} {isAuthenticated && <Greeting />} </nav> ); } ``` - Fragments 是用來顯示多個元件的特殊元件,無需向 DOM 添加額外的元素 - Fragments 適合用在條件邏輯 ``` // we can improve the logic in the previous example // if isAuthenticated is true, how do we display both AuthLinks and Greeting? function Header() { const isAuthenticated = checkAuth(); return ( <nav> <Logo /> {/* we can render both components with a fragment */} {/* fragments are very concise: <> </> */} {isAuthenticated ? ( <> <AuthLinks /> <Greeting /> </> ) : ( <Login /> )} </nav> ); } ``` ### 列表和 keys - 使用 .map() 將資料列表(陣列)轉換為元素列表 ``` const people = ["John", "Bob", "Fred"]; const peopleList = people.map(person => <p>{person}</p>); ``` - .map() 也可用來轉換為元件列表 ``` function App() { const people = ['John', 'Bob', 'Fred']; // can interpolate returned list of elements in {} return ( <ul> {/* we're passing each array element as props */} {people.map(person => <Person name={person} />} </ul> ); } function Person({ name }) { // gets 'name' prop using object destructuring return <p>this person's name is: {name}</p>; } ``` - 迭代的每個 React 元素都需要一個特殊的 `key` prop - key 對於 React 來說是必須的,以便能夠追蹤每個正在用 map 迭代的元素 - 沒有 key,當資料改變時,React 較難弄清楚元素應該如何更新 - key 應該是唯一值,才能讓 React 知道如何分辨 ``` function App() { const people = ['John', 'Bob', 'Fred']; return ( <ul> {/* keys need to be primitive values, ideally a generated id */} {people.map(person => <Person key={person} name={person} />)} </ul> ); } // If you don't have ids with your set of data or unique primitive values, // you can use the second parameter of .map() to get each elements index function App() { const people = ['John', 'Bob', 'Fred']; return ( <ul> {/* use array element index for key */} {people.map((person, i) => <Person key={i} name={person} />)} </ul> ); } ``` ### 事件和事件處理器 - React 和 HTML 中的事件略有不同 ``` // Note: most event handler functions start with 'handle' function handleToggleTheme() { // code to toggle app theme } // in html, onclick is all lowercase <button onclick="handleToggleTheme()"> Submit </button> // in JSX, onClick is camelcase, like attributes / props // we also pass a reference to the function with curly braces <button onClick={handleToggleTheme}> Submit </button> ``` - 最該先學的 React 事件是 onClick 和 onChange - onClick 處理 JSX 元素上的點擊事件(也就是按鈕動作) - onChange 處理鍵盤事件(也就是打字動作) ``` function App() { function handleChange(event) { // when passing the function to an event handler, like onChange // we get access to data about the event (an object) const inputText = event.target.value; const inputName = event.target.name; // myInput // we get the text typed in and other data from event.target } function handleSubmit() { // on click doesn't usually need event data } return ( <div> <input type="text" name="myInput" onChange={handleChange} /> <button onClick={handleSubmit}>Submit</button> </div> ); } ``` ## React Hooks ### State and useState - useState 為我們提供 function component 中的本地狀態 ``` import React from 'react'; // create state variable // syntax: const [stateVariable] = React.useState(defaultValue); function App() { const [language] = React.useState('javascript'); // we use array destructuring to declare state variable return <div>I am learning {language}</div>; } ``` - 注意:此段文章中的任何 hook 都來自 React 套件,且可以單獨導入 ``` import React, { useState } from "react"; function App() { const [language] = useState("javascript"); return <div>I am learning {language}</div>; } ``` - useState 還為我們提供了一個 `setter` 函數來更新它的狀態 ``` function App() { // the setter function is always the second destructured value const [language, setLanguage] = React.useState("python"); // the convention for the setter name is 'setStateVariable' return ( <div> {/* why use an arrow function here instead onClick={setterFn()} ? */} <button onClick={() => setLanguage("javascript")}> Change language to JS </button> {/* if not, setLanguage would be called immediately and not on click */} <p>I am now learning {language}</p> </div> ); } // note that whenever the setter function is called, the state updates, // and the App component re-renders to display the new state ``` - useState 可以在單個元件中使用一次或多次 ``` function App() { const [language, setLanguage] = React.useState("python"); const [yearsExperience, setYearsExperience] = React.useState(0); return ( <div> <button onClick={() => setLanguage("javascript")}> Change language to JS </button> <input type="number" value={yearsExperience} onChange={event => setYearsExperience(event.target.value)} /> <p>I am now learning {language}</p> <p>I have {yearsExperience} years of experience</p> </div> ); } ``` - useState 可以接受 primitive value 或物件 ``` // we have the option to organize state using whatever is the // most appropriate data type, according to the data we're tracking function App() { const [developer, setDeveloper] = React.useState({ language: "", yearsExperience: 0 }); function handleChangeYearsExperience(event) { const years = event.target.value; // we must pass in the previous state object we had with the spread operator setDeveloper({ ...developer, yearsExperience: years }); } return ( <div> {/* no need to get prev state here; we are replacing the entire object */} <button onClick={() => setDeveloper({ language: "javascript", yearsExperience: 0 }) } > Change language to JS </button> {/* we can also pass a reference to the function */} <input type="number" value={developer.yearsExperience} onChange={handleChangeYearsExperience} /> <p>I am now learning {developer.language}</p> <p>I have {developer.yearsExperience} years of experience</p> </div> ); } ``` - 如果新狀態依賴於之前的狀態,我們可以在 setter 函數中使用一個函數來為我們提供之前狀態的值 ``` function App() { const [developer, setDeveloper] = React.useState({ language: "", yearsExperience: 0, isEmployed: false }); function handleToggleEmployment(event) { // we get the previous state variable's value in the parameters // we can name 'prevState' however we like setDeveloper(prevState => { return { ...prevState, isEmployed: !prevState.isEmployed }; // it is essential to return the new state from this function }); } return ( <button onClick={handleToggleEmployment}>Toggle Employment Status</button> ); } ``` ### side effects 和 useEffect - useEffect 讓我們在 function component 中執行副作用。什麼是副作用? - 副作用是我們需要接觸外部世界的地方。例如,從 API 獲取資料或操作 DOM。 - 副作用是可能以不可預測的方式,改變我們元件狀態的操作(導致「副作用」)。 - useEffect 接受 callback function(稱為 effect 函數),預設情況下,每次重新渲染時都會運行 - useEffect 在我們的元件載入後運行,可以準確在元件的各個生命週期觸發 ``` // what does our code do? Picks a color from the colors array // and makes it the background color function App() { const [colorIndex, setColorIndex] = React.useState(0); const colors = ["blue", "green", "red", "orange"]; // we are performing a 'side effect' since we are working with an API // we are working with the DOM, a browser API outside of React useEffect(() => { document.body.style.backgroundColor = colors[colorIndex]; }); // whenever state is updated, App re-renders and useEffect runs function handleChangeIndex() { const next = colorIndex + 1 === colors.length ? 0 : colorIndex + 1; setColorIndex(next); } return <button onClick={handleChangeIndex}>Change background color</button>; } ``` - 如果不想在每次渲染後都執行 callback function,可以在第二個參數傳一個空陣列 ``` function App() { ... // now our button doesn't work no matter how many times we click it... useEffect(() => { document.body.style.backgroundColor = colors[colorIndex]; }, []); // the background color is only set once, upon mount // how do we not have the effect function run for every state update... // but still have it work whenever the button is clicked? return ( <button onClick={handleChangeIndex}> Change background color </button> ); } ``` - useEffect 讓我們能夠在 dependencies 陣列的內容改變時,才執行 - dependencies 陣列是第二個參數,如果陣列中的任何一個值發生變化,effect function 將再次運行 ``` function App() { const [colorIndex, setColorIndex] = React.useState(0); const colors = ["blue", "green", "red", "orange"]; // we add colorIndex to our dependencies array // when colorIndex changes, useEffect will execute the effect fn again useEffect(() => { document.body.style.backgroundColor = colors[colorIndex]; // when we use useEffect, we must think about what state values // we want our side effect to sync with }, [colorIndex]); function handleChangeIndex() { const next = colorIndex + 1 === colors.length ? 0 : colorIndex + 1; setColorIndex(next); } return <button onClick={handleChangeIndex}>Change background color</button>; } ``` - 可以在 useEffect 最後回傳一個函數,來取消訂閱某些效果 ``` function MouseTracker() { const [mousePosition, setMousePosition] = useState({ x: 0, y: 0 }); React.useEffect(() => { // .addEventListener() sets up an active listener... window.addEventListener("mousemove", handleMouseMove); // ...so when we navigate away from this page, it needs to be // removed to stop listening. Otherwise, it will try to set // state in a component that doesn't exist (causing an error) // We unsubscribe any subscriptions / listeners w/ this 'cleanup function' return () => { window.removeEventListener("mousemove", handleMouseMove); }; }, []); function handleMouseMove(event) { setMousePosition({ x: event.pageX, y: event.pageY }); } return ( <div> <h1>The current mouse position is:</h1> <p> X: {mousePosition.x}, Y: {mousePosition.y} </p> </div> ); } // Note: we could extract the reused logic in the callbacks to // their own function, but I believe this is more readable ``` - 使用 useEffect 撈取資料 - 請注意,如果要使用更簡潔的 async/awit 語法處理 promise,則需要建立一個單獨的函數(因為 effect callback function 不能是 async) ``` const endpoint = "https://api.github.com/users/codeartistryio"; // with promises: function App() { const [user, setUser] = React.useState(null); React.useEffect(() => { // promises work in callback fetch(endpoint) .then(response => response.json()) .then(data => setUser(data)); }, []); } // with async / await syntax for promise: function App() { const [user, setUser] = React.useState(null); // cannot make useEffect callback function async React.useEffect(() => { getUser(); }, []); // instead, use async / await in separate function, then call // function back in useEffect async function getUser() { const response = await fetch("https://api.github.com/codeartistryio"); const data = await response.json(); setUser(data); } } ``` ### 效能和 useCallback - useCallback 是一個用來提高元件性能的 hook - 如果你有一個經常重新渲染的元件,useCallback 可以防止元件內的 callback function 在每次元件重新渲染時都重新創建(導致元件重新運行) - useCallback 僅在其依賴項之一改變時重新運行 ``` // in Timer, we are calculating the date and putting it in state a lot // this results in a re-render for every state update // we had a function handleIncrementCount to increment the state 'count'... function Timer() { const [time, setTime] = React.useState(); const [count, setCount] = React.useState(0); // ... but unless we wrap it in useCallback, the function is // recreated for every single re-render (bad performance hit) // useCallback hook returns a callback that isn't recreated every time const inc = React.useCallback( function handleIncrementCount() { setCount(prevCount => prevCount + 1); }, // useCallback accepts a second arg of a dependencies array like useEffect // useCallback will only run if any dependency changes (here it's 'setCount') [setCount] ); React.useEffect(() => { const timeout = setTimeout(() => { const currentTime = JSON.stringify(new Date(Date.now())); setTime(currentTime); }, 300); return () => { clearTimeout(timeout); }; }, [time]); return ( <div> <p>The current time is: {time}</p> <p>Count: {count}</p> <button onClick={inc}>+</button> </div> ); } ``` ### Memoization 和 useMemo - useMemo 和 useCallback 非常相似,都是為了提高效能。但就不是為了暫存 callback function,而是為了暫存耗費效能的運算結果 - useMemo 允許我們「記憶」已經為某些輸入進行了昂貴計算的結果(之後就不用重新計算了) - useMemo 從計算中回傳一個值,而不是 callback 函數(但可以是一個函數) ``` // useMemo is useful when we need a lot of computing resources // to perform an operation, but don't want to repeat it on each re-render function App() { // state to select a word in 'words' array below const [wordIndex, setWordIndex] = useState(0); // state for counter const [count, setCount] = useState(0); // words we'll use to calculate letter count const words = ["i", "am", "learning", "react"]; const word = words[wordIndex]; function getLetterCount(word) { // we mimic expensive calculation with a very long (unnecessary) loop let i = 0; while (i < 1000000) i++; return word.length; } // Memoize expensive function to return previous value if input was the same // only perform calculation if new word without a cached value const letterCount = React.useMemo(() => getLetterCount(word), [word]); // if calculation was done without useMemo, like so: // const letterCount = getLetterCount(word); // there would be a delay in updating the counter // we would have to wait for the expensive function to finish function handleChangeIndex() { // flip from one word in the array to the next const next = wordIndex + 1 === words.length ? 0 : wordIndex + 1; setWordIndex(next); } return ( <div> <p> {word} has {letterCount} letters </p> <button onClick={handleChangeIndex}>Next word</button> <p>Counter: {count}</p> <button onClick={() => setCount(count + 1)}>+</button> </div> ); } ``` ### Refs 和 useRef - Refs 是所有 React 組件都可用的特殊屬性。允許我們在元件安裝時,創建針對特定元素/元件的引用 - useRef 允許我們輕鬆使用 React refs - 我們在元件開頭呼叫 useRef,並將回傳值附加到元素的 ref 屬性來引用它 - 一旦我們創建了一個引用,就可以用它來改變元素的屬性,或者可以呼叫該元素上的任何可用方法(比如用 .focus() 來聚焦) ``` function App() { const [query, setQuery] = React.useState("react hooks"); // we can pass useRef a default value // we don't need it here, so we pass in null to ref an empty object const searchInput = useRef(null); function handleClearSearch() { // current references the text input once App mounts searchInput.current.value = ""; // useRef can store basically any value in its .current property searchInput.current.focus(); } return ( <form> <input type="text" onChange={event => setQuery(event.target.value)} ref={searchInput} /> <button type="submit">Search</button> <button type="button" onClick={handleClearSearch}> Clear </button> </form> ); } ``` ## 進階 Hook ### Context 和 useContext - 在 React 中,盡量避免創建多個 props 來將資料從父元件向下傳遞兩個或多個層級 ``` // Context helps us avoid creating multiple duplicate props // This pattern is also called props drilling: function App() { // we want to pass user data down to Header const [user] = React.useState({ name: "Fred" }); return ( // first 'user' prop <Main user={user} /> ); } const Main = ({ user }) => ( <> {/* second 'user' prop */} <Header user={user} /> <div>Main app content...</div> </> ); const Header = ({ user }) => <header>Welcome, {user.name}!</header>; ``` - Context 有助於將 props 從父組件向下傳遞到多個層級的子元件 ``` // Here is the previous example rewritten with Context // First we create context, where we can pass in default values const UserContext = React.createContext(); // we call this 'UserContext' because that's what data we're passing down function App() { // we want to pass user data down to Header const [user] = React.useState({ name: "Fred" }); return ( {/* we wrap the parent component with the provider property */} {/* we pass data down the computer tree w/ value prop */} <UserContext.Provider value={user}> <Main /> </UserContext.Provider> ); } const Main = () => ( <> <Header /> <div>Main app content...</div> </> ); // we can remove the two 'user' props, we can just use consumer // to consume the data where we need it const Header = () => ( {/* we use this pattern called render props to get access to the data*/} <UserContext.Consumer> {user => <header>Welcome, {user.name}!</header>} </UserContext.Consumer> ); ``` - useContext hook 可以移除這種看起來很怪的 props 渲染模式,同時又能在各種 function component 中隨意使用 ``` const Header = () => { // we pass in the entire context object to consume it const user = React.useContext(UserContext); // and we can remove the Consumer tags return <header>Welcome, {user.name}!</header>; }; ``` ### Reducers 和 useReducer - Reducers 是簡單、可預測的(純)函數,它接受一個先前的狀態物件和一個動作物件,並回傳一個新的狀態物件。例如: ``` // let's say this reducer manages user state in our app: function reducer(state, action) { // reducers often use a switch statement to update state // in one way or another based on the action's type property switch (action.type) { // if action.type has the string 'LOGIN' on it case "LOGIN": // we get data from the payload object on action return { username: action.payload.username, isAuth: true }; case "SIGNOUT": return { username: "", isAuth: false }; default: // if no case matches, return previous state return state; } } ``` - Reducers 是一種強大的狀態管理模式,用於流行的 Redux 狀態管理套件(這套很常與 React 一起使用) - 與 useState(用於本地元件狀態)相比,Reducers 可以通過 useReducer hook 在 React 中使用,以便管理我們應用程序的狀態 - useReducer 可以與 useContext 配對來管理資料並輕鬆地將資料傳遞給元件 - useReducer + useContext 可以當成應用程式的整套狀態管理系統 ``` const initialState = { username: "", isAuth: false }; function reducer(state, action) { switch (action.type) { case "LOGIN": return { username: action.payload.username, isAuth: true }; case "SIGNOUT": // could also spread in initialState here return { username: "", isAuth: false }; default: return state; } } function App() { // useReducer requires a reducer function to use and an initialState const [state, dispatch] = useReducer(reducer, initialState); // we get the current result of the reducer on 'state' // we use dispatch to 'dispatch' actions, to run our reducer // with the data it needs (the action object) function handleLogin() { dispatch({ type: "LOGIN", payload: { username: "Ted" } }); } function handleSignout() { dispatch({ type: "SIGNOUT" }); } return ( <> Current user: {state.username}, isAuthenticated: {state.isAuth} <button onClick={handleLogin}>Login</button> <button onClick={handleSignout}>Signout</button> </> ); } ``` ### 編寫 custom hooks - 創建 hook 就能輕鬆在元件之間重用某些行為 - hook 是一種比以前的 class component 更容易理解的模式,例如 higher-order components 或者 render props - 根據需要,隨時可以自創一些 hook ``` // here's a custom hook that is used to fetch data from an API function useAPI(endpoint) { const [value, setValue] = React.useState([]); React.useEffect(() => { getData(); }, []); async function getData() { const response = await fetch(endpoint); const data = await response.json(); setValue(data); }; return value; }; // this is a working example! try it yourself (i.e. in codesandbox.io) function App() { const todos = useAPI("https://todos-dsequjaojf.now.sh/todos"); return ( <ul> {todos.map(todo => <li key={todo.id}>{todo.text}</li>)} </ul> ); } ``` ### Hooks 規則 - 使用 React hooks 有兩個核心規則,要遵守才能正常運作: - 1. Hooks 只能在元件開頭呼叫 - Hooks 不能放在條件、迴圈或嵌套函數中 - 2. Hooks只能在 function component 內部使用 - Hooks 不能放在普通的 JavaScript 函數或 class component 中 ``` function checkAuth() { // Rule 2 Violated! Hooks cannot be used in normal functions, only components React.useEffect(() => { getUser(); }, []); } function App() { // this is the only validly executed hook in this component const [user, setUser] = React.useState(null); // Rule 1 violated! Hooks cannot be used within conditionals (or loops) if (!user) { React.useEffect(() => { setUser({ isAuth: false }); // if you want to conditionally execute an effect, use the // dependencies array for useEffect }, []); } checkAuth(); // Rule 1 violated! Hooks cannot be used in nested functions return <div onClick={() => React.useMemo(() => doStuff(), [])}>Our app</div>; } ``` --- 以上是快速整理,希望對您有幫助!