mirror of
https://github.com/tiddly-gittly/TidGi-Desktop.git
synced 2025-12-06 02:30:47 -08:00
fa9751e5ea
* refactor: only use message id * feat: stream update, but react don't react on it dny * feat: start stop * feat: basic resoponse handlers * feat: load handler config schema * chore: upgrade to "moduleResolution": "bundler" * fix: prompt concat * feat: rjfs with mui * fix: mui v7 upgrade * fix: field editor tabs * feat: Description field is i18n key, use i18nAlly extension to see it on VSCode. And use react-i18next to translate it on frontend. * refactor: extract some shared components * refactor: remove @mui/lab * feat: faster editor * feat: beautify editor * refactor: better style * fix: fullscreen style * fix: array sort * fix: editor not saved * fix: broken array * chore: upgrade mui deps * Update type.d.ts * feat: upgrade and install react dev tools * refactor: simplify the code * refactor: simplify the code * refactor: simplify ai generated garbage code * fix: translated label * feat: translate enum and conditional show field based on enum * feat: click to open source * fix: ai generated garbage code to solve id and index problem * refactor: simplify * refactor: shorten * test: add jest * chore: node-linker=hoisted as electron forge suggested * test: e2e * fix: test failed on ci * test: faster by disable some typecheck * fix: ci * fix: ci * refactor: remove unused * fix: use macos and take screenshot in ci * docs: (cherry picked from commitb1a9706264) * fix: ci crash due to no dugite binary, due to NODE_ENV=test not set * Update test.yml * fix: ci * Update release.yml * docs: test * refactor: move folders and lint * refactor: organize pages to reduce level * refactor: merge page service and workspace service to allow drag & drop of pages * Update ipc-syncadaptor.ts * Update browserViewMetaData.ts * fix: view.webContents.loadURL won't reload on new electron version * fix: try to disable annoying useless disabling pasting "feature" https://github.com/electron/electron/issues/40995 * fix: initial page * feat: allow sort add icon * test: use vitest instead * refactor: use emotion instead * fix: all ts error after migrate to vitest and emotion * fix: Component selectors can only be used in conjunction with @emotion/babel-plugin, the swc Emotion plugin, or another Emotion-aware compiler transform. fix by not using this usage * fix: too many open files * test: add basic main page test * refactor: split workspace type to make it more specific * test: support mock react lazy import component by re-export them * test: sidebar * refactor: move mock to global * test: testing library fix * test: new wiki form * docs: test with vitest * lint: fix * docs: pronounication & remove toc as gh build in * feat: tools and rag * feat: prompt for build-in tools * fix: i18n * fix: tool using * test: fix * refactor: Auto-create default wiki workspace if none exists (handled in backend) * fix: create wiki workspace so it is on first one * refactor: remove some useless feature * refactor: use plugin instead of handler * chore: make concatPrompt async iterator * feat: show progress on frontend * fix: errors * fix: ConfigError: Config (unnamed): Key "overrides": This appears to be in eslintrc format rather than flat config format. * Update package.json * feat: allow windows to hide titlebar * fix: logger error when ctrl+c on mac * lint * refactor: use plugin everywhere * refactor: move more logic to plugin * refactor: run tool in plugin, simplify abstraction * refactor: remove ai generated duplicate * refactor * refactor: less plugins * test: simplify wiki search test * test: remove any * fix: not streaming, tool order wrong * test: plugin system and streaming * refactor: remove useless auto reply plugin * fix: hide duration expired tool calling message, so long tool result only show to ai once * fix: $ props should lower cased * test: support run as electron so can use sqlite3 compiled bin for electron * docs: better electron as node run and doc * test: restore to use threads instead of fock. Seems this also works for inmemory database * test: fix frontend ui test * lint: ai content * test: fix coverage * test: Refactor test mocks to dedicated __mocks__ directory Moved common test mocks from setup-vitest.ts into separate files under src/__tests__/__mocks__ for better organization and maintainability. Updated documentation to reflect the new structure. Removed fileMock.js and updated setup-vitest.ts to import and use the new centralized mocks. * Update ErrorDuringStart.md * Update messageManagementPlugin.test.ts * test: Fix Electron test process cleanup and update test config Update test:unit script to use cross-env for ELECTRON_RUN_AS_NODE, ensuring child processes inherit the variable and preventing resource leaks. Remove manual process.env setting from vitest.config.ts and add documentation on handling related errors in Testing.md. Also, add 'hanging-process' reporter to Vitest config for improved diagnostics. * fix: warning in test * Create AgentInstanceWorkflow.md * fix: duration bug * fix: ai not response for tool result * test: Add agent workflow feature and step definitions Introduces a new Cucumber feature for agent workflow, including multi-round conversation and tool usage. Adds agent-specific step definitions for message input and chat history validation. Refactors application step definitions for improved selector handling and element interaction. Updates test scripts in package.json to include a preparation step for e2e tests. * Update application.ts * test: Add mock OpenAI server and tests Introduces a MockOpenAIServer for simulating OpenAI chat completions, including tool call and tool result responses. Adds corresponding tests and updates vitest config to include tests in the features directory. * test: simplify steps * test: Add separate test userData and wiki folders Introduces 'userData-test' and 'wiki-test' folders for test environments, updating appPaths, fileNames, and paths constants to distinguish between development and test modes. This helps prevent conflicts when running test and development instances simultaneously. * Update eslint.config.mjs @typescript-eslint/no-unnecessary-condition': 'off' * test: Add data-testid attributes for test automation Added data-testid attributes to form inputs and buttons in ExternalAPI components to improve test automation reliability. Updated feature files and step definitions to use these selectors, refined window switching logic, and improved timing for UI interactions. Also exposed isElectronDevelopment and added isMainWindowPage utility for window identification. * test: fix wront page type by ai written garbage code * Update application.ts * Update Testing.md * fix: no log during e2e test, and error creating wiki blocks other init steps * test: click agent workspace button, and refactor mock server * test: mock streamable openai api * rename * chore: try esbuild loader, renderer build too slow * chore: organize webpack * chore: ignore service code from hot reload * Update Testing.md * test: use full agentinstance in test * chore: webpack error * Update Testing.md * chore: remove useless spectron * test: EsbuildPlugin's `define` doesn't work, it won't set env properly. * test: e2e mock openai * lint: disable @typescript-eslint/require-await * test: Add quick access to create default agent tab Introduces a 'Create Default Agent' quick access button in the Agent New Tab page, with localization support. Adds utility to close all tabs and create a default agent tab for fallback scenarios, improves test selectors for tab and close actions, and refactors agent chat tab creation logic for consistency and testability. * feat: remove unuse favorite * feat: Add wiki operation and workspaces list plugins Introduces wikiOperationPlugin and workspacesListPlugin for agent instance prompt and response handling. Updates plugin registry, test coverage, and default agent configuration to support wiki workspace listing and wiki note operations (create, update, delete) via tool calls. Refactors wikiSearchPlugin to delegate workspace list injection to workspacesListPlugin. * Refactor plugin schema system for dynamic registration Introduces a dynamic plugin schema registry, allowing plugins to register their parameter schemas and metadata at runtime. Refactors prompt concat schema generation to use dynamically registered plugin schemas, removes static plugin schema definitions, and updates all plugin files to export their parameter schemas. Adds new modelContextProtocolPlugin and schemaRegistry modules, and updates plugin initialization to register schemas and metadata. This enables extensibility and type safety for plugin configuration and validation. * refactor: move PromptConfigForm to inside PromptPreviewDialog * test: frontent render tool usage info * test: split file * test: fix error * test: wiki operation * test: remove log and clean up test deps * Update i18next-electron-fs-backend.ts * fix: wikiOperationInServer not called due to no message * test: fix * test: Refactor agent feature setup and improve mock server URL handling Moved AI provider and model configuration to a shared setup scenario in agent.feature to avoid redundant steps in each test. Enhanced application.ts to use a fallback localhost URL for MOCK_SERVER_URL when the mock server is not available, improving test reliability. * Remove retry logic from basicPromptConcatHandler Eliminated retryCount and maxRetries from basicPromptConcatHandler, simplifying the control flow and removing the retry limit for LLM calls. Logging and yield logic were updated to reflect the removal of retries. * Update agent.feature * test: agent and default wiki * test: refactor wiki cleanup * Update agent.feature * Refactor AI settings setup and add preference feature Moved AI provider and model configuration steps from agent.feature to a new preference.feature file for better separation of concerns. Updated step definitions in agent.ts to improve robustness when reading and writing settings files, including type usage and directory checks. * test: try debug can't stop bug * Update Testing.md * fix: cancel agent not update cancel button * refactor: update Frontend use `void window.service.native.log` to log to file. * test: log from renderer test * feat: add default embedding model config and victor search service and search preference panel * test: default embedding form and * test: default agent * refator: unused tool listing methods * Refactor test database setup for integration tests Centralizes in-memory SQLite test database initialization and cleanup in shared utilities for all integration tests. Updates agentDefinition and messageManagementPlugin tests to use the shared test database, improving reliability and reducing code duplication. * fix: app path wrong in unit test * feat: externalAPIDebug * test: embedding service and let db use real in memory one * test: provide at least one tab for close tab test * fix: victor db not loaded in new connection * test: disable hot reload * Update DeveloperTools.tsx * feat: tool message & wiki tool schema * feat: pref to open db * chore: skip ExternalsPlugin on dev * fix: APIs doesn't accept 'tool' role, and it won't return anything when API calls * fix: docs and remove ai fake fix * refactor: get agent list don't need to have message * Update basicPromptConcatHandler.failure.test.ts * Update basicPromptConcatHandler.test.ts * fix: role wrong cause e2e failed * test: allow e2e create .db file to check * feat: create new agent * feat: getAgentDefinitionTemplatesFromWikis * fix: Prevent update non-active (hiding) wiki workspace, so it won't pop up to cover other active agent workspace * fix: Ensure the config change is fully persisted before proceeding * test: Don't bring up window when running e2e test, otherwise it will annoy the developer who is doing other things. * feat: Edit existing agent definition workflow * Update StyledArrayContainer.tsx * test: prevent mock server hang * Refactor UI step definitions into separate file Moved UI-related Cucumber step definitions from application.ts to a new ui.ts file for better separation of concerns and maintainability. application.ts now only contains application-specific logic. * lint: ai auto fix * Clean up feature files and improve test coverage Removed unnecessary blank lines and improved formatting in feature files for better readability. Updated tsconfig to use 'bundler' module resolution. Enhanced EditAgentDefinitionContent test to mock and verify console.error output. Added type annotation for IAgentDefinitionService in basicPromptConcatHandler.failure.test.ts for improved type safety. * test: simplify message text match * feat: transcription and image generation model config * fix: style props error * chore: remove unused file * chore: try to imporove dev start speed but no sig improvment Replaces 'pnpm start' with 'pnpm start:init' for initial setup and documents slow startup in development. Adds a debug Webpack script, disables polling in renderer file watching, and only enables CircularDependencyPlugin in production/CI for faster dev builds. WebpackBar now only shows when DEBUG includes 'electron-forge:*'. ForkTsCheckerWebpackPlugin is now configured for async checks with memory limits and test file exclusion. Updates documentation to reflect these changes. * docs: why not vite * refactor: basic vite usage, but wiki worker is not running and view is not showing * refactor: remove lazy inject so vite works and wiki worker works, wiki in browser view can loads * Replace electron-squirrel-startup with inline implementation Added a custom Squirrel event handler in src/helpers/squirrelStartup.ts to handle Windows install/update/uninstall events, replacing the electron-squirrel-startup package. This avoids ESM/CommonJS compatibility issues and simplifies event handling in src/main.ts. * refactor: vite build and test * fix: e2e still use dev wiki folder * fix: provide env * chore: "tiddlywiki": "5.3.7" * Apply suggestion from @Copilot Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com> * Apply suggestion from @Copilot Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com> * Apply suggestion from @Copilot Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com> * Apply suggestion from @Copilot Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com> * fix: missing i18n * test: Module did not self-register: '/home/runner/work/TidGi-Desktop/TidGi-Desktop/node_modules/better-sqlite3/build/Release/better_sqlite3.node'. * feat: i18n * feat: zh-Hant i18n * Update test.yml * feat: zh-Hans in test * test: i18n * refactor: ts forge config * lint: fix * chore: update wiki * Update pnpm-lock.yaml * chore: update github action versions * Update wiki * fix: pnpm then node * fix: Multiple versions of pnpm specified * Update test.yml * fix: Failed to take screenshot: page.screenshot: Target page, context or browser has been closed * chore: CodeQL Action major versions v1 and v2 have been deprecated. * chore: parallel codeql * Update test.yml * fix: i18n test not passing * test: step screenshot in each folder * fix: can't unzip, may due to file path * test: increase timeout in CI * docs: more log about wiki creation, and add log to criticial path * Refactor: logging for structured and consistent output Replaces string-based logger messages with structured logging throughout the codebase, providing function names, error messages, and relevant context as objects. This improves log readability, enables better filtering and searching, and standardizes error and debug reporting across services. * chore: debug wiki copy * fix: wiki submodule not cloned * Revert "test: increase timeout in CI" This reverts commiteff8583a01. * test: reduce wait time, because most check already will wait for a while * test: batch some e2e steps to reduce screenshot count * Update index.ts * chore: remove webpack files, use vite plugins --------- Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
165 lines
17 KiB
Gherkin
165 lines
17 KiB
Gherkin
Feature: Vector Search - Embedding Generation and Semantic Search
|
||
As a user
|
||
I want to use vector database to perform semantic search in my wiki
|
||
So that I can find relevant content based on meaning rather than exact keywords
|
||
|
||
Background:
|
||
Given I add test ai settings
|
||
Then I launch the TidGi application
|
||
And I wait for the page to load completely
|
||
And I should see a "page body" element with selector "body"
|
||
# Ensure we are in the agent workspace (not wiki workspace) for agent interaction
|
||
When I click on an "agent workspace button" element with selector "[data-testid='workspace-agent']"
|
||
And I should see a "new tab button" element with selector "[data-tab-id='new-tab-button']"
|
||
|
||
@vectorSearch @mockOpenAI
|
||
Scenario: Agent workflow - Create notes, update embeddings, then search
|
||
Given I have started the mock OpenAI server
|
||
| response | stream | embedding |
|
||
| <tool_use name="wiki-operation">{"workspaceName":"wiki","operation":"wiki-add-tiddler","title":"AI Agent Guide","text":"智能体是一种可以执行任务的AI系统，它可以使用工具、搜索信息并与用户交互。"}</tool_use> | false | |
|
||
| 已成功在工作区 wiki 中创建条目 "AI Agent Guide"。 | false | |
|
||
| <tool_use name="wiki-operation">{"workspaceName":"wiki","operation":"wiki-add-tiddler","title":"Vector Database Tutorial","text":"向量数据库用于存储和检索高维向量数据，支持语义搜索和相似度匹配。"}</tool_use> | false | |
|
||
| 已成功在工作区 wiki 中创建条目 "Vector Database Tutorial"。 | false | |
|
||
| <tool_use name="wiki-update-embeddings">{"workspaceName":"wiki","forceUpdate":false}</tool_use> | false | |
|
||
| | false | note1 |
|
||
| | false | note2 |
|
||
| 已成功为工作区 wiki 生成向量嵌入索引。总计2个笔记，2个嵌入向量。 | false | |
|
||
| <tool_use name="wiki-search">{"workspaceName":"wiki","searchType":"vector","query":"如何使用AI智能体","limit":5,"threshold":0.7}</tool_use> | false | |
|
||
| | false | query-note1 |
|
||
| 根据向量搜索结果，在工作区 wiki 中找到以下相关内容：\n\n**Tiddler: AI Agent Guide** (Similarity: 95.0%)\n这篇笔记介绍了AI智能体的基本概念和使用方法。 | false | |
|
||
# Step 1: Open agent chat interface
|
||
When I click on a "new tab button" element with selector "[data-tab-id='new-tab-button']"
|
||
And I should see a "search interface" element with selector ".aa-Autocomplete"
|
||
When I click on a "search input box" element with selector ".aa-Input"
|
||
And I should see an "autocomplete panel" element with selector ".aa-Panel"
|
||
When I click on an "agent suggestion" element with selector '[data-autocomplete-source-id="agentsSource"] .aa-ItemWrapper'
|
||
# Step 2: Create first note
|
||
When I click on a "message input textarea" element with selector "[data-testid='agent-message-input']"
|
||
When I type "在 wiki 工作区创建一个名为 AI Agent Guide 的笔记，内容是：智能体是一种可以执行任务的AI系统，它可以使用工具、搜索信息并与用户交互。" in "chat input" element with selector "[data-testid='agent-message-input']"
|
||
And I press "Enter" key
|
||
Then I should see 4 messages in chat history
|
||
# Step 3: Create second note
|
||
When I click on a "message input textarea" element with selector "[data-testid='agent-message-input']"
|
||
When I type "再创建一个名为 Vector Database Tutorial 的笔记，内容是：向量数据库用于存储和检索高维向量数据，支持语义搜索和相似度匹配。" in "chat input" element with selector "[data-testid='agent-message-input']"
|
||
And I press "Enter" key
|
||
Then I should see 8 messages in chat history
|
||
# Step 4: Update vector embeddings using agent tool
|
||
When I click on a "message input textarea" element with selector "[data-testid='agent-message-input']"
|
||
When I type "为 wiki 工作区更新向量索引" in "chat input" element with selector "[data-testid='agent-message-input']"
|
||
And I press "Enter" key
|
||
Then I should see 12 messages in chat history
|
||
# Step 5: Perform vector search
|
||
When I click on a "message input textarea" element with selector "[data-testid='agent-message-input']"
|
||
When I type "使用向量搜索在 wiki 中查找关于如何使用AI智能体的内容" in "chat input" element with selector "[data-testid='agent-message-input']"
|
||
And I press "Enter" key
|
||
Then I should see 16 messages in chat history
|
||
# Verify the last message contains vector search results
|
||
And I should see "search result in last message" elements with selectors:
|
||
| [data-testid='message-bubble']:last-child:has-text('Tiddler: AI Agent Guide') |
|
||
|
||
@vectorSearch @mockOpenAI
|
||
Scenario: UI workflow - Generate embeddings via preferences, then search
|
||
Given I have started the mock OpenAI server
|
||
| response | stream | embedding |
|
||
| <tool_use name="wiki-operation">{"workspaceName":"wiki","operation":"wiki-add-tiddler","title":"Machine Learning Basics","text":"机器学习是人工智能的一个分支，通过算法让计算机从数据中学习规律。"}</tool_use> | false | |
|
||
| 已成功在工作区 wiki 中创建条目 "Machine Learning Basics"。 | false | |
|
||
| | false | note3 |
|
||
| <tool_use name="wiki-search">{"workspaceName":"wiki","searchType":"vector","query":"机器学习","limit":5,"threshold":0.7}</tool_use> | false | |
|
||
| | false | query-note3 |
|
||
| 根据向量搜索结果，在工作区 wiki 中找到以下相关内容：\n\n**Tiddler: Machine Learning Basics** (Similarity: 98.0%)\n这篇笔记介绍了机器学习的基本概念。 | false | |
|
||
# Step 1: Create a test note via agent
|
||
When I click on "new tab button and create default agent button" elements with selectors:
|
||
| [data-tab-id='new-tab-button'] |
|
||
| [data-testid='create-default-agent-button'] |
|
||
And I should see a "message input box" element with selector "[data-testid='agent-message-input']"
|
||
When I click on a "message input textarea" element with selector "[data-testid='agent-message-input']"
|
||
When I type "在 wiki 工作区创建一个名为 Machine Learning Basics 的笔记，内容是：机器学习是人工智能的一个分支，通过算法让计算机从数据中学习规律。" in "chat input" element with selector "[data-testid='agent-message-input']"
|
||
And I press "Enter" key
|
||
Then I should see 4 messages in chat history
|
||
# Step 2: Open preferences and manually generate embeddings via UI
|
||
When I click on a "settings button" element with selector "#open-preferences-button"
|
||
When I switch to "preferences" window
|
||
# Navigate to Search section (which contains vector database settings)
|
||
When I click on a "search section" element with selector "[data-testid='preference-section-search']"
|
||
# Wait for workspace list to load
|
||
# The Search.tsx renders workspace cards with name, status, and buttons
|
||
And I should see a "wiki workspace card" element with selector "*:has-text('wiki')"
|
||
# Click the generate button - use button text "生成" instead of data-testid
|
||
# The button shows "生成" for initial generation, "更新嵌入" after generation
|
||
When I click on a "generate button with text" element with selector "button:has-text('生成')"
|
||
# Verify generation completed with detailed status information
|
||
# Should show: workspace name, embedding count, note count, last updated time and action buttons
|
||
Then I should see "workspace name in status and embedding count status and embedding word and last updated label and update button after generation and delete button after generation" elements with selectors:
|
||
| *:has-text('wiki') |
|
||
| *:has-text('个笔记') |
|
||
| *:has-text('嵌入') |
|
||
| *:has-text('最后更新') |
|
||
| button:has-text('更新嵌入') |
|
||
| button:has-text('删除') |
|
||
# Close preferences
|
||
When I close "preferences" window
|
||
And I switch to "main" window
|
||
# Step 3: Perform vector search and verify results match agent workflow
|
||
When I click on a "message input textarea" element with selector "[data-testid='agent-message-input']"
|
||
When I type "使用向量搜索在 wiki 中查找关于机器学习的内容" in "chat input" element with selector "[data-testid='agent-message-input']"
|
||
And I press "Enter" key
|
||
Then I should see 8 messages in chat history
|
||
# Verify the last message contains vector search results
|
||
And I should see a "ML search result in last message" element with selector "[data-testid='message-bubble']:last-child:has-text('Tiddler: Machine Learning Basics')"
|
||
|
||
@vectorSearch @mockOpenAI
|
||
Scenario: Vector search with low similarity - No results below threshold, then lower threshold
|
||
Given I have started the mock OpenAI server
|
||
| response | stream | embedding |
|
||
| <tool_use name="wiki-operation">{"workspaceName":"wiki","operation":"wiki-add-tiddler","title":"AI Technology","text":"人工智能技术正在改变世界。"}</tool_use> | false | |
|
||
| 已成功在工作区 wiki 中创建条目 "AI Technology"。 | false | |
|
||
| <tool_use name="wiki-operation">{"workspaceName":"wiki","operation":"wiki-add-tiddler","title":"Machine Learning","text":"机器学习算法和应用。"}</tool_use> | false | |
|
||
| 已成功在工作区 wiki 中创建条目 "Machine Learning"。 | false | |
|
||
| <tool_use name="wiki-update-embeddings">{"workspaceName":"wiki","forceUpdate":false}</tool_use> | false | |
|
||
| | false | note4 |
|
||
| | false | note5 |
|
||
| 已成功为工作区 wiki 生成向量嵌入索引。总计2个笔记，2个嵌入向量。 | false | |
|
||
| <tool_use name="wiki-search">{"workspaceName":"wiki","searchType":"vector","query":"天气预报","limit":5,"threshold":0.7}</tool_use> | false | |
|
||
| | false | unrelated |
|
||
| 在Wiki工作空间"wiki"中未找到符合条件的向量搜索结果（相似度阈值：0.7）。 | false | |
|
||
| <tool_use name="wiki-search">{"workspaceName":"wiki","searchType":"vector","query":"天气预报","limit":5,"threshold":0.1}</tool_use> | false | |
|
||
| | false | unrelated |
|
||
| 根据向量搜索结果，在工作区 wiki 中找到以下相关内容：\n\n**Tiddler: AI Technology** (Similarity: 15.0%)\n低相似度结果。 | false | |
|
||
# Step 1: Open agent chat interface
|
||
When I click on "new tab button and create default agent button" elements with selectors:
|
||
| [data-tab-id='new-tab-button'] |
|
||
| [data-testid='create-default-agent-button'] |
|
||
And I should see a "message input box" element with selector "[data-testid='agent-message-input']"
|
||
# Step 2: Create first note about AI
|
||
When I click on a "message input textarea" element with selector "[data-testid='agent-message-input']"
|
||
When I type "在 wiki 工作区创建一个名为 AI Technology 的笔记，内容是：人工智能技术正在改变世界。" in "chat input" element with selector "[data-testid='agent-message-input']"
|
||
And I press "Enter" key
|
||
Then I should see 4 messages in chat history
|
||
# Step 3: Create second note about ML
|
||
When I click on a "message input textarea" element with selector "[data-testid='agent-message-input']"
|
||
When I type "再创建一个名为 Machine Learning 的笔记，内容是：机器学习算法和应用。" in "chat input" element with selector "[data-testid='agent-message-input']"
|
||
And I press "Enter" key
|
||
Then I should see 8 messages in chat history
|
||
# Step 4: Update vector embeddings
|
||
When I click on a "message input textarea" element with selector "[data-testid='agent-message-input']"
|
||
When I type "为 wiki 工作区更新向量索引" in "chat input" element with selector "[data-testid='agent-message-input']"
|
||
And I press "Enter" key
|
||
Then I should see 12 messages in chat history
|
||
# Step 5: Search for unrelated content with high threshold (should find nothing)
|
||
When I click on a "message input textarea" element with selector "[data-testid='agent-message-input']"
|
||
When I type "使用向量搜索在 wiki 中查找关于天气预报的内容，阈值设为0.7" in "chat input" element with selector "[data-testid='agent-message-input']"
|
||
And I press "Enter" key
|
||
Then I should see 16 messages in chat history
|
||
# Verify the 16th message contains "no results found" with threshold info
|
||
Then I should see "no results in 16th message and threshold 0.7 in 16th message" elements with selectors:
|
||
| [data-testid='message-bubble']:nth-child(16):has-text('未找到符合条件') |
|
||
| [data-testid='message-bubble']:nth-child(16):has-text('0.7') |
|
||
# Step 6: Lower threshold and search again (should find low-similarity results)
|
||
When I click on a "message input textarea" element with selector "[data-testid='agent-message-input']"
|
||
When I type "再次搜索天气预报，但这次把阈值降低到0.1" in "chat input" element with selector "[data-testid='agent-message-input']"
|
||
And I press "Enter" key
|
||
Then I should see 20 messages in chat history
|
||
# Verify the 20th message contains low-similarity result
|
||
Then I should see "AI Technology and low similarity in 20th message" elements with selectors:
|
||
| [data-testid='message-bubble']:nth-child(20):has-text('Tiddler: AI Technology') |
|
||
| [data-testid='message-bubble']:nth-child(20):has-text('15') |
|