Close [AI-75](https://linear.app/affine-design/issue/AI-75)
<!-- This is an auto-generated comment: release notes by coderabbit.ai -->
## Summary by CodeRabbit
- **New Features**
- Improved formatting for AI-generated reasoning and web search results, displaying them as callout blocks for enhanced readability.
- Expanded support for rendering callout blocks in the text renderer.
- **Style**
- Adjusted layout to ensure callout blocks retain appropriate spacing and appearance.
- **Refactor**
- Simplified and unified the integration of web search tools for AI providers.
<!-- end of auto-generated comment: release notes by coderabbit.ai -->
Close [AI-85](https://linear.app/affine-design/issue/AI-85)
<!-- This is an auto-generated comment: release notes by coderabbit.ai -->
## Summary by CodeRabbit
- **New Features**
- Enhanced AI chat experience with improved web search integration, including a new "AUTO" mode that allows the AI to search the web only when needed.
- Updated AI model for chat to "o4-mini" for improved performance.
- **Improvements**
- Refined instructions to reduce hallucinations and ensure the AI admits uncertainty when unsure.
- Enhanced streaming responses to display web search queries and results more clearly.
- Updated web search tool modes for more accurate and flexible search behavior.
- Centralized AI provider options and improved tool selection logic for better response quality.
- **Bug Fixes**
- Improved handling of tool calls and reasoning steps in AI chat responses.
<!-- end of auto-generated comment: release notes by coderabbit.ai -->
<!-- This is an auto-generated comment: release notes by coderabbit.ai -->
## Summary by CodeRabbit
- **New Features**
- Updated the AI model used for the "Chat With AFFiNE AI" feature to enhance chat capabilities.
<!-- end of auto-generated comment: release notes by coderabbit.ai -->
<!-- This is an auto-generated comment: release notes by coderabbit.ai -->
## Summary by CodeRabbit
- **Chores**
- Updated test setup to automatically skip onboarding steps during environment initialization.
- Simplified test utility methods by removing notification handling logic from chat panel and editor mode switching processes.
<!-- end of auto-generated comment: release notes by coderabbit.ai -->
Close [AI-59](https://linear.app/affine-design/issue/AI-59)
<!-- This is an auto-generated comment: release notes by coderabbit.ai -->
## Summary by CodeRabbit
- **New Features**
- Added support for two new providers, Anthropic and Exa, in test environments and configurations.
- **Chores**
- Updated workflow and test environment setup to include API keys for Anthropic and Exa providers.
- **Refactor**
- Changed the AI model used for the "Chat With AFFiNE AI" prompt to "claude-3-7-sonnet-20250219".
<!-- end of auto-generated comment: release notes by coderabbit.ai -->
Support [AI-60](https://linear.app/affine-design/issue/AI-60).
<!-- This is an auto-generated comment: release notes by coderabbit.ai -->
## Summary by CodeRabbit
- **Refactor**
- Updated naming conventions for the web search tool to ensure consistency across AI chat features.
- Simplified internal handling of web search tool options for a more streamlined experience.
- Unified parameter naming from "mustSearch" to "webSearch" across AI chat inputs and actions.
- **Chores**
- Removed unused configuration options related to web search from prompt settings.
- Added support for enabling or disabling web search via new parameters in chat requests.
<!-- end of auto-generated comment: release notes by coderabbit.ai -->
Close [AI-81](https://linear.app/affine-design/issue/AI-81).
<!-- This is an auto-generated comment: release notes by coderabbit.ai -->
## Summary by CodeRabbit
- **New Features**
- Added support for uploading and handling multiple images in the AI chat interface, with a maximum image limit enforced.
- Chat assistant avatar now visually indicates different statuses, including a transmitting animation.
- **Improvements**
- Enhanced file upload experience by allowing batch image uploads and clearer notifications for files exceeding size limits.
- Stronger status handling for chat assistant messages, improving reliability and user feedback.
- Unified image count limit across chat input components using a shared constant.
- **Bug Fixes**
- Improved consistency in how image limits are enforced across chat components.
<!-- end of auto-generated comment: release notes by coderabbit.ai -->
Close [AI-58](https://linear.app/affine-design/issue/AI-58).
<!-- This is an auto-generated comment: release notes by coderabbit.ai -->
## Summary by CodeRabbit
- **New Features**
- Introduced a toggleable AI reasoning feature in the chat interface, allowing users to enable or disable advanced reasoning during AI chat interactions.
- Added a new reasoning button to the chat input for quick access and control.
- **Enhancements**
- Improved chat configuration options to include reasoning settings, providing more flexibility for AI responses.
- Streamlined network search and image upload interactions for a smoother user experience.
<!-- end of auto-generated comment: release notes by coderabbit.ai -->
Close [AI-40](https://linear.app/affine-design/issue/AI-40).
### What changed?
- Add anthropic provider.
<!-- This is an auto-generated comment: release notes by coderabbit.ai -->
## Summary by CodeRabbit
- **New Features**
- Added support for Anthropic as a new provider option in Copilot, enabling integration with Anthropic's text generation models.
- Users can now configure Anthropic provider settings, including API key management, through the admin interface.
- **Chores**
- Updated dependencies to include the Anthropic SDK for backend operations.
<!-- end of auto-generated comment: release notes by coderabbit.ai -->
Close [BS-3079](https://linear.app/affine-design/issue/BS-3079).
- Separate the create session logic from the `createMessage function`.
- Ensure the session is created before executing any chat or actions.
- Convert the `AIActions` into asynchronous functions.
- Transfer the update prompt name logic to the chat action.
- Introduce a networkSearch field in `AITextActionOptions`.
- Eliminate the redundant `LAST_ROOT_SESSION_ID`.
Close [BS-2758](https://linear.app/affine-design/issue/BS-2758).
Support [BS-2583](https://linear.app/affine-design/issue/BS-2583).
### What changed?
- Extend `ChatPanelInput` and `ChatBlockInput` from the same abstract class `AIChatInput` to reduce duplication of code.
- Unify the context interface of `chat-panel` and `chat-block`.
- Rename `items` field to `messages`.
- Remove duplicated type declare.
Support [BS-2583](https://linear.app/affine-design/issue/BS-2583).
### What changed?
- Move chat panel chip components to `/components` folder.
- Separate `chips` and `embeddingProgress` from ChatContextValue.
### Why make this change?
Seperate front end timeout errors from server side errors.
### What changed?
- Add `RequestTimeoutError` which extends from `BaseAIError`.
- Track as `request timeout` instead of `server error`.
Close [BS-2834](https://linear.app/affine-design/issue/BS-2834).
### What Changed?
- Change `reference_index` from chip order to increasing positive integer.
- Add matched context documents to ai prompt.
Close [BS-2834](https://linear.app/affine-design/issue/BS-2834).
### What Changed?
- Add `addContextCategoryMutation` and `removeContextCategoryMutation` graphql apis.
- Provide tag and collection apis for front-end components.
Close [BS-2850](https://linear.app/affine-design/issue/BS-2850).
### What Changed?
- After flattening the search results, filter the current document.
- Remove duplicate results based on document id.
Close [BS-2791](https://linear.app/affine-design/issue/BS-2791).
### What Changed?
- Add status filed to `CopilotContextDoc` to querying document embedding processing progress.
- Change `ChipState` from `success` to `finished` to better align with backend server status.
- Add `pollContextDocsAndFiles` API for embedding status polling.
### About Polling
- Set the minimum interval to 1 second and the maximum interval to 30 seconds
- Use exponential backoff and increase the interval by 50% after each poll
- Make sure the interval does not exceed the maximum
Close [BS-2751](https://linear.app/affine-design/issue/BS-2751).
### What Changed?
- Do not send AI gql request when chat-panel is not open.
- When the chat-panel width is greater than 540px, adjust the padding to 24px.
- Optimize the display and hide logic of scroll to end button.
Why make this fix?
The network search flag is saved locally. If the user opens multiple clients, the local status and server status may be inconsistent.
Fix issue [AF-2252](https://linear.app/affine-design/issue/AF-2252).
## What Changed?
- Remove useless `generateAnswer` configuration for inline ask ai
- Refactor the common `updateAIPanelConfig` function
- Use empty string instead of meaningless `placeholder` string
- Remove unnecessary high-order function wrappers
Fix issue [BS-2522](https://linear.app/affine-design/issue/BS-2522).
### Why make this change?
If the user data contains illegal content, carrying the user data in the system prompt will run the risk of having the account banned.
### What Changed?
- Move the `Context Documents` to the user prompt.
- Add `withPrompt` in `QueryChatHistoriesInput` options.
- Get chat histories without prompt messages.
- Omit document context when saving messages to the `aiSessionMessage` db.
Support issue [BS-2353](https://linear.app/affine-design/issue/BS-2353).
### What changed?
- Remove `Open AI Chat` menu item.
- Remove `chat-cards` web component.
- Add `extractAll` function for page and edgeless doc full content extract.
For an existing AI chat session, the prompt will not be updated when the front-end calls the `createChatSession` gql interface. Thus update it manually here.
In the future, the architecture needs to be adjusted to support multiple LLM model calls in one chat session.
Fix issue [AF-1632](https://linear.app/affine-design/issue/AF-1632).
### What Changed?
- Keep the note selection and do not set the widget tool to copilot if asking AI from selection in a note in edgeless mode.
- Remove `content` from `AIImageActionOptions`, using `input` as AI context like other actions.
Fix issue [BS-1779](https://linear.app/affine-design/issue/BS-1779).
The attribute `contenteditable` will be `true` if the document is not read-only. In this case, the note preview can be focused on and users can edit the note content.
But we can not simply make the document read-only. If the user changes the editor settings, we also need to update the document model, which requires that the document is not read-only.
Thus following code sets the document editable before the model updates and resets the document to read-only after the model is updated.
akumatus
