AI 对话框
Get AI writing and research help from the in-editor assistant panel.
Last updated: February 14, 2026
What AI 对话框 Is
AI 对话框 is embedded in the editor and connected to your current article context.
You can use it to:
- Brainstorm ideas, outlines, hooks, and topic angles
- Draft, revise, and improve sentences or paragraphs
- Search for references and material, then cite them in follow-up tasks
- Write generated content back into the editor or save it to inspiration
Where to Find It
- Open any article in the editor.
- Locate the AI 对话框 panel on the right side (desktop) and start typing.
- On mobile, swipe to the third page of the editor workspace (AI chat view).
- The panel is resizable on desktop, so you can allocate more/less width as needed.
Quick Start
- Choose an AI model from the top-left model selector.
- Type your request in the input area.
- Press
Enterto send. - When a response appears, review it and apply what you need.
- Use New Conversation if you want a clean context for a fresh question set.
Editor Selection as Context
When you select text in the editor, that selection is automatically added to AI request context.
Typical flow:
- Select one paragraph or multiple blocks in the editor.
- Switch to the AI panel and type your request.
- The selected text appears as a small preview at the top of the input area.
- Send the message.
The AI will receive both your question and the selected text automatically (no need to copy-paste manually).
If you do not want this selection included, click the clear button on the preview.
A message can be sent even if the input is empty, as long as a valid editor selection is present.
Supported Actions
Core Text Assistance
- Request drafting support directly in Chinese/English.
- Ask for rewriting of selected text for tone/style changes.
- Ask for structure changes (e.g., “make this section more concise”).
References and Research
- Use
@mention to reference project articles/materials in your prompt. - Upload images (max 5 files, each up to 5MB) when using a model that supports vision.
- Upload files (
pdf,docx,txt,md) for file-based context.
Editor & Knowledge Flow
- Ask the assistant to generate content and write it to the editor.
- Keep generated text in memory for one prompt by context.
- Save useful outputs directly to the inspiration library.
How Outputs Are Applied
- Some responses offer Apply actions (for example, “Apply to Editor”, “Apply Diff”, “Copy All”).
- Diff mode can be used when tool-based updates are returned.
- If an output looks off, do not apply it directly; regenerate or refine with follow-up prompts.
Common Issues
The answer is not as expected
- Reduce ambiguity in your prompt.
- Add constraints (tone, audience, length, format, and language).
- Ask for one task only per message if the answer becomes noisy.
Tool features are missing
- Some models may not support all tool calls.
- Image input is only effective with vision-enabled models.
@references and file search depend on project context availability.
Generation is interrupted
- Use the Stop Generating control to cancel a stalled stream.
- Retry after checking network status and upload completion state.
Chat history seems unavailable
- Use the history panel in the chatbot header to open, browse, or delete previous sessions.
- If you started a new tab/session, create a new conversation as needed.