For the complete documentation index, see llms.txt. This page is also available as Markdown.

Knowledge and Content Workflow

Knowledge Base, Notes, Drawing, and Translation

Knowledge bases, notes, drawing, and translation are not isolated tools from one another. They are responsible for organizing materials, accumulating content, creating visual assets, and handling multilingual files respectively, and all can continue to be passed to the Agent for use.

资料进入知识库、通过召回测试、交给 Agent 并形成文字图片或译文的流程图
First use real questions for retrieval testing, then let the Agent create reports, translations, or illustrations based on the materials.

How the four entry points are divided

Entry
What it is suitable for
Common next step

[Knowledge Base]

Make materials searchable by question

Bind to a chat assistant or Agent

[Notes]

Write, organize, and maintain Markdown content long-term

Export documents or add to the knowledge base

[Drawing]

Generate, edit, merge, and enhance images

Download, or hand off for content tasks

[Translation]

Handle text, images, and documents

Save results, reuse history, or continue editing

新建知识库时的名称、嵌入模型和保存入口
When creating a new knowledge base, first name it, then choose the embedding model used for retrieval.

From materials to deliverables

1

1. Organize the sources first

Add long-term reference materials to the knowledge base, and keep drafts currently being revised in Notes or the Work directory. Do not import all unconfirmed temporary chats into the knowledge base.

2

2. Verify whether key information can be retrieved

Use retrieval testing in the knowledge base to confirm that real questions can find the correct snippets. If retrieval fails, fix the materials and settings first, and do not let the Agent guess.

3

3. Then create text, translations, and images

Have the Agent draft content based on the materials; polish images and translations separately in [Drawing] and [Translation]. Finally, have a human check facts and wording.

Last updated

Was this helpful?