Knowledge Base

Configure Models

To use the knowledge base, you need to configure embedding models (required) and reranking models (optional). For related configuration, please refer to Model Configuration.

After configuration, you can enable the knowledge base function in the record page dialog:

Model availability status will be detected during the enabling process.

Knowledge Base Vector Computation

The knowledge base functionality is implemented based on RAG principles, converting Markdown files into vectors for storage and search. The entire process is completed through embedding models.

When using for the first time, if Markdown files already exist in the writing page, you can first perform full computation (optional):

For subsequent writing, you don't need to perform full computation every time. It will automatically compute during auto-save while writing.

This process will consume embedding model resources. Use judiciously or adopt free models.

Parameter Settings

If you're not familiar with knowledge base-related knowledge, you can ignore this and use default values.

When using the knowledge base, you may not always get complete matches for the information you need. By adjusting parameters, you can more precisely control the retrieval effectiveness of the knowledge base:

• Chunk Size: Maximum number of characters in text chunks. Larger chunks may contain more context but increase vector computation complexity.
• Overlap Size: Number of overlapping characters between text chunks. Larger overlaps can maintain context coherence.
• Retrieval Count: Number of relevant documents returned during retrieval. More documents may provide richer information but may also introduce noise.
• Similarity Threshold: Minimum similarity threshold between documents and queries. Only documents exceeding this threshold will be returned. Value range 0.0-1.0, higher values are more strict.