Embed and Add to Knowledge Base¶
Generate document embeddings and write them to a knowledge base index for retrieval.
This node saves vectors to the knowledge base. Use Text Embedding if you only need vectors for further processing.
Configure the knowledge base¶
Bind documents from parsing or chunking and select the target knowledge base.
Setting |
Description |
|---|---|
Knowledge base |
Required; select or create a target that provides the index table and embedding model settings |
Multi-level retrieval |
Enabled by default; builds document, section, and chunk indexes |
Chunks per section |
Default: 3; groups consecutive chunks into sections when multi-level retrieval is enabled |
The selected knowledge base supplies table_name and embedding_model. Retain document source metadata for index management and tracing.
Use the output¶
Output |
Purpose |
|---|---|
|
Number of document or index rows actually written |
|
Table receiving the index |
|
Model used to build the index |
With multi-level retrieval, written includes rows at document, section, and chunk levels.
Example: index manuals¶
Flow: Read MOI Volume → General Document Parsing → Segment → Embed and Add to Knowledge Base.
Select the manual knowledge base, keep multi-level retrieval enabled, and set chunks per section to 3. After running, check the target table and row count, then verify the content through knowledge base retrieval.