Manage files and structured tables¶
Open the Knowledge tab in knowledge-base details to inspect imported files and tables before using them through an agent.
Check file-processing status¶
Status |
Meaning |
Next action |
|---|---|---|
Processing |
Parsing, chunking, or indexing is incomplete |
Wait before validating answers |
Completed |
The processing workflow finished |
Inspect originals and chunks, then test retrieval |
Failed or error |
A processing step did not succeed |
Read the error and check format, legibility, and content |
Library-level readiness does not replace individual file status. Completed processing also does not guarantee every answer: relevance and configuration still matter.
Read and maintain files¶
Select a file to open its preview and confirm the source. In file information, review tags and validity dates. In the chunks area, inspect the text supplied to retrieval. Use the add-material entry to import additional long-term information and check processing again.
Include the business object and date in file names. When material changes, use the available update or import action and verify that the processed text contains the new content rather than relying on the filename.
Structured tables¶
After import, inspect table names, fields, sample values, date ranges, and row-and-column meanings. Correct table counts do not establish correct business definitions: currency units, refunds, and test-order exclusions need explicit interpretation.
Add these through NL2SQL semantics. Validate a small statistical question against manually calculated source data, not just a successful query status.
Delete or clean up material¶
Use the actual file or table’s additional-actions menu and verify the target name before submitting. Removing material affects agents that retrieve it. Check continuing business dependencies first.