What this category's data looks like
In the social evaluation scenario of vocational skill level certification, the materials that enter the indexing and vector retrieval stage usually include the evaluation implementation plan, candidate registration forms, and assessor dispatch forms. The evaluation implementation plan gives the certification occupation, level, assessment method, and time arrangement, and is the backbone of the basis query; the candidate registration form carries the candidate's identity, application conditions, and material list; the assessor dispatch form records the assessors, dates, and signatures. The three types of materials are updated by batch or certification cycle, and the fields include name, quantity, date, signature, and receipt. After entering this function, the implementation plan is split into the index as the main basis text, and the registration form and dispatch form are used as supporting materials to supplement recall, for subsequent nodes to trim and cite.
Configuration and implementation checks
This step is executed by the indexing service and the vector database, and outputs vectors and index status (whether recall is possible, whether data in the same group is merged). "Basis entries that can be traced back to the original text" are then completed by subsequent AI nodes, code nodes, or manual work (retrieval and citation trimming only take supporting materials). The index model must be consistent with the vector model actually used by the knowledge base, and the index field of a data row must retain multiple vector identifiers for one piece of data.
| Object | Check | Verification action |
|---|---|---|
| Evaluation implementation plan | The name, quantity, date, and signature must match the registration form and dispatch form item by item, and it must be the current version | Check the index model in the knowledge base settings; open a data row and check defaultToken and maxToken to confirm the index field is complete |
| Candidate registration form | The name, quantity, and date of the same candidate in the two materials are consistent; missing signatures or receipts must be marked | Retrieve the same candidate record and see whether the same group of original data is recalled and merged, and whether the status is indexed |
| Assessor dispatch form | The assessor name and dispatch date are consistent with the implementation plan batch; if the version is not synchronized, update it first | Check the data range covered by the rebuild task record, and confirm that the data that has completed indexing after the model change was triggered |
| Data row index field | Multiple vector identifiers for one piece of data are all present and correspond to the vector identifiers in the vector database | Open the data row and check the index field, then retrieve again to see whether data in the same group is merged and whether the highest score is taken |
Three common errors
- Symptom: duplicate entries appear in retrieval results. Inspect the following: the same group of original data corresponds to multiple vector identifiers, and after recall they are not merged back to the original data by identifier.
- Symptom: recall quality drops after changing the model. Inspect the following: the index model is inconsistent with the vector model actually used by the knowledge base, and the rebuild task record does not cover all data.
- Symptom: the name, quantity, and date of the same record in the two materials do not match. Inspect the following: the version is not synchronized, signatures or receipts are missing.
How to verify the setup
This function's acceptance: take one data row and look at the index field to confirm that multiple vector identifiers are all present; then retrieve once more to see whether the same group is merged, whether the status is indexed, and check the rebuild task record.
Business test: ask a basis question that spans materials, see whether the answer gives a traceable source, and check whether the number of citations and the tokens of the cited text are recorded separately.
- Check in the knowledge base settings whether the index model is consistent with the vector model actually used by the knowledge base, and check the data range covered by the rebuild task record.
- Open a data row and check the index field to confirm that
defaultToken,maxToken, and the vector identifiers are complete, then retrieve to verify same-group data merging and the highest score.