What this category's data looks like
In the enterprise independent skill talent evaluation scenario, the category materials that enter the business basis query function include three types: the enterprise evaluation filing form, the evaluation specification, and the candidate roster. The enterprise evaluation filing form carries the evaluation entity, filing date, and signature/seal, and is used to confirm that the evaluation activity has a filing basis; the evaluation specification carries the occupation direction, level, assessment method, and scoring rules, and is the core source for determining whether the basis holds; the candidate roster carries candidate names, identity information, declared level, and batch, and is used to check the personnel scope and record consistency. The three types of materials are updated by batch or year, and when versions change, the old and new versions may be retained at the same time. At the field level, attention is paid to whether names, quantities, dates, and signatures are complete; the same record may have inconsistent names, quantities, or dates across different materials, or versions may not be synchronized, with missing signatures and receipts. After the materials enter this function, they are first split and vectorized and written into the vector database; the original data and index fields remain in the business database, and during retrieval they are recalled by vector and point back to the original text, supporting replies to parents or being written into proposals.
Configuration and implementation checks
This step is executed by the index service and the vector database, outputting vectors and index status (whether recall is possible, whether data in the same group is merged). The basis entries that can point 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).
| Object | Check | Verification action |
|---|---|---|
| Enterprise evaluation filing form | Whether the filing entity, filing date, and signature are complete, and whether they can support the establishment of the evaluation activity | Check the filing form fields and signature/seal, confirm the version is current, and check whether signatures or receipts are missing |
| Evaluation specification | Whether the occupation direction, level, and assessment rules are consistent with the filing form, and whether the versions are synchronized | Compare against the filing form and roster, check whether names, quantities, and dates match, and confirm the specification is the current version |
| Candidate roster | Whether candidate names, declared level, and batch can correspond to the filing form and evaluation specification | Compare the roster with the filing form and specification item by item for names and quantities, and check whether signatures and receipts are complete |
| Data row index fields | Whether multiple vector identifiers of one data row are all present, and whether retrieval can recall and merge them | Open a data row to check the index fields, sample a data row to look at the index fields, then retrieve once to see whether the same group is merged; for status values, check indexed and the rebuild task records |
Three common errors
- Symptom: the name, quantity, or date of the same record does not match between the filing form and the roster. Inspect the following: the two materials were entered at different times, and field-level cross-checking was not performed, resulting in contradictions when the basis entries point back to the original text.
- Symptom: the evaluation specification has been updated, but retrieval still recalls the old version content. Inspect the following: the versions are not synchronized; after changing models, rebuild is triggered only for data whose indexing has already been completed, and the old version data is still in the vector database and can be recalled.
- Symptom: the filing form or roster lacks signatures or receipts, but the basis entries are still cited. Inspect the following: retrieval and citation trimming only take supporting materials and do not block materials with missing items.
How to verify the setup
Acceptance for this function: sample a data row to look at the index fields and confirm that multiple vector identifiers are all present; retrieve once more to see whether data in the same group is merged and whether the highest score is taken; for status values, only check indexed and the rebuild task records.
Business test: ask a basis question that spans materials, see whether the answer gives a citable source, and check that the source comes from the institution’s materials, that the version is current, and that the entries match the original text.
- In the knowledge base settings, check whether the index model is consistent with the vector model actually used by the knowledge base, and view the data scope covered by the rebuild task.
- Open a data row to check the index fields and confirm that multiple vector identifiers are all present; during retrieval, check whether data in the same group is merged, record the number of cited entries and the cited text tokens separately, and refer to
limitfor the cited content budget.