What this category's data looks like
This category's data revolves around three types of materials: training plans, thesis proposals, and mid-term assessment forms. Materials enter the knowledge base in batches according to semester milestones. The training plan carries course and credit arrangements; the thesis proposal records the topic selection and supervisor comments; the mid-term assessment form summarizes stage progress and review conclusions. The fields of the three overlap: student name, student ID, course name, credit count, proposal and assessment dates, supervisor signature, review panel signature, and college receipt. The update cadence is typically: the training plan is finalized after enrollment, the thesis proposal is submitted in the second semester, and the mid-term assessment is completed at the end of the third semester. When materials enter this feature, each participates in retrieval as an independent source; the citation output is only responsible for pointing back to the original text fragments, while field comparison and ledger generation are left to subsequent steps.
Configuration and implementation checks
This step is executed by the citation output of the knowledge base retrieval node. It outputs an array of citation content, each entry containing an original text fragment and a source identifier, and the length may be 0. The exception list and review conclusions are completed subsequently by AI nodes, code nodes, or manual work.
| Object | Check | Verification action |
|---|---|---|
| Training plan | Whether the course name, credit count, study semester, and signature fields correspond item by item with the thesis proposal and mid-term assessment form | Open the citation superscript to point back to the original paragraph of the training plan, verify the fields and dates, and record the quoteMaxToken budget and the number of array entries |
| Thesis proposal | Whether the topic title, supervisor signature, and proposal date are consistent with those recorded in the mid-term assessment form, and whether the receipt is complete | View the citation array in responseData (non-streaming module results returned with detailed output enabled) within the conversation, and compare each original text fragment with its source identifier one by one |
| Mid-term assessment form | Whether the assessment conclusion, review panel signature, and assessment date are synchronized with the versions of the training plan and thesis proposal | Run a test with a question that has no matching materials, confirm that the array is empty while the pipeline still runs through, then check quoteList to locate the matched materials |
Three common errors
- Symptom: The course name and credit count for the same student do not match between the training plan and the mid-term assessment form. Cause: The two materials were entered in different semesters, and no field-level comparison or version marking was performed.
- Symptom: The thesis proposal has been updated, but the mid-term assessment form still cites the old topic title and date. Cause: The versions were not synchronized, retrieval matched the old fragment, and the citation did not mark the version difference.
- Symptom: The mid-term assessment form lacks the review panel signature or college receipt, and the original record cannot be found during tracing. Cause: The scanned materials are incomplete, and the cited fragments cover only the body text but not the signature page.
How to verify the setup
Feature acceptance: ask a question that has original-text support, and check whether the citation array and superscripts can point back to the original text; then ask a question with no matching materials, confirm that the array is empty while the pipeline still runs through, and record the number of array entries and the quoteMaxToken budget.
Business testing: create an exception record and see whether the source material can be traced, confirming that the exception is supported by an original record and, after the reviewer signs, is handed to the responsible person for handling.
- In the node, verify whether the selected knowledge base collections and their vector models are consistent, and open the search parameter panel to check the values item by item.
- In the conversation, open the citation superscripts, compare them against the source documents, confirm that each citation can point back to the original paragraph, and separately record the number of array entries.