What this category's data looks like
This page addresses the scenario of annual progress and acceptance of scientific research projects. Material sources include annual progress reports, mid-term inspection materials, acceptance reports, and their attachment receipts. The update rhythm changes on a rolling basis with project milestones and supplementary explanations. Field characteristics focus on project name, undertaking unit, number of indicators, start and end dates, signatures, and receipts. Annual progress reports carry phased results and plans for the next year; mid-term inspection materials carry indicator comparisons and problem explanations; acceptance reports carry final indicators, funding, and conclusions. After materials enter document parsing and chunking, retrievable data rows are formed according to chunkSize and chunkSplitter. trainingType determines the vector or QA processing path. Upon completion of parsing, they are stored in the database and bound to a unique data identifier, providing a traceable data foundation for subsequent version ledgers and change records.
Configuration and implementation checks
This step is executed by parsing and training tasks (server-side tasks), outputting data rows and index statuses (parsed, indexing, indexed). "Version ledger and change records" are then completed by retrieval and subsequent processing.
| Object | Check | Verification action |
|---|---|---|
| Annual progress report | Project name, number of indicators, start and end dates, and signature receipts are consistent between the annual progress report and mid-term inspection materials | Open the data rows of the two materials separately, and verify the name, quantity, date, and signature receipt fields |
| Mid-term inspection materials | The name, quantity, and date of the same record match between the mid-term inspection materials and the acceptance report, and the version number and references are synchronized | Compare the data rows of the two materials, and check whether the version number and references are updated synchronously when one content change is made |
| Acceptance report | The signatures and receipts of the acceptance report are complete, the current version is unique, and old versions are traceable | Check whether the data row status is parsed, indexing, or indexed, and verify the coverage scope of the rebuild task |
| Data row | It is in one of the three states and consistent with the interface display, and the file record and data row identifier are distinguishable | Open the data row to verify the status, verify the file record and data row separately, and confirm the correspondence of chunkSize, chunkSplitter, and trainingType |
Three common errors
- Symptom: The name, quantity, and date of the same record do not match between the annual progress report and mid-term inspection materials. Cause: The two materials are parsed and stored separately, and fields are not aligned by unique data identifier.
- Symptom: After changing one content, the version number and references are not updated synchronously. Cause: The old data row is still in indexed state, and the retrieval layer continues to recall old content.
- Symptom: The acceptance report lacks signatures or receipts, but retrieval can still hit it. Cause: Successful parsing does not mean retrieval will definitely hit, but the index status does not intercept data with missing items.
How to verify the setup
Acceptance of this feature: Upload one material, observe the data row status transitions and the coverage scope of the rebuild task. Statuses are only recognized as parsed, indexing, indexed, and verify chunkSize, chunkSplitter, trainingType, and the data row identifier.
Business test: Change one content, and see whether the version number and references are updated synchronously.
- Open the data row to verify the status, verify the file record and data row separately, and confirm that the current version is unique and old versions are traceable.
- Check the data rows covered by the rebuild task, and confirm that only data with completed indexing is covered, and data pending indexing does not participate in the rebuild.