What this category's data looks like
Annual progress and acceptance of research projects involve three types of materials: annual progress reports, mid-term review materials, and acceptance reports. The sources are paper or electronic texts submitted periodically by the project team, and the update cadence advances with the review milestones. Annual progress reports record stage tasks, indicators, and budget execution; mid-term review materials supplement stage results and rectification notes; acceptance reports summarize final indicators, dates, and signed receipts. When these three types of materials enter the parsing and chunking stage, each is stored as an independent file record and bound to a unique data identifier. The fields include name, quantity, date, signature, and version information, for use in subsequent retrieval and comparison.
Configuration and implementation checks
This stage is executed by the parsing and training task (server-side task), which outputs data rows and index status (parsed, indexing, indexed). "Basis entries that can be traced back to the original text" are completed next by retrieval and subsequent processing. After parsing is completed, the parsed chunks are immediately stored as visible data rows and bound to a unique data identifier. The index transitions through three states, and data pending indexing does not participate in recall.
| Object | Check | Verification action |
|---|---|---|
| Annual progress report | Fields include stage tasks, indicator quantities, budget dates, and signatures, and must be aligned item by item with the mid-term review materials | Check whether the names, quantities, and dates in the report are consistent with the mid-term review materials; record separately any entries missing signatures or receipts |
| Mid-term review materials | The version must be current, and the rectification notes and corresponding items in the progress report must not show deviations in name or quantity | Check the file records and data rows separately, and confirm that entries whose versions have not been synchronized are marked |
| Acceptance report | Summarizes indicators, dates, and signed receipts, and must form a closed loop with the aforementioned materials | Look at the data rows covered by the rebuild task, trigger rebuilds only for data whose indexing is completed, and trainingType distinguishes processing modes |
| Data row | Is in one of the three states and is consistent with the interface display; data pending indexing is not recalled | Open the data row to check its status, and confirm that the chunking results corresponding to chunkSize and chunkSplitter can be traced back to the original text |
Three common errors
- Symptom: The name and quantity of the same record do not match between the annual progress report and the mid-term review materials. Cause: The two materials are submitted separately, and the fields are not compared in a unified way.
- Symptom: The version cited in the acceptance report is inconsistent with the current mid-term review materials. Cause: The version is not synchronized, and the old draft remains in the knowledge base and participates in retrieval.
- Symptom: The materials lack signatures or receipts, and the basis entries cannot be traced back to the original text. Cause: After parsing and storage, the signature fields are not checked, so retrieval hits have no source to verify.
How to verify the setup
Acceptance of this feature: upload one material, observe the status transitions of the data rows and the coverage of the rebuild task, and confirm that the statuses are parsed, indexing, and indexed in sequence, that rebuilds cover only data whose indexing is completed, and that data pending indexing does not participate in rebuilds.
Business test: ask a basis question that spans materials, see whether the answer provides a traceable source, and check and record separately the tokens of the cited text and the number of array entries.
- Check the file records and data rows separately, and confirm that the identities of the file records and data rows are distinguishable from each other
- Open the data row to check its status, confirm it is consistent with the interface display, and note that data rows whose training failed can be directly edited or deleted