What data for this category looks like
Data sources for this category include raw datasets, ETL run logs, output from annotation platforms, and sampling quality inspection results. Data updates occur daily or after each batch task completes. Daily generated documents include cleaning rule configuration files, quality inspection sampling statistics tables, abnormal data lists, and final delivered quality reports. Data fields include data identifiers, anomaly types, processing status, sampling ratios, quality inspection conclusions, and more. There are no uniform fixed units; values adjust based on specific data types.
What constraints do these characteristics impose on the source citation and traceability link
The batch update nature of raw datasets and ETL logs requires traceability to support associating multi-source data by batch, to avoid fragmented traceability information. The versioned attribute of cleaning rule documents requires traceability to bind citation identifiers corresponding to the rule version, ensuring traceability to specific cleaning configurations. The structured output requirement for abnormal data lists and delivery reports requires traceability-returned sources to match field hierarchies, to facilitate subsequent quality inspection review. The daily or per-batch update cadence requires traceability timestamps to align with data update cycles, to avoid referencing expired cleaning or quality inspection results. Multi-source annotation results from annotation platforms require traceability to distinguish source identifiers from different annotation links, to ensure the accuracy of quality inspection traceability.
How to configure settings
This configuration guide applies to FastGPT v4.17.0.
| Configuration Item | Recommended Value Range | Rationale |
|---|---|---|
quoteMaxToken | 800–1200 token | Source documents for data cleaning and quality inspection are mostly structured fields. This range covers complete cited content without exceeding context limits |
| Source citation limit | Top 10 entries | Data sources for single-batch data in cleaning and quality inspection are mostly limited batches. This limit covers sources from major processing links |
detail | Enable full traceability | Abnormal data lists and delivery reports need to display specific processing nodes. Full traceability returns logs and results from corresponding links |
responseData | Include source ID, update time, processing link | This category needs to associate data sources, quality inspection links and other information, to ensure traceability covers the complete processing chain |
| Timestamp alignment rule | Match daily or per-batch update cycles | Align with actual data update rhythm, to avoid referencing expired cleaning or quality inspection results |
Three common configuration errors
- Phenomenon: The number of cited sources returned by interface calls is less than the set limit. Cause: The source citation limit is not configured correctly, or the value of the
quoteMaxTokenparameter does not match the content length of a single source. - Phenomenon: The traceability information returned by the interface has no download link or the download fails. Cause: The
detailconfiguration item is not enabled, or the storage path of the source data is not correctly associated with FastGPT’s knowledge base index. - Phenomenon: The granularity of exported traceability data cannot match a single cleaning rule or sampling batch. Cause: The knowledge base dimension is mistakenly used as the only export granularity, and the traceability association dimension is not configured according to the data update cycle.
How to confirm configuration is complete
- Call the FastGPT question answering interface, check whether traceability-related field information is included in the returned results.
- Adjust the configuration value of the source citation limit, verify whether the number of cited entries returned by the interface matches the set value.
- Export traceability data, check whether the export granularity can be split by dimensions such as cleaning rules and sampling batches.
- Test reference to expired data, confirm that the returned traceability information includes correct update timestamps.
The values given are common starting points and should be measured against the reader's own samples.
Question material comes from public community discussions. Configuration values are common starting points and should be measured against your own samples. Verified on 2026-10-19.