What the data for this category looks like
Jewelry yield rate related data comes primarily from internal brand inventory and sales systems, offline retail POS ledgers, and third-party jewelry market trend aggregation APIs. Full data synchronization runs daily at midnight. The maximum delay for updated data on individual jewelry items is no more than 2 hours. Data uses structured JSON format. Each record includes a unique SKU identifier, jewelry name, material type, daily supply unit price, daily terminal retail price, and daily second-hand circulation reference price. The unit of price fields varies by material: precious metals use yuan/gram, non-precious metals use yuan/item. The update timestamp is an ISO 8601 formatted string.
What Constraints These Characteristics Impose on Conversation Logging and Auditing
Multiple data sources require audit logs to fully record the call chain for each data entry, eliminating data traceability blind spots. Fixed daily update schedules require logs to archive incremental data daily, supporting monthly review and compliance audits. Differentiated field units require logs to track unit information for each numeric field, preventing ambiguity in valuation methods during audits. Fine-grained SKU-level data requires logs to link to specific SKU identifiers, enabling quick location of abnormal records by individual item and avoiding issues with unfilterable massive logs. Additionally, conversation logs must fully retain broadcast request parameters and returned results to ensure yield rate broadcast content can be traced and verified.
Configuration Settings
| Configuration Item | Recommended Value | Rationale |
|---|---|---|
WORKFLOW_LOG_ENABLE | true | Enable full logging for workflow nodes, covering every execution step of the jewelry yield rate broadcast |
RECORD_DATA_SOURCE_DETAIL | all | Fully record request parameters and returned fields for third-party APIs and internal systems used to call jewelry data, adapting to audit requirements for multiple data sources |
LOG_FIELD_UNIT_TRACK | enabled | Track unit information for each numeric field, resolving valuation unit ambiguity between precious and non-precious metal jewelry |
LOG_RETENTION_DAYS | 90 days | Meet compliance retention requirements for financial audits, adapting to monthly review needs for jewelry yield rate daily reports |
MAX_LOG_ENTRY_PER_WORKFLOW | 500 entries | Balance log storage and query efficiency, adapting to scenarios where a single workflow processes data for multiple SKU jewelry items |
API_HISTORY_LOG_SOURCE_ENABLE | true | Record all conversation context passed through API calls, meeting traceability requirements for cross-system audits |
The parameter values provided on this page are common recommended starting points for configuration. Actual values are affected by material form, data volume, and business rules. Specific issues require individual analysis, and it is recommended to test on your own samples before finalizing settings.
Three Common Mistakes
- Phenomenon: No interface return data is recorded in workflow logs after calling the third-party jewelry market trend API. Cause: The
RECORD_DATA_SOURCE_DETAILconfiguration item is not enabled, resulting in failure to write data source call details to logs. - Phenomenon: When querying historical conversations via the API, the
sourcefield only displays a basic call identifier, and cannot distinguish the specific call source between internal inventory and sales systems and third-party market trend APIs. Cause: TheLOG_SOURCE_DETAILparameter is not configured, only general call source types are recorded, and data source identifiers are not detailed. - Phenomenon: After the MCP service fails to execute in a workflow, corresponding error details cannot be found in system logs. Cause: The
MCP_SERVICE_LOG_ENABLEconfiguration item is not enabled, resulting in failure to synchronize MCP service execution logs to the global audit log database.
How to Confirm Configuration is Correct
- Navigate to the FastGPT workflow log management page, select the jewelry yield rate broadcast workflow, and check whether execution time, input parameters, and returned results for each node are included.
- Call the API to trigger a jewelry yield rate broadcast, query the global audit log, and confirm that the log includes data source call identifiers and field unit information.
- Navigate to the system log management page, filter logs related to the MCP service, and confirm that execution records for the MCP component of the corresponding workflow can be found.
- Check the log retention days configuration item, and confirm that the current retention duration meets internal audit requirements.
Question material comes from public community discussions. Configuration values are common starting points and should be measured against your own samples. Verified on 2026-09-14.