What the Data for This Category Looks Like
Hotel and catering investment research data mainly comes from store POS systems, supply chain purchase ledgers, OTA user reviews, and regional competitor dynamic documents. Update cycles vary significantly: POS revenue data is updated daily, ingredient purchase ledgers are updated weekly, and industry competitor dynamics are released irregularly.
Single-document structures include fields such as store ID, customer unit price (unit: yuan), passenger flow (unit: people), ingredient SKU, purchase price (unit: yuan/kg), user rating (1-5 points). Some documents include detailed lists of multiple SKUs.
What Constraints These Characteristics Impose on the "Conversation Logs and Auditing" Link
Hotel and catering investment research data has scattered sources and inconsistent update cycles. Audit logs must mark the source and update time of each referenced data to avoid using expired revenue or purchase data.
The multi-field document structure requires logs to record specific fields referenced in conversations, to facilitate tracing data deviations. Some conversations involve real-time operational adjustments, so logs must fully record trigger times, operating subjects, and call links to meet compliance audit requirements.
How to Configure the Settings
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
AUDIT_LOG_ENABLE | Enabled | Hotel and catering investment research involves operational data. Audit logs must fully record the entire conversation chain, and enabling this generates traceable operation records |
LOG_LEVEL | info | Key information such as conversation triggers, data source associations, and result returns must be covered. The info level balances log volume and audit needs |
LOG_RETENTION_DAYS | 90 days | Audits of catering industry operational data typically require at least quarterly records, and 90 days meets conventional compliance requirements |
LOG_INCLUDE_SOURCES | Enabled | Hotel and catering knowledge bases reference multi-source store, supply chain, and review data. Enabling this allows marking the data source ID and update time of each reference in logs |
LOG_MAX_SIZE_PER_DAY | 500 MB | Maximum daily log file size, preventing excessive disk usage from impacting store system operations |
PARSE_AUDIT_LOG_TIMEOUT | 300 seconds | Catering supply chain documents may include multiple SKU details. Parsing audit logs requires sufficient time to complete field extraction and association |
The parameter values provided on this page are common recommendations used as a starting point for configuration. Actual values are affected by material form, data volume, and business rules. Specific issues require case-by-case analysis, and it is recommended to test against your own samples before finalizing settings.
Three Common Configuration Mistakes
- Symptom: After deployment, the conversation log page fails to load after a long time, with a 502 status code. Cause:
LOG_RETENTION_DAYSis not configured to a reasonable range, resulting in excessively large log index files that time out during loading. - Symptom: Conversations from the Feishu channel have no record in FastGPT logs, but Feishu backend shows the message was sent successfully. Cause: The
LOG_INCLUDE_CHANNELconfiguration is not enabled, or the channel callback address is not correctly bound to the log reporting node. - Symptom: After submitting a question, the system logs do not show the actual called model version, which does not match the configured model A and the model B configured for the knowledge base. Cause: The
LOG_INCLUDE_MODEL_INFOconfiguration is not enabled, resulting in logs not recording the actual called model parameters, making it impossible to troubleshoot model configuration deviation issues.
How to Confirm the Configuration Is Correct
- Initiate a test conversation referencing store revenue or supply chain data, and check if the log includes the conversation trigger time, question content, and referenced data source identifiers.
- View the system log panel, confirm that the log level is
info, and the daily log file size does not exceed the preset limit. - Trigger a channel callback, and confirm that the log records the channel source and message content.
- Modify the log retention days configuration, restart the service, and check if archived files for the corresponding number of days are generated in the log directory.
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.