What the data for this category looks like
Data sources for delivery effect optimization scenarios include advertising platform data, attribution systems, CRM leads, and landing page data. Update frequency is real-time or daily. Regularly produced documents include daily reports, weekly reports, price adjustment records, material test sheets, and optimization retrospective documents. Data structure is categorized by advertising application, time, and material dimensions. Core fields include advertising plan ID, material ID, conversion lead identifier, token consumption, session identifier, and others. Data is only archived by business dimensions, with no fixed unified statistical caliber.
What constraints these characteristics impose on the conversation logs and auditing link
The data in delivery effect optimization scenarios is multi-source heterogeneous, has flexible update frequencies, and is associated with specific business operations. This brings three constraints to the conversation logs and auditing link. First, session logs must be isolated by advertising application dimension to avoid confusion of operation records from different delivery projects. Second, log collection must match the real-time or daily update rhythm of the business, to ensure that when an abnormality occurs, the operations of the corresponding time period can be quickly traced back. Third, core business fields such as advertising plan ID and token consumption must be collected, so that subsequent resource consumption statistics and abnormal indicator positioning can be performed by application dimension.
How to set the configurations
| Configuration Item | Recommended Approach | Rationale |
|---|---|---|
chatId isolation rule | Divide by advertising application ID | Match the application dimension statistics and permission isolation requirements of delivery effect optimization |
| Log retention period | 30–90 days | Cover the cycle of regular delivery retrospectives, while controlling storage costs |
responseChatItemId collection granularity | Return full conversation content | Fully record the complete conversation link of operations such as price adjustments and material tests |
detail field collection scope | Include token consumption and advertising plan ID | Meet the business requirement of counting token consumption by application dimension |
| Abnormal log alert threshold | Trigger when single session token consumption exceeds 5000 tokens | Match the operation volume of regular delivery sessions, filter normal fluctuations |
Three common mistakes
- Symptom: Only some Agents can view conversation records, while other Agents have no permission to access | Cause: The advertising application ID is not bound to
chatId, causing the session isolation configuration to not take effect - Symptom: The log page shows no content, and model call records cannot be viewed | Cause: Core business identifier collection for the
detailfield is not enabled, causing logs to not be associated with valid business information - Symptom: Conversation logs are lost, and abnormal operations cannot be traced back | Cause: A reasonable log retention period is not configured, or the docker configuration file does not correctly mount the log storage path
How to confirm the configuration is correct
- Log in to the FastGPT v4.17.0 management backend, enter the session log page of the target advertising application, and verify that only Agents bound to this application can view the corresponding conversation records
- Perform a conversation operation for price adjustment or material testing, check the
detailfield in the log details, and confirm that it includes business identifiers such as token consumption and advertising plan ID - Enter the token statistics page, filter data by application dimension, and verify that the statistical results match the actual token consumption of the conversation operations
- Check the log storage path in the docker configuration file, confirm that log files are not automatically cleaned, and the retention period meets the configuration requirements
The values provided here 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.