This document applies to FastGPT v4.17.0.
What data for this category looks like
Data for this category primarily comes from advertising placement platforms, creative material libraries, user interaction data, and conversion data. Update frequency is mostly daily or weekly. Regular associated documents include creative scripts, material test sheets, placement data reports, and iteration checklists. Fields typically include creative material ID, placement time period, interaction metrics, conversion data, and more. Common units are material count, impression count, click count, and conversion cost. This data centers on creative combination testing for information feed ads, with core records of placement performance for different image, video, and copy combinations. It is used to evaluate creative effectiveness and complete iterative optimization.
What constraints these characteristics impose on the conversation logs and auditing workflow
Decentralized data sources require logs to integrate placement, material, and interaction data from multiple platforms. Cross-source data pull rules must be configured to ensure data completeness. Differences in update frequency require logs to be archived in daily or weekly batches, to avoid performance pressure from real-time storage. Daily documents are tied to test data, so logs must be linked to unique identifiers from creative scripts and test sheets to enable traceability by test batch. The large number of creative variables and complex test combinations require logs to record variable parameters and corresponding replies for each round of conversation, to facilitate summarizing test conclusions. Errors directly impact placement effectiveness, so the auditing workflow must retain complete conversation links to quickly locate material violations or parameter errors.
How to set configurations
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
chatId | Bind to creative test batch ID | Ensure conversation logs for each creative test are linked to their corresponding test batch, enabling archiving and retrieval by test dimension |
responseChatItemId | Generate unique identifiers per material version | Distinguish conversation replies for different material versions within the same test batch, avoiding log confusion |
detail | Enable full field return | Cover all information including creative variables, placement data, and user interactions to meet retrospective and auditing requirements |
| Log storage period | 30–90 days | Match the regular weekly or monthly retrospective cycle for this category, balancing storage costs and audit traceability needs |
| Context recall limit | Top 10 conversation records | Adapt to the multi-round variable adjustment conversation links in creative testing, avoiding interference from redundant data during troubleshooting |
| Conversation log export format | CSV format | Compatible with conventional editing and analysis tools for placement data reports and iteration checklists |
Three common misconfigurations
- When calling the conversation API to view log details, the returned content for multiple creative tests remains consistent and does not match the actual generated replies. Cause:
chatIdis not bound to the unique identifier of the creative test batch, leading to incorrect association or overwriting of logs from different batches. - When configuring multiple knowledge base categories, conversation auditing only reads the current question and does not include historical context records, causing classification results to lean toward fallback categories. Cause: Context recall configuration is not enabled, only single-round conversation data is captured, and the multi-round variable adjustment links of creative testing are not supported.
- When attempting to pull all conversation records or export logs, the number of returned results is incomplete and cannot cover the full test cycle data. Cause: No cycle threshold for full log storage is configured, or the
detailfull field return switch is not enabled, leading to truncation of partial data.
How to confirm proper configuration
- Generate one round of creative test conversation, retrieve logs using
chatId, and confirm that the linked test batch identifier is correct. - View the
detailfield of a single conversation entry, and confirm that it includes all associated information such as creative variables and placement data. - Export conversation logs for a specified period, and verify that the number of exported entries matches the actual number of test records.
- Trigger a multi-round conversation, retrieve context records, and confirm that historical conversation content is fully recalled.
The values provided here are common starting points and should be evaluated against local deployment 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.