What this category of data looks like
Daily documents for this category include keyword lists, negative keyword tables, bid records, and search term reports. Data sources include search ad backends, search term reports, auction ranking data, and landing page data. Data update cycles are typically daily or weekly. Fields include keyword match type, bid amount, search volume, negative keyword match mode, click and conversion data, and others. Units include yuan, times, characters, and others. There is no fixed unified standardized field format, and data must be organized according to the export rules of the corresponding advertising platform.
What constraints do these characteristics impose on the citation source and traceability link
Data sources for this category are scattered, involving multiple platforms such as advertising backends and search term reports. Accurately associate the unique identifiers of keywords across different data sources when performing citation traceability, to avoid confusing content from different delivery cycles. Data updates occur daily or weekly, so ensure that the cited materials match the keywords and negative keyword tables of the current delivery cycle; otherwise, compliance review deviations will occur. Document fields have no unified format, so configure field mapping rules for different export files to ensure precise matching of keywords and corresponding review materials during traceability. At the same time, incorrect traceability can lead to issues such as invalid clicks and brand term hijacking, so strictly bind the corresponding relationship between data sources and keywords.
How to set configurations
| Configuration Item | Recommended Value | Rationale |
|---|---|---|
quoteMaxToken | 800–1200 token | Adapt to the average content length of documents such as search ad keyword lists and negative keyword tables, avoiding truncation of key information required for compliance reviews |
responseData | Return only matched keywords and source fields | Filter irrelevant data to ensure traceability results focus on core compliance materials for advertising delivery |
detail | Enable full source display | Show the export file name of raw data and advertising platform identifier for quick source verification |
| Single search limit | Top 10 entries | Cover the core data scope of daily keyword expansion and negative keyword screening, avoiding excessive redundant information that interferes with reviews |
| Maximum context length | 5000–7000 characters | Adapt to batch review requirements for multiple advertising documents, ensuring all citation sources can be fully associated |
Three common mistakes
- Knowledge base variable reference dropdown is blank, with no variables available for selection. The cause is that global variables bound to advertising data sources have not been predefined in the workflow, or the variables have not been configured as a type that can be referenced by knowledge base search nodes.
- Source data for different keywords cannot be distinguished in logs. The cause is that the
detailparameter has not been configured to enable full source display, or keyword unique identifier fields have not been bound in field mappings. - Large language model returns forged citation IDs. The cause is that the
responseDataparameter has not been enabled to restrict returned fields, causing the model to fail to accurately match identifier information of original data sources, or insufficient context length leading to truncation of key source information.
How to confirm the configuration is complete
- Enter the workflow configuration page of FastGPT v4.17.0, open the variable reference settings for the knowledge base search node, and check if defined advertising data source variables appear in the dropdown list.
- Run a keyword review task, view the traceability information of the returned results, and confirm that the file name or platform identifier of the raw data is included.
- Check the task logs to confirm that each cited keyword is bound to the corresponding data source field, with no null values or confusion.
- After adjusting the
quoteMaxTokenparameter, test keyword lists of different lengths to confirm that no key information is truncated.
The values provided 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.