What the data for this category looks like
The data for this category primarily comes from advertising platforms, internal material libraries, user interaction data, and conversion data. Daily produced documents include creative scripts, material test sheets, delivery data reports, and iteration lists. Update frequencies are mostly daily or weekly. Data fields focus on creative material types, delivery time periods, conversion links, and test variable combinations. Units include impressions, clicks, cost thresholds, and more. The data is generated across the full workflow of information flow advertising creative combination testing, serving creative optimization and delivery effect review.
What constraints these characteristics impose on the knowledge base retrieval and recall link
Data for this category is scattered across multiple sources and has inconsistent update rhythms. This requires the knowledge base retrieval and recall link to support multi-source data docking and incremental updates. Diverse document types require adapted content parsing for different structures, ensuring that content in different formats such as creative scripts and test sheets can be effectively recalled. Data fields focus on creative combinations and conversion effects, requiring retrieval to accurately match variable combinations and delivery indicators, avoiding irrelevant results that interfere with operations and work order processing. Frequent creative testing and review requirements also require the recall link to have low latency characteristics, supporting real-time creative fatigue judgment and test conclusion summarization.
How to configure settings
All following configuration items are based on the official support list for FastGPT v4.17.0.
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
similarity | 0.75–0.85 | Matches the precise retrieval needs of information flow creative testing, filtering low-relevance historical test data |
limit | Top 8 entries | Covers the variable combination scenarios of a single creative test, avoiding excessive redundant results occupying context space |
searchMode | hybrid | Combines keyword and semantic retrieval, adapting to recall of multiple document types such as creative scripts and delivery data reports |
usingReRank | Enabled | Performs secondary sorting on recall results, correcting semantic matching deviations and improving the recall priority of high-conversion creative materials |
datasetSearchUsingExtensionQuery | Enabled | Automatically expands query terms for work orders or test questions, covering synonyms of creative variables and industry terminology |
collectionIds | Bound by business group | Only retrieves knowledge base collections related to creative testing, excluding irrelevant internal administrative documents and narrowing the recall scope |
Three common configuration errors
- An
invalid configuration parameter name "hnsw.iterative_scan"error occurs. This is caused by using non-standard configuration parameters outside the official support list, and not using retrieval configuration items allowed by FastGPT. - Knowledge base retrieval results include irrelevant internal administrative documents. This is caused by failing to correctly configure
collectionIds, and not limiting retrieval to knowledge base collections related to creative testing. - Extra spaces or incorrect capitalization appear in model outputs. This is caused by not enabling
usingReRankordatasetSearchUsingExtensionQuery, leading to unadapted retrieval result formats, or format deviations introduced by expanded query terms.
How to confirm the configuration is complete
- Upload category-specific documents such as creative scripts and material test sheets, perform a test query, and verify that retrieval results only include content within the target collection.
- Adjust the
similarityparameter range, verify that the relevance of retrieval results meets the business needs of creative testing, and adjust the threshold as needed. - Trigger an internal work order query, check whether industry-related query terms are automatically expanded, and confirm that
datasetSearchUsingExtensionQueryis active. - Check system logs to confirm that there are no unauthorized configuration parameter errors, and verify that all used configuration items are on the official support list.
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.