What the data for this category looks like
The data for this category comes from four main sources: advertising platform APIs, internal material libraries, user interaction data, and conversion data. Daily produced documents include creative scripts, material test sheets, delivery data reports, and iteration checklists. Creative scripts contain copy content and associated material IDs. Material test sheets mark test batches and performance metrics. Delivery data reports record fields such as impressions, click-through rate, and conversion cost. Most documents are updated daily or weekly. Some high-frequency business data requires daily synchronization to the knowledge base.
What constraints these characteristics impose on knowledge base retrieval and recall
The data sources for this category are scattered, and update rhythms vary significantly. This requires the knowledge base retrieval to support unified recall of multi-source data. Daily documents have distinct structural differences. Fields of different documents are strongly tied to specific business scenarios. Retrieval must accurately match specific fields for different document types, such as creative scripts and test data. Creative variable design and test conclusion induction rely on real-time or near-real-time recall results. This places high requirements on retrieval response speed. Inaccurate recall results directly affect the accuracy of creative testing. This creates risks of reduced click-through rates and increased delivery costs. Therefore, retrieval recall accuracy must meet the confidence requirements of business judgment.
How to set the configurations
| Configuration Item | Recommended Value | Rationale |
|---|---|---|
similarity | 0.75–0.85 | Matches the semantic relevance requirements of creative testing businesses, and balances recall accuracy and result filtering effects |
limit | Top 8 entries | Controls the number of results returned per recall, to avoid overly long context that affects model understanding |
searchMode | Vector recall | Adapts to semantic matching requirements for multi-source business data, and supports associated recall of creative scripts and delivery data |
usingReRank | Enabled | Improves the sorting priority of highly relevant results, and optimizes the recall accuracy of key data in creative testing |
datasetSearchUsingExtensionQuery | Enabled | Expands query terms based on business scenarios, and supplements potential retrieval needs related to creative variables |
collectionIds | Configured according to business-specific knowledge base IDs | Limits the retrieval scope to avoid interference from irrelevant data on recall results |
Three common mistakes
- Knowledge base retrieval takes too long. This happens when the
collectionIdsretrieval scope is not limited, or theusingReRankoptimization for recall logic is not enabled. This results in an overly large retrieval scope and redundant return results. - The open-source version does not support custom prompt words and reference templates. This is because FastGPT v4.17.0 and earlier open-source versions require advanced AI configuration to be enabled in the corresponding function module. FastGPT v4.8.22 only supports basic parameter adjustments, and does not include custom template functions.
- Knowledge base chat works during debugging, but the online chat menu or guest link throws the error common:core.chat. This occurs when cross-domain access permissions for the knowledge base are not configured, or knowledge base data source configurations are not updated after going live. This causes parameter inconsistencies between the online environment and the debugging environment.
How to Confirm Configurations Are Set Correctly
- Run a retrieval test to confirm that the number of returned results matches the
limitparameter, and that results cover target document types such as creative scripts and delivery data. - Adjust the
similarityparameter to verify that retrieval result relevance meets business judgment standards, with no excessive irrelevant materials or data. - Check the open-source version configuration page to confirm that
datasetSearchUsingExtensionQueryandusingReRankfunctions are enabled, and corresponding modules are visible. - Troubleshoot online error logs to confirm that cross-domain and data source configurations related to common:core.chat have been completed and synchronized.
The values provided here are common starting points and should be evaluated using samples tailored to the specific deployment context.
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.