What This Category of Data Looks Like
This category of data comes from ticket systems, internal chat logs, product operation logs, and customer contract SLA documents. Update frequency is daily or real-time. Daily generated documents include customer tickets, problem classification tables, processing records, and service reports. Fields include ticket unique identifiers, problem description text, classification tags, associated processors, SLA constraint items, and more. Each data entry is linked to a specific service scenario and responsible entity.
What Constraints These Characteristics Impose on Knowledge Base Retrieval and Recall
Dispersed data sources and flexible update frequencies require the knowledge base retrieval link to support cross-source recall of multiple data types, including ticket systems, chat logs, and product logs. This prevents missing real-time generated tickets and abnormal information. Structured documents such as problem classification tables and processing records require retrieval to match corresponding classification tags and responsible entity fields. This ensures recall results align with the current ticket’s classification dimensions. The real-time requirements of duty scenarios require the retrieval link to support low-latency responses. Without this, rapid problem reproduction and impact scope judgment cannot be completed. Documents in different formats require unified preprocessing. This avoids reduced recall accuracy caused by differences in text structure.
How to Set the Configuration
Configuration items under FastGPT v4.17.0 are as follows:
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
similarity | 0.75–0.85 | Matches the semantic similarity between ticket problem descriptions and historical processing records, filtering out low-relevance recall results |
limit | Top 10 entries | Controls the total token count of recall results to adapt to model context window limits |
searchMode | Vector search + keyword hybrid | Balances semantic matching and precise keyword matching, adapting to both standardized and colloquial problem phrasing for tickets |
usingReRank | Enabled | Performs secondary ranking on initial recall results, raising the priority of highly relevant historical tickets and classification tables |
datasetSearchUsingExtensionQuery | Enabled | Expands user-input questions into synonymous phrasing to cover more potentially relevant internal documents |
collectionIds | Bound to the operation ticket-specific knowledge base collection | Limits the retrieval scope to avoid mixing in non-operation business documents |
Three Common Mistakes
- Symptom: When clicking the reading link returned by the API retrieval knowledge base, an error
Only support .txt, .mis prompted. Cause: The document upload verification rules of the knowledge base are not correctly configured, only allowing uploads of specified file formats. Non-compliant format documents cannot generate accessible reading links. - Symptom: The number of retrieval results does not match the configured
limitparameter, with recall results exceeding or falling short of the set quantity. Cause: ThecollectionIdsparameter is not correctly bound, and the retrieval scope covers multiple unrelated knowledge base collections, leading to abnormal recall result quantities. - Symptom: The imported knowledge base fails to generate accurate matching results, and the response content deviates from the core demands of the ticket problem. Cause: The
usingReRankparameter is not enabled, and no secondary ranking is performed on the initial recall results, causing low-relevance documents to occupy high-priority positions.
How to Confirm the Configuration Is Correct
- Submit a test ticket, check the number of results returned by retrieval, and confirm that it matches the configured
limitparameter. - Input a test question containing classification tags, check whether the recall results include the content of the corresponding classification table, and confirm that the
similaritythreshold configuration meets business requirements. - Click the document link returned by retrieval, confirm that it can be accessed normally, and check whether the document format verification rules are correctly configured.
- Compare the retrieval results when
usingReRankis enabled and disabled, and confirm that the sorting priority of highly relevant documents meets expectations.
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.