Operations Duty and Internal Ticket Q&A for SLA and Service Reports: Knowledge Base Retrieval and Recall

Data sources for this category include monitoring systems, ticket systems, contract documents, and customer communication records. Update frequency is

What This Category of Data Looks Like

Data sources for this category include monitoring systems, ticket systems, contract documents, and customer communication records. Update frequency is monthly or quarterly. Daily documents include SLA clauses, monthly service reports, availability statistics tables, and customer review minutes. Fields include service time indicators, customer contract identifiers, event occurrence times, compliance check items, and more. There is no fixed uniform single-document length. Data is grouped and stored according to business archive cycles.

What Constraints These Characteristics Impose on Knowledge Base Retrieval and Recall

This category of data is scattered across multiple systems. It includes structured contract clauses and ticket records, plus unstructured review minutes. Retrieval logic must support multiple formats. Update frequency is monthly or quarterly. High-frequency incremental synchronization is not required. Indexes must be updated according to archive cycles to maintain data timeliness. Documents contain compliance check-related fields. Non-authorized content must be filtered during retrieval to avoid compliance risks. SLA-related indicators are the core retrieval targets. Retrieval ranking must combine field relevance and keyword matching accuracy. Irrelevant content must be prevented from interfering with SLA indicator calculation and report caliber alignment.

How to Configure Settings

The following configurations are based on FastGPT v4.17.0.

Configuration ItemRecommended ValueRationale
similarity0.75–0.85Matches keyword accuracy for SLA clauses and ticket records, balances recall completeness and relevance
limitTop 10 entriesControls the number of documents retrieved per request, avoids excessive content slowing response, meets the small-batch precise retrieval needs of SLA reports
searchModehybridCombines keyword and semantic retrieval, adapts to mixed structured and unstructured SLA category data
usingReRankEnabledRe-ranks retrieval results, prioritizes passages related to SLA core indicators, improves report caliber alignment efficiency
datasetSearchUsingExtensionQueryEnabledAutomatically optimizes user-input SLA-related questions, supplements implicit retrieval terms such as contract numbers and time thresholds
collectionIdsConfigured by business archive groupsOnly retrieves corresponding monthly/quarterly archived SLA document sets, narrows retrieval scope and improves response speed

Three Common Configuration Errors

  • API calls return no knowledge base matching results. The cause is failure to bind the correct collectionIds parameter in the API request, or failure to enable knowledge base retrieval permissions.
  • Retrieval response time exceeds 30 seconds. The cause is enabling both usingReRank and datasetSearchUsingExtensionQuery, with an excessively high retrieval limit value, exceeding single-request processing thresholds.
  • Retrieval results do not link to specified file names. The cause is failure to set searchMode to a multi-source document compatible mode, or failure to enable file name association indexing in retrieval parameters.

How to Confirm Correct Configuration

  • Check that API request parameters include the configured collectionIds, and that they match the target SLA document sets.
  • Review retrieval logs to confirm that the switch states of usingReRank and datasetSearchUsingExtensionQuery match the configured settings.
  • Initiate a test retrieval, input a question containing SLA time indicator keywords, verify that the relevance and quantity of returned results meet the configured limit and similarity requirements.
  • Verify response time, confirm that the duration falls within an acceptable business range for the configured parameter combination.

The values listed are common starting points, and should be evaluated against samples specific to the target use case.

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.