What Data for This Category Includes
Data for this category originates from CMDB, cloud resource inventories, automated discovery tools, and procurement ledgers. Daily documentation covers configuration item lists, service component relationship topologies, change records, and asset inventory sheets. Data is updated on a daily or weekly basis. Document structures include structured tabular configuration item lists, topology diagrams with associated relationships, and time-sequenced change logs. Fields include configuration item unique identifiers, associated asset models, associated dependencies, update times, procurement batch numbers, and more. Units include device numbers, IPv4 addresses, GB, days, and others.
Constraints for Knowledge Base Retrieval and Recall
Data for this category is scattered and sourced from multiple locations. Retrieval must support associated recall across multiple datasets such as CMDB and cloud resource inventories, to avoid missing core dependency relationships. Data is updated daily or weekly, so incremental data must be synchronized regularly to ensure retrieval results align with the latest configuration status. Documentation includes both structured and unstructured formats, so retrieval and parsing must adapt to different content types. Dependency relationships between configuration items are core retrieval targets. Content with high relevance must be recalled first, and isolated configuration items should be excluded. Errors from mismatched asset book and physical records require retrieval results to match the confidence level of official CMDB data, to avoid deviations.
Configuration Settings
These configurations apply to FastGPT v4.17.0.
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
similarity | 0.75–0.85 | Matches the precision requirement for CMDB configuration items, to avoid recalling irrelevant, non-associated asset data |
limit | Top 20 entries | Adapts to the maximum number of configuration items returned by a single CMDB retrieval, to avoid excessive redundant data causing context overflow |
searchMode | hybrid | Combines keyword and semantic retrieval, to balance configuration item name matching and semantic association of dependency relationships |
usingReRank | Enabled | Re-ranks recalled configuration items by relevance, to prioritize returning core dependency items and change records |
datasetSearchUsingExtensionQuery | Enabled | Expands the retrieval scope to associated cloud resource inventories and procurement ledgers, to cover all category data sources |
collectionIds | Configured per CMDB-specific dataset | Limits the retrieval scope to the corresponding CMDB dataset, to avoid mixing asset data from other categories |
Three Common Configuration Mistakes
- Symptom: The number of configuration items returned by retrieval stays fixed at 10, and cannot be adjusted via the interface or API. Cause: The
limitparameter is not configured correctly, or the bound dataset has a default return count limit. - Symptom: The
usingReRankfunction works normally during knowledge base test retrieval, but re-ranking does not activate when calling the large model. Cause: The re-ranking switch for knowledge base retrieval is not enabled during large model calls, and the configuration was only enabled in separate knowledge base tests. - Symptom: Duplicate configuration item entries appear in the knowledge base, and automatic deduplication does not function. Cause: The knowledge base file deduplication configuration is not enabled, or duplicate configuration items exported from CMDB have not undergone preprocessing.
How to Verify Successful Configuration
- Initiate a retrieval targeting core configuration items, and verify that the number of returned results matches the upper limit set by the
limitparameter. - Compare results from knowledge base test retrieval and large model calls, to confirm that the sorted order of re-ranked results is consistent.
- View the metadata fields of retrieval results, to confirm that each result is associated with the corresponding data source, such as CMDB or cloud resource inventory.
- Simulate an updated data change, and verify that retrieval results can synchronize the latest configuration item information within the update cycle.
The values provided here are common starting points, and should be evaluated against the user’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.