What this category of data looks like
This category of data primarily comes from switch consoles, wireless controllers, office area monitoring systems, and frontline user trouble ticket records. Daily maintenance documents include network topology diagrams, change tickets, fault troubleshooting records, and wireless coverage inspection forms. Data fields cover device IP addresses, SSID names, VLAN numbers, fault response times, inspection cycles, and more. Data updates occur daily or immediately following network changes. Most individual data entries are small structured records and topology-related graphic and text files.
What constraints these characteristics impose on knowledge base retrieval and recall
Multi-source heterogeneous data sources require the retrieval system to support structured field matching and mixed graphic and text recall, adapting to content in different formats such as topology diagrams, operation and maintenance logs, and inspection forms. The high-frequency update feature requires configuring an incremental synchronization mechanism to avoid recalling outdated network change information. Small structured records are mostly short fields such as device IP addresses and VLAN numbers. This requires limiting the redundancy of single recall results while optimizing the matching weight between keywords and fields. Topology-related graphic and text files require enabling mixed retrieval mode to ensure that questions about network paths and coverage ranges can be accurately recalled.
Configuration settings
This configuration is based on FastGPT v4.17.0.
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
similarity | 0.75–0.85 | Matches the precision requirements for structured fields such as internal network device IP addresses and VLAN numbers. Prevents irrelevant results if set too low, and misses valid fault records if set too high |
limit | Top 8 entries | Internal network operation and maintenance issues mostly involve single-device or single-region faults. 8 results cover common fault troubleshooting paths, and avoids excessive results that increase troubleshooting costs |
searchMode | hybrid | Supports a combination of text keyword matching and vector retrieval, adapting to mixed data types including topology graphics and structured logs |
usingReRank | Enabled | Performs secondary ranking on recall results to optimize the semantic matching degree between fault records and user questions |
datasetSearchUsingExtensionQuery | Enabled | Automatically expands synonyms for operation and maintenance-related questions, such as expanding "WiFi outage" to "wireless access interruption", "SSID unable to connect", and similar terms |
collectionIds | Fill in according to the dedicated operation and maintenance knowledge base collection | Restricts the retrieval scope to only datasets related to internal network and WiFi operations, avoiding mixing documents from other categories |
Three common configuration mistakes
- Entering the same question twice before a result is recalled, with no match on the first retrieval. The cause is that
datasetSearchUsingExtensionQueryis not enabled. The initial retrieval does not cover synonym expansion for operation and maintenance scenarios, leading to a failed first match. - Retrieval results include non-operation and maintenance documents. The cause is that
collectionIdsis not configured correctly, and the retrieval scope is not limited to the dedicated operation and maintenance knowledge base collection. - Unable to find the knowledge base collection ID. The cause is that the collection details page in the dataset management section of the FastGPT backend is not viewed. The collection ID is displayed by default in the URL parameter at the top of the page or in the details card.
How to confirm the configuration is correct
- Enter a typical operation and maintenance question, such as "Poor WiFi coverage on a certain floor", and check whether the retrieval results include documents related to wireless coverage inspection forms.
- View the retrieval log to confirm whether synonym expansion is triggered, verifying that
datasetSearchUsingExtensionQueryis working. - Check the source collection of the retrieval results to confirm that only datasets related to internal network and WiFi operations are included.
- Adjust the
similarityparameter value, test the number of recall results under different thresholds, and ensure that the matching accuracy meets the requirements of the operation and maintenance scenario.
Note: The configuration values provided here are common starting points. They should be evaluated against the reader's own specific samples and use cases.
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.