What the data for this category looks like
The internal office data for urban water utilities primarily comes from pipeline operation and maintenance logs, water quality monitoring reports, customer service ticket systems, internal job training manuals, and equipment maintenance procedures. Data update cycles fall into two categories: daily and periodic. Real-time monitoring data from inspection points is updated daily, while annual maintenance plans and cross-department collaboration process documents are updated quarterly. Most individual documents are structured tables or step-by-step instructions, with fixed fields including monitoring point number, monitoring time, water quality indicator values, and ticket number. Supporting units include mg/L, MPa, hours, and similar units.
What constraints these characteristics impose on knowledge base retrieval and recall
A high share of structured data requires retrieval to support field-level precise matching, to avoid semantic recall that mixes monitoring data from unrelated points. Real-time data updated daily requires incremental synchronization rules, to prevent full synchronization from consuming excessive system resources. Long documents such as equipment maintenance procedures include multi-step workflows; when chunking, context relevance must be preserved to avoid losing operational logic after splitting. Unique identifier fields such as ticket numbers and point numbers must be used as mandatory recall conditions, to ensure results match specific business scenarios.
How to configure settings
Reference only; verify against the target FastGPT release and deployment documentation.
| Configuration Item | Recommended Value | Rationale |
|---|---|---|
similarity | 0.72–0.80 | Matches the field precision requirements for structured urban water utility data. A low setting risks recalling unrelated point data, while a high setting risks missing valid ticket records. |
limit | Top 8 entries | Covers common information needs for single-round employee inquiries, such as multiple sets of monitoring data for inspection points. Prevents excessive results from interfering with reading. |
searchMode | vector | Prioritizes matching semantic and keyword associations for structured fields such as water quality indicators and point numbers, adapting to the structured nature of water utility data. |
usingReRank | Enabled | Re-ranks recall results by business relevance, prioritizing recently updated inspection data and urgent tickets. |
datasetSearchUsingExtensionQuery | Disabled | Most water utility data uses structured fields, so no additional query term extension is needed. This avoids introducing content from unrelated business scenarios. |
collectionIds | Specify the exclusive collection ID of the internal water utility knowledge base | Isolates knowledge base data from other categories, ensuring recall results only come from internal office documents for urban water utilities. |
Three common configuration mistakes
- A 401 Unauthorized error occurs when calling an external knowledge base API. This happens because FastGPT automatically adds a Bearer token to requests, which does not match the target API’s authentication rules.
- Recall results include a large number of non-water utility related documents. This is because
collectionIdsis not configured to specify the exclusive knowledge base collection, resulting in mixed datasets from other categories. - The number of single recall results exceeds expectations. This is because a reasonable value for
limitis not set, returning more entries than required by the business scenario.
How to confirm configurations are properly set
- View the knowledge base collection configuration page, confirm that
collectionIdsis bound to the current urban water utility internal office dataset. - Initiate a test query containing a point number or water quality indicator, check whether the matching score of returned results falls within the preset range.
- After enabling
usingReRank, verify that the order of recall results prioritizes recently updated documents. - After disabling
datasetSearchUsingExtensionQuery, confirm that no additional extended content is automatically added to the query terms.
The values provided here 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 2026-09-28.