What the data for this category looks like
Urban water utility document and material drafting involves pipeline operation and maintenance logs, water quality test reports, water supply scheduling plans, drainage emergency response plans, water utility project bidding documents, water fee accounting ledgers, and more. Data sources include internal operation and maintenance systems, third-party testing institution reports, and project archived files. Update cadence is as follows: operation and maintenance logs are synced in real time, test reports are updated quarterly or monthly, and project files are updated according to project progress. Document structure includes structured fields such as pipeline number, test point, water quality indicator value, and timestamp, as well as unstructured documents such as emergency response plan PDFs and Word plans. Units include professional water utility units such as cubic meters per hour, mg/L, and pH value. Document lengths range from a few pages of operation guides to dozens of pages of special plans.
What constraints these characteristics impose on the knowledge base retrieval and recall link
The mixed structured and unstructured data characteristic requires the retrieval link to balance field matching and semantic matching, to avoid retrieving irrelevant non-water utility data. Different update frequencies require setting incremental sync frequencies to regularly update the latest operation and maintenance data and test reports in the knowledge base. The presence of multiple fields and professional units requires matching corresponding fields and units during retrieval, to avoid retrieval errors caused by unit confusion. Large differences in document lengths require adapting chunking parameters, to avoid over-cutting long documents or insufficient chunking for short documents.
How to set the configurations
| Configuration Item | Recommended Value | Rationale |
|---|---|---|
similarity | 0.75–0.85 | Matches the precision requirement for water utility professional terminology, to avoid retrieving irrelevant non-water utility data |
limit | Top 8 entries | Single-round retrieval sufficiently covers the core content of water utility documents, to avoid context length exceeding the context window |
searchMode | Professional mode | Adapts to the field matching requirements for structured water utility data, and matches water utility professional content more accurately than general mode |
usingReRank | Enabled | Performs secondary ranking for long documents such as water quality test reports and emergency response plans, to improve relevance |
datasetSearchUsingExtensionQuery | Enabled | Performs extended queries for water utility professional terminology, for example, "pipeline leak" is extended to "water supply pipe leakage" to improve the recall rate of similar queries |
collectionIds | Corresponding water utility-specific knowledge base collection ID | Isolates knowledge base data from other categories, to ensure only water utility-related documents are retrieved |
Three common mistakes
- Phenomenon: Retrieval results do not call knowledge base content, and generated water utility documents deviate from professional water utility content. Cause:
collectionIdsis not configured correctly, and non-water utility knowledge base collections are incorrectly associated, resulting in retrieved data deviating from the target dataset. - Phenomenon: Extracted knowledge base information is empty in the workflow. Cause: The
similaritythreshold is set too high, exceeding 0.85, resulting in no matching qualified water utility documents, or chunking does not adapt to the length of water utility documents. - Phenomenon: Similar queries retrieve incorrect content. Cause:
datasetSearchUsingExtensionQueryis not enabled, and water utility professional similar terms are not extended, resulting in similar queries failing to accurately match corresponding water utility documents.
How to confirm the configuration is correct
- View the knowledge base management interface in FastGPT v4.17.0, and confirm that
collectionIdsonly associates water utility-specific knowledge base collections. - Test retrieval for "pipeline leak emergency response", check that the retrieval results include water utility-related documents, and that the
similaritythreshold meets the preset range. - After enabling
usingReRank, test the ranking results of long documents, confirm that documents with higher relevance are ranked first. - Test retrieval of similar queries, confirm that after enabling
datasetSearchUsingExtensionQuery, the retrieval results of similar queries are more accurate.
The values given 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.