What Data for This Category Looks Like
Data sources for water utility due diligence reports include four types:
- Daily pipeline operation logs of water utility operators
- Monthly water quality monitoring bulletins from water conservancy departments
- Revenue settlement reports of water supply enterprises
- As-built drawings of pipeline renovation projects
Update frequencies vary across sources:
- Operation logs are updated daily
- Water quality monitoring bulletins are updated weekly
- Revenue reports are updated monthly
- Engineering drawings are only updated after renovation completion
Document structures include three types:
- Structured tables, such as water quality indicators and pipeline pressure data
- Unstructured text, such as operation records
- Vector graphics, such as pipeline topology diagrams
Fields included are:
- Water supply scale (unit: cubic meters per day)
- Pipeline pressure (unit: megapascals)
- Turbidity (unit: NTU)
- Water loss (unit: cubic meters)
Generic percentage-based statistical fields are not used.
Constraints for Citation Sources and Traceability
These characteristics create the following constraints for the traceability workflow:
- Multi-source and heterogeneous data sources require traceability to associate at least three types of data sources, to avoid information bias from a single source.
- Different update frequencies require verifying data timestamps during traceability, prioritizing the latest released monitoring data and operation records.
- Complex document structures require parsing to distinguish traceability markers for structured tables and unstructured text, to avoid confusing chart annotations with core parameters.
- Specific field units require matching parameter standard units during traceability, preventing mixed use of megapascals and kilopascals.
- Low-frequency updated engineering drawings need separate version marking to ensure the latest pipeline topology data is referenced.
Configuration Settings
| Configuration Item | Recommended Value | Rationale |
|---|---|---|
recall count | top 8 | Water utility due diligence reports include multi-dimensional professional parameters, requiring sufficient context to support citation accuracy |
similarity threshold | 0.72–0.78 | Water utility documents have high density of professional terminology, requiring high matching accuracy to filter irrelevant results |
chunk length | 1000–1200 characters | Water utility documents contain long sections of pipeline parameters and inspection records, to avoid truncation of critical information |
PARSE_FILE_TIMEOUT_SECONDS | 300 seconds | Large water utility reports include multi-page charts and tables, resulting in longer parsing time |
source data metadata extraction toggle | enabled | Metadata such as publication time and testing institution of water utility documents must be extracted for traceability |
reordered return count | top 3 | Prioritize displaying the most authoritative data sources, aligning with the rigor requirements of due diligence reports |
The parameter values provided on this page are common starting points for configuration. Actual values are affected by material format, data volume and business rules. Specific issues require case-by-case analysis. It is recommended to test on your own samples before finalizing settings.
Three Common Misconfigurations
- A
Cannot redefine property: toStringerror occurs when uploading large water utility operation log files. The cause is incorrect use of global property overrides in custom parsing scripts, which causes FastGPT's built-in toString method to be redefined. - Recall results do not associate the detection time of the original data source. The cause is that the
source data metadata extraction toggleis not enabled, and key traceability metadata of water utility documents is not extracted. - The number of recall results is less than the configured
recall count, failing to meet the information completeness requirements of due diligence reports. The cause is that the similarity threshold is set too high, filtering out some matching results that meet the requirements for water utility professional parameters.
How to Verify Correct Configuration
- Upload a standard water utility inspection report, check the metadata fields in the knowledge base parsing results, and confirm that information such as publication time and testing institution has been correctly extracted.
- Initiate a query related to water utility due diligence, check the traceability tags of recall results, and confirm that each result is associated with the corresponding data source name and publication time.
- Adjust the similarity threshold and initiate a test query, verify that the number of recall results changes as expected.
- Click the traceability link of the recall result, confirm that it can jump to the original data source page.
Question material comes from public community discussions. Configuration values are common starting points and should be measured against your own samples. Verified on 2026-09-14.