What the Data for This Category Looks Like
Water utility financial and operational data primarily comes from annual and semi-annual public reports of listed water utilities, public regulatory documents from housing and urban-rural development departments and ecological environment departments, and monthly operational reports of water supply and wastewater treatment enterprises. Data update cycles fall into three categories: annual (annual reports), quarterly (operational briefings), and monthly (real-time monitoring data). Document structures include fields such as tap water supply revenue, wastewater treatment costs, pipeline operation and maintenance length, and pollutant discharge indicators. Units cover professional measurement standards including ten thousand yuan, cubic meters, tons, mg/L, and other standard units. A single core data entry is usually tied to specific business scenarios and corresponding compliance requirements.
Constraints Imposed by These Characteristics on Citation Source and Traceability
The multi-source and decentralized nature of water utility financial reports requires traceability to link two types of data sources: enterprise announcements and regulatory documents. This avoids incomplete traceability caused by relying solely on a single channel. Different update frequency data requires corresponding time filtering rules. For example, annual financial reports must match full accounting cycles, and monthly monitoring data must be limited to the last 30 days. Precise matching of professional fields such as COD and BOD discharge values requires the retrieval system to retain field-level associations. Matching only general text will result in irrelevant content being included. Additionally, the compliance attribute of water utility data requires traceability information to clearly point to document numbers, page numbers, or monitoring site identifiers. Using only file names cannot serve as a complete traceability basis.
How to Configure the System
| Configuration Item | Recommended Range | Rationale |
|---|---|---|
similarity threshold | 0.65–0.75 | Water utility financial reports contain a large number of professional terms and specific business scenarios. A threshold that is too low will mix in general text from non-water utility fields |
recall count | Top 8–12 entries | Water utility-related data is scattered across annual reports, regulatory documents, and operational reports. A sufficient recall volume can cover all relevant corpus |
maxContext | 1500–2000 characters | A single water utility financial report entry contains multiple sets of professional fields. An overly long context will exceed the model's processing limit |
citation source field mapping | Configured as "data source type + document number + page number" | Traceability must clearly point to specific regulatory documents or enterprise announcements. Extracting only file names cannot meet compliance traceability requirements |
data source time range filtering | Configured as "annual reports by year, operational data by quarter" | Adapts to the multi-cycle update characteristics of water utility data, avoiding recall of expired or irrelevant historical data |
PARSE_FILE_TIMEOUT_SECONDS | 120 seconds | Water utility financial reports may contain a large number of pipeline data tables. Parsing takes longer than general documents |
The parameter values provided on this page are common recommended starting points for configuration. Actual values are affected by material format, data volume, and business rules. Specific issues require individual analysis. It is recommended to test on your own samples before finalizing settings.
Three Common Configuration Mistakes
- Phenomenon: The number of citations displayed on the interface exceeds the configured
recall count. Cause: The association betweenmaxContextand recall count is not bound, and the system automatically supplements redundant corpus. - Phenomenon: COD discharge data from water utility financial reports cannot be recalled. Cause: The
similarity thresholdis set to 0.4, which is too low and mixes in a large amount of general text from non-water utility fields. - Phenomenon: Citation sources only display file names without specific page numbers or document numbers. Cause:
citation source field mappingis not configured, and only basic file name information is extracted.
How to Verify Successful Configuration
- Upload a public annual report of a water utility and a monitoring report from the ecological environment department. After executing a retrieval, check the citation source field of the returned results to confirm that it includes data source type, document number, and page number information.
- Adjust the
similarity thresholdto the set range, retrieve content related to water utility pipeline operation and maintenance, and check whether the relevance of the returned results meets expectations. - Initiate retrieval requests for different time ranges, and confirm that only water utility financial reports and operational data within the corresponding cycle are recalled.
- View the document parsing log to confirm that the parsing time of a single water utility financial report does not exceed the set value of
PARSE_FILE_TIMEOUT_SECONDS.
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.