What the data for this category looks like
Water utility intelligent due diligence report data mainly comes from monthly operation reports of water utility operators, pipe network monitoring logs, government water regulatory public documents, project feasibility study and bidding documents. The update rhythm varies by document type: operation reports are updated monthly, monitoring logs are collected in real time or hourly, and project documents are only generated during project cycles. Document structures include standardized tables (such as pipe network parameter tables, water quality test details), paragraph-style descriptions (such as operation and maintenance plans), and scattered data entries. Fields and units include pipe inner diameter (millimeters), average daily water supply (cubic meters), COD concentration (mg/L), operation and maintenance duration (hours) and other professional water utility parameters.
What constraints do these characteristics impose on document parsing and chunking
Structured tables (such as water quality test sheets) in water utility due diligence reports contain multiple numerical fields with units. Parsing processes must retain the correspondence between fields and units, to avoid breaking data associations during chunking. Long, dense monitoring log documents from real-time or high-frequency collection must not have cross-entry truncation during chunking, to prevent incomplete data in single chunks. Project documents and operation documents coexist, so it is necessary to distinguish long text paragraphs from short data entries, and avoid forcibly merging operation and maintenance plan paragraphs with pipe network parameter tables. Some documents contain nested tables. Parsing processes must retain the hierarchical structure, otherwise subsequent chunking cannot accurately associate context.
How to set the configurations
| Configuration Item | Recommended Value | Rationale |
|---|---|---|
chunkSize | 800–1200 characters | Adapts to the parsing needs of long text paragraphs and structured tables in water utility documents, avoiding single chunks that are too long or too short |
chunkOverlap | 10–15% | Retains contextual association across chunks, preventing loss of association between structured entries such as pipe network parameters and water quality test data after splitting |
parse_table_mode | Retain original format | Fully retains the correspondence between table fields and units in water utility documents, avoiding separation of fields and units after parsing |
PARSE_FILE_TIMEOUT_SECONDS | 600 seconds | Adapts to the parsing duration requirements of large-scale water utility operation logs and multi-page monitoring reports, preventing mid-parsing timeout interruptions |
enable_ocr | Enable only for scanned documents | For paper-based water utility reports stored as scanned files, perform OCR recognition before parsing and chunking |
The parameter values provided on this page are conventional recommendations used as a starting point for configuration. Actual values are affected by material format, data volume and business rules. Specific issues require targeted analysis. It is recommended to test on your own samples before finalizing settings.
Three common mistakes
- After uploading some water utility PDF documents, the content is displayed as empty, while some documents can be recognized normally. The cause is that some scanned water utility reports do not have OCR parsing enabled, or the PDF is encrypted and cannot read text content.
- Changing the suffix of a Java interface document to TXT and importing it results in no parsed data. The cause is that the document content is in code format, code block parsing mode is not enabled, and the semantic association of pure code text differs significantly from the structured data of water utility due diligence reports. The default chunking logic cannot effectively extract valid information.
- After using chunk mode to call the pushdata API for upload, the index status is displayed for a long time. The cause is that the uploaded water utility document chunk size exceeds the system default threshold, or the API request times out without correctly returning index progress, resulting in task blocking.
How to confirm the configuration is correct
- Upload a typical water utility operation report PDF, and check whether the parsed text fully retains the field and unit information of the table.
- Adjust the
chunkSizeparameter, upload the same document, and compare the length and content integrity of the chunking results to confirm that they meet business requirements. - Call the document parsing test interface, input a text fragment of a water utility document, and verify whether the contextual association after chunking is correct.
- Check the system logs to confirm that no timeout or format error prompts appear for parsing tasks, and that the index progress is completed within a reasonable time frame.
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.