What the data for this category looks like
Urban gas application materials and case review data mainly comes from the business management systems of urban gas operation enterprises and public service platforms of administrative examination and approval departments. Data is updated on a per single application project basis. Each individual case includes application subject qualification documents, gas pipeline design drawings, site investigation records, review and approval documents, and other content. Documents contain structured fields and unstructured attachments. Structured fields cover application number, gas usage address, pipe diameter specification, application date, review status, and other items. Units include engineering standard units such as millimeters, cubic meters, and kilopascals. Some attachments are PDF format scans or CAD drawings.
What constraints these characteristics impose on the citation sources and traceability link
The multi-format mixed characteristics of urban gas declaration cases require the citation traceability link to support both structured field matching and unstructured attachment association. Data updated by declaration batches needs to bind a unique application number to achieve version locking, avoiding confusion from cross-batch reference. Engineering-specific units and fields require automatic association of standard definitions for corresponding fields during traceability, preventing unit conversion errors. Multi-page PDF and CAD drawing attachments need to extract specified page numbers or blocks as citation sources; referencing entire files cannot meet precise review requirements.
How to configure
Reference only; verify against the target FastGPT release and deployment documentation.
| Configuration Item | Suggested Value | Rationale |
|---|---|---|
quoteMaxToken | 1500–2000 token | Urban gas case documents have large single-text volume, so sufficient token must be reserved to accommodate citation fragments and source annotations |
| Max Context | Top 8 matching results | Supporting materials associated with a single case usually do not exceed 10 items; limiting the number of results avoids context overload |
detail | Enabled | Specific fields and page number information of citation sources must be returned to meet the precision requirements of review and traceability |
responseData | Include source metadata | Fields such as application number and review date must be returned synchronously for case association verification |
| Citation Source Display Location | Line-break annotation at the end of main text | Complies with reading habits of engineering review documents without interfering with main content |
Three common mistakes
- Phenomenon: Citation content displays garbled characters, while downloaded original files are normal. Cause: Correct text extraction rules for gas engineering drawing CAD formats or OCR results of scanned PDFs are not configured, leading to encoding parsing failure.
- Phenomenon: When
quoteMaxTokenis set above 3000, the large language model cannot receive context content. Cause: The FastGPT platform has an upper limit on the total token of single-round context; exceeding the threshold triggers automatic truncation, resulting in context loss. - Phenomenon: Uploading declaration data returned by the HTTP interface as a knowledge base citation has no effect. Cause: The data is not converted to the structured format required by the FastGPT knowledge base, and the unique case identification field is not bound.
How to confirm the configuration is correct
- Initiate a search for a gas declaration case, check whether the returned results include source annotations for exclusive fields such as application number and gas usage address.
- Adjust the
quoteMaxTokenparameter, verify that the length of the citation fragment changes with the parameter, and no context truncation error occurs. - Upload a gas engineering PDF drawing, confirm that the search result can accurately locate the specific page number and block of the drawing.
- Test the HTTP interface data import, verify that the imported data source can be correctly associated with the review process of the corresponding case.
The values provided are common starting points and should be evaluated 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.