What data for this category looks like
Data for municipal gas service inquiries is primarily sourced from official service guides of local urban gas companies, offline lobby display materials, and annual service update documents. Updates occur quarterly or annually, with most changes involving adjustments to regional service rules or fee standards. Each document centers on an individual service item, and includes fields such as item name, processing location, required materials, processing workflow, fee standards, and contact numbers. Units include yuan, cubic meters, and working days, among others. Some official documents are released as digitally signed PDFs, and structured table documents make up a relatively high share of total files.
What constraints do these characteristics impose on document parsing and chunking?
The structured nature of municipal gas service documents requires chunking to preserve the integrity of individual service items. Independent items such as account opening and ownership transfer must not be split across different chunks. Digitally signed PDF documents cannot be recognized by the default parsing engine. Additional compatible rules must be configured to support these files. Long regional service collection documents may exceed the default parsing time limit. The timeout threshold must be adjusted to avoid parsing failures. Fee tables, process diagrams, and other content in documents must be chunked together with surrounding text. Failure to do so will prevent the large model from associating images with their corresponding descriptions. Service document formats vary across regions, and some documents include cross-regional general service descriptions. Chunking must balance regional attributes and item independence, requiring flexible chunking rules.
How to set configurations
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
chunkSize | 800–1200 characters | Municipal gas service documents mostly consist of coherent process descriptions and independent item lists. This length preserves complete context for a single item, and avoids splitting key processing information |
chunkSplitter | Split by title and paragraph boundaries | Service documents divide independent modules using titles such as "Account Opening Process" and "Ownership Transfer Notice". Splitting in this way preserves the integrity of each service item |
customPdfParse | Enabled | Some official service documents include digital signatures. Enabling custom PDF parsing supports this format type, and avoids empty content extraction |
PARSE_FILE_TIMEOUT_SECONDS | 120 seconds | Large regional gas service collection documents have lengthy content. This duration ensures complete parsing, and avoids timeout failures |
UPLOAD_FILE_MAX_SIZE | 1000 MB | Allows uploading batch multi-region service guide collections, and adapts to the need for bulk import of multiple documents |
chunkSettingMode | Custom mode | Allows adjusting chunking rules for documents such as fee standard tables, and adapts to chunking and association requirements for structured data |
Three common misconfigurations
- After uploading a digitally signed gas service PDF, the knowledge base displays empty file content. The cause is that the
customPdfParseconfiguration is not enabled. The default parsing engine cannot be compatible with digitally signed official file formats, resulting in failure to extract text content. - After the API knowledge base returns a document reading link, clicking it triggers an error prompting "Only support .txt, .m". The cause is that the uploaded document format is not supported by the default parsing engine, and custom parsing rules are not enabled via the
customPdfParseconfiguration. - After uploading a normally parsed service document, the large model’s output answer does not include content related to images or tables in the document. The cause is that the chunking configuration does not associate images and tables with surrounding text, or parsing binding for multimedia elements within the document is not enabled.
How to confirm configurations are correct
- Upload one digitally signed gas service PDF. Verify if complete text content can be extracted in the knowledge base. Confirm that the
customPdfParseconfiguration is enabled. - Upload a gas service collection document larger than 100 MB. Check whether the parsing task completes within 120 seconds without timeout errors. Confirm that the
PARSE_FILE_TIMEOUT_SECONDSconfiguration is appropriate. - Upload a Word document containing a fee table. Check whether the chunking result preserves the complete context of the paragraph containing the table. Confirm that the
chunkSplitterandchunkSizeconfigurations adapt to structured documents. - In FastGPT v4.17.0, test uploading gas service documents of different formats. Confirm that all formats can be parsed normally without format errors.
The values provided 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.