Reference only; verify against the target FastGPT release and deployment documentation.
What This Category of Data Looks Like
Coal chemical policy and regulatory basis data is mainly sourced from official documents publicly issued by the National Development and Reform Commission, National Energy Administration, Ministry of Emergency Management, and industry competent authorities. Updates are synchronized with new policy releases and revisions of existing clauses. Most documents are in PDF format, with some having official online web versions. Their structure includes document numbers, issuing authorities, release dates, main text clauses, attachments and supplementary tables. Fields include policy ID, applicable scenarios, and implementation standards. Commonly used units include tons of standard coal, ten thousand tons of production capacity, operating hours, and similar units.
What Constraints Do These Characteristics Impose on Document Parsing and Chunking
Coal chemical policy documents are mostly PDFs with attachments, some containing embedded charts and process formulas. Some documents are available via public Yuque links. Clause hierarchies are clear but long text is common. Splitting must avoid breaking policy clause logic. Public link formats must comply with official release standards; non-standard links may cause parsing failures. Embedded images must be extracted and linked to corresponding text blocks, otherwise the large language model cannot access the policy explanation content associated with the images. Long documents have high risk of parsing timeout, so timeout configuration must be adjusted to match file size.
Configuration Settings
| Config Item | Recommended Value | Rationale |
|---|---|---|
chunkSize | 800–1200 characters | Single clause length of coal chemical policy documents mostly falls between 500-1000 characters. This range preserves clause integrity and avoids over-splitting or overly large chunks |
chunkSplitter | Split by paragraph + heading | Policy documents are mostly structured by clauses and sections. Splitting by paragraph + heading preserves the complete logic of individual policy items |
customPdfParse | Enable embedded image extraction | Coal chemical policy documents often include embedded images such as process flow diagrams and energy consumption supplementary tables. Enabling this feature links images to corresponding text blocks |
PARSE_FILE_TIMEOUT_SECONDS | 300–600 seconds | Single policy documents typically have 10-50 pages. This timeout range covers the parsing time of long documents and avoids premature interruption |
UPLOAD_FILE_MAX_SIZE | 1000 MB | Some coal chemical policy collection documents have large file sizes. This upper limit supports complete file upload and parsing |
chunkSettingMode | Manual configuration | Adjustment of chunking rules is required to match the clause structure of coal chemical policies. Manual configuration allows flexible adaptation to document characteristics |
Three Common Mistakes
- When a public shared Yuque link is uploaded, the knowledge base parsing task fails, and the interface prompts that the data source type cannot be identified. The cause is that a fully public share permission link is not used. Yuque links with internal or partial share permissions cannot be parsed normally.
- Parsed coal chemical policy document text blocks do not include embedded process flow diagrams or energy consumption supplementary table images. The large language model cannot access the policy explanation content associated with the images. The cause is that the embedded image extraction function of the
customPdfParseconfiguration is not enabled. - Parsing of a single coal chemical policy collection document with more than 50 pages is interrupted, and a timeout error prompt is returned. The cause is that the
PARSE_FILE_TIMEOUT_SECONDSsetting value is too low, and does not cover the full parsing time of the long document.
How to Confirm the Configuration Is Correct
- A coal chemical policy PDF containing embedded process flow diagrams may be uploaded, and the text in the parsing result checked for links to corresponding image markers.
- The parsed text blocks may be checked for retention of complete policy clause hierarchies and document number information, with no incorrect clause splitting observed.
- A large coal chemical policy collection document may be uploaded, and the parsing process confirmed to have no timeout interruptions.
- The
chunkSizeparameter may be manually adjusted, chunking results across different configurations compared, and the chunking logic confirmed to match the structural characteristics of the policy documents.
The values given 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.