What the data for this category looks like
This category's data comes from regulatory and rule databases, higher-level law databases, departmental policy document libraries, and administrative normative document filing systems. Update frequency is irregular. Daily processed documents include normative document drafts, drafting explanatory notes, legality review opinions, summary of solicited comments, and filing reports. Most documents contain chapters, clauses, and attachments. Fields include drafting entities, review process nodes, clause content, summary of opinions, and more. Documents use document numbers and clauses as basic identification units.
What constraints these characteristics impose on the document parsing and chunking link
Data sources are scattered and formats are inconsistent. This requires the parsing module to support multiple official export formats. Documents have clear hierarchical structures. Chunking must retain the integrity of clauses and chapters to avoid breaking logical units required for review. Review scenarios require chunked content to accurately match higher-level legal basis. Chunk length must balance retrieval accuracy and contextual completeness. Batch file processing must meet upload and parsing timeout and size limits, and adapt to the scale of batch documents exported by filing systems.
How to configure the settings
Reference only; verify against the target FastGPT release and deployment documentation.
| Configuration Item | Recommended Value | Rationale |
|---|---|---|
chunkSize | 800–1200 characters | The length of a single clause in normative documents mostly falls within this range, avoiding cross-clause splitting that breaks review logic |
chunkSplitter | Split by chapter + paragraph | Documents have clear chapter hierarchies, retaining logical units facilitates subsequent legality comparison |
customPdfParse | Enabled | Government normative documents are mostly formatted official PDFs, enabling this enhances parsing to retain text structure and layout information |
PARSE_FILE_TIMEOUT_SECONDS | 300–600 seconds | Parsing batch filing documents takes a long time, this range covers parsing needs for most scenarios |
UPLOAD_FILE_MAX_SIZE | 500 MB | Matches the maximum file size limit for batch files exported from administrative normative document filing systems |
chunkSettingMode | Manual configuration | Chunking logic must be adjusted based on review scenarios to adapt to different types of review documents |
Three common mistakes
- The phenomenon is parsing failure when calling functions related to
customPdfParse, with logs showing API call errors. The cause is failure to configure the mineru API key or activate corresponding interface permissions. - The phenomenon is that boolean type parameters parsed from HTTP components become empty after being processed by conditional judgment components. The cause is that FastGPT's internal type conversion logic does not support the judgment rules for native boolean values.
- The phenomenon is an online environment error: Cannot redefine property: toString. The cause is that a global toString method is repeatedly defined in a custom script, conflicting with the built-in method of the runtime environment.
How to confirm the configuration is correct
- Upload a single normative document draft PDF, view the parsed chunk list, and confirm that each chunk contains a complete clause or paragraph unit.
- View the running logs of the parsing task, confirm that no timeout errors occur, and verify that the configuration of
PARSE_FILE_TIMEOUT_SECONDStakes effect. - Upload a batch file package exported from a filing system, confirm that the upload size limit is not triggered, and verify that the configuration of
UPLOAD_FILE_MAX_SIZEtakes effect. - Retrieve the chunked document content, confirm that the review basis for the corresponding clause can be accurately matched, and verify that the chunking logic meets business requirements.
The values provided are common starting points and should be measured against the user'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.