What this category of data looks like
Data for survey and design declaration materials and file review comes from project submission documents and archived files. This includes construction drawing review submission forms, survey results reports, design plan documents, project qualification certification documents, and similar items. Data updates are triggered on demand alongside project submission progress, with no fixed cycle. Document types cover structured Word submission forms, PDF survey reports with charts, scanned drawing instructions, and other mixed formats. Fields include project number, survey scope, qualification level, submission date, and other relevant fields. Units involved include square meters, survey depth (meters), qualification level, and other professional engineering units.
What constraints these characteristics impose on the document parsing and chunking workflow
Structured submission forms in survey materials require precise extraction of fixed fields to avoid splitting professional terms into disconnected fragments. Long, highly specialized non-structured survey report paragraphs require chunk lengths adapted to content density. Mixed-format files may experience parsing format errors, requiring targeted parsing rule configuration. File review requires chunking by project batch to avoid mixing content across separate files. Large archived files include multiple attachments, requiring support for batch parsing and chunking to prevent single-file parsing timeouts.
Configuration guidelines
Reference only; verify against the target FastGPT release and deployment documentation.
| Configuration Item | Recommended Value | Rationale |
|---|---|---|
chunkSize | 800–1200 characters | Survey and design reports have dense professional content. Chunks that are too long cause semantic fragmentation, while chunks that are too short lose contextual relevance. This range is calibrated for professional documents based on real-world testing. |
chunkSplitter | Split by paragraph + heading | Survey materials have clear chapter headings and paragraph structures. Splitting content this way preserves the contextual integrity of professional content. |
customPdfParse | Enable structured extraction | Survey drawing instructions and submission PDFs often contain fixed-format tables and fields. Enabling this setting allows accurate extraction of structured content. |
UPLOAD_FILE_MAX_SIZE | 1000 MB | A single survey project's archived files may include multiple large PDF drawings and Word reports, requiring support for large file uploads. |
PARSE_FILE_TIMEOUT_SECONDS | 600 seconds | Large survey report parsing takes significant time. This setting prevents parsing failures due to timeout. |
chunkSettingMode | Manual configuration | Adjusting chunk parameters is required for different material types such as submission forms, survey reports, and drawing instructions. Manual configuration offers greater flexibility. |
Three common configuration errors
- Symptom: Boolean-type parsed parameters become empty in conditional judgment components. Cause: Standardized extraction of boolean value fields in unstructured text was not performed during document parsing, leading to incorrect parameter type conversion.
- Symptom: Parsing error "Cannot redefine property: toString". Cause: Custom PDF parsing configuration enabled non-standard text rewriting rules, which conflict with the system's built-in toString method.
- Symptom: Search results behave abnormally after the total number of chunks exceeds 3000. Cause: Files were not split by project file batch, causing the total number of chunks for a single knowledge base to exceed the system's default index limit. Large archived files were not split into multiple independent files.
How to verify your configuration
- Upload a typical survey submission material, view the parsed chunk list, and confirm each chunk contains complete professional paragraphs and chapter headings.
- Test the upload and parsing of a single large survey report, confirm that the upload progress and parsing status are normal, and no timeout errors are triggered.
- Enter search terms related to survey and design specialties, and confirm that the search result chunks cover the complete content of the corresponding chapters.
- After adjusting chunk parameters, compare chunk results across different parameter settings to confirm they align with the document structure characteristics of survey materials.
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.