Reference only; verify against the target FastGPT release and deployment documentation.
What data in this category looks like
Daily documents for this category mainly include transfer announcements, transaction guidelines, bid applications, earnest money vouchers, transaction confirmation letters, listing result announcements, and property transfer prospectuses. Data sources are the national public resource trading platform, natural resource trading system, property right exchange system, and state-owned asset supervision system. The update frequency is daily.
Documents contain core fields such as ownership information, target parameters, earnest money requirements, and transaction clauses. Units include square meters, ten thousand yuan, parcels, and others. Document formats vary slightly across different issuing entities.
What constraints do these characteristics impose on the "document parsing and chunking" link
Multiple system sources lead to inconsistent document formats. Some documents contain nested tables and seals, so structured information must be retained during parsing. The daily update frequency requires the parsing process to adapt to high-frequency batch processing. Core information such as ownership and valuation in documents has strong correlation, so chunking must not break logical connections. Some single documents have large file sizes, requiring adaptation to large-file parsing requirements. Parsing errors may lead to consequences such as invalid transactions or disputes over state-owned asset loss, so parsing accuracy must meet required standards.
How to configure the settings
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
chunkSize | 800–1200 characters | Documents in this category include long paragraphs of ownership descriptions and transaction clauses. This range avoids splitting key logical units |
chunkSplitter | Semantic chunking | Information in property rights and transaction documents is closely linked. Semantic chunking preserves contextual logic |
customPdfParse | Enabled | Most documents in this category contain form fields and structured data. Enabling this option retains the original document structure and field information |
PARSE_FILE_TIMEOUT_SECONDS | 600 seconds | Parsing large individual property transfer prospectuses takes a long time. This duration covers parsing requirements for most scenarios |
UPLOAD_FILE_MAX_SIZE | 1000 MB | Some documents include multiple attachments. This upper limit adapts to file upload requirements for most business scenarios |
Three common mistakes
- Phenomenon: After uploading a large property transfer prospectus, the interface shows parsing in progress but has no result for a long time, finally returning a timeout error. Cause: The
PARSE_FILE_TIMEOUT_SECONDSvalue is set too small, failing to cover the full parsing time of large documents. - Phenomenon: Parsed document content has format errors and cannot be imported into the knowledge base normally. Cause:
customPdfParseis not enabled, and the native parsing logic destroys nested structured fields in the document. - Phenomenon: In a locally deployed service, the parsing function does not respond after uploading a document. Cause: The associated parsing service is not configured correctly, and a reasonable
UPLOAD_FILE_MAX_SIZElimit is not set, leading to large file upload failure.
How to confirm the configuration is correct
- Upload a standard transfer announcement document, check whether the parsed text retains core fields such as plot coordinates, transfer term, and earnest money amount.
- Check the current
PARSE_FILE_TIMEOUT_SECONDSvalue, confirm it is greater than the average parsing time of a single document. Reference values can be obtained by uploading a small batch of test documents. - Test uploading multiple attachment documents within 1000 MB, confirm that the parsing queue can process normally without timeout errors.
- Check the enabled status of
customPdfParse, confirm that the parsed document structure has no missing form information or seal area annotations.
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.