Reference only; verify against the target FastGPT release and deployment documentation.
What data for this use case looks like
Service entities for this use case include central enterprise strategic investment departments, project companies, state-owned assets supervision departments, and external consulting evaluation institutions. Daily processed documents cover annual investment plans, project proposals, feasibility study reports, due diligence reports, asset appraisal reports, and post-investment evaluation reports. Data is primarily sourced from state-owned assets supervision systems, enterprise ERP and project management systems, industry market data, audit and financial statements, with a quarterly update frequency. The most time-consuming steps include compiling technical, market, financial, and legal feasibility study materials, organizing investment decision-making meeting materials and risk control opinions, and collecting post-investment data to conduct post-evaluations.
What constraints do these characteristics impose on the document parsing and chunking link
Documents have diverse types and complex structures, including structured fields, long text chapters, charts, and appendices. Adaptation is required for format differences across source systems. The quarterly batch update feature requires chunking workflows to efficiently process multiple documents without single-run timeouts. Fields such as investment amounts, progress indicators, and quantified post-evaluation results must be fully preserved, without semantic integrity being broken during chunking. Additionally, chapter separation rules vary across different documents, so chunking logic needs flexible adaptation.
How to set the configurations
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
chunkSize | 800–1200 characters | Adapts to the mixed structure of long text chapters and short field indicators in documents for this use case, balancing semantic integrity and chunking precision |
chunkSplitter | Custom separator + semantic chunking | Matches fixed chapter titles and line-break separated paragraphs in documents, avoiding chunking errors caused by relying on a single separator |
customPdfParse | Enabled | Most documents for this use case are in PDF format, with complex elements including headers, footers, and charts. Custom parsing can extract structured content and filter invalid information |
PARSE_FILE_TIMEOUT_SECONDS | 600 seconds | Single feasibility study reports have significant length. Batch processing requires sufficient parsing time to avoid timeout interruptions |
chunkSettingMode | Manual mode | Adjust chunking rules for different document types such as feasibility study reports and post-evaluation reports, to adapt to multi-format input requirements |
UPLOAD_FILE_MAX_SIZE | 1000 MB | Large feasibility study reports for this use case may contain a large number of high-definition charts, leading to large individual file sizes. This requires adjustment of the upload size limit |
Three common mistakes
- After configuring a custom separator, chunking results either merge multiple independent paragraphs or split a single complete paragraph into multiple chunks. The cause is failing to set
chunkSplitterto "custom separator + semantic chunking" mode, relying only on a single line break for separation, and not adjusting thechunkSizeparameter to match the document chapter structure. - After deploying Marker via Docker and configuring environment variables, processing PDF documents returns an error response. The cause is failing to enable the
customPdfParseparameter, or incorrectly configuring Marker's dependency path, leading to PDF parsing failure. - Timeout errors occur when batch processing quarterly updated documents. The cause is failing to adjust the
PARSE_FILE_TIMEOUT_SECONDSparameter; the default timeout duration is insufficient for processing long feasibility study reports.
How to confirm the configuration is correct
- Upload a typical project proposal PDF, check that chunking results retain chapter titles and complete paragraphs, with no excessive merging or splitting.
- Run a single document parsing test, verify that parsing time does not exceed the preset
PARSE_FILE_TIMEOUT_SECONDSparameter value, with no timeout errors. - Review chunked metadata, confirm that each chunk includes basic traceability information such as the original document filename and upload time.
- Adjust the
chunkSizeparameter to different ranges, compare the semantic integrity of chunking results, and confirm alignment with the reading logic of business documents.
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.