What data in this category looks like
Data for this category originates from on-site collected materials including customer computer room environments, equipment lists, software packages, and network topologies. Updates occur once per order or project phase. Commonly processed documents include deployment manuals, installation records, configuration lists, and on-site check-in forms. Most document structures follow linear process records, parameter listings, or topology connection annotations, with no unified standardized format. Typical fields include equipment models, IP addresses, installation paths, and check-in times. There are no fixed units, or only the parameters native to the equipment are used.
What constraints do these characteristics impose on document parsing and chunking
Most daily documents are unstructured, with linear processes and scattered parameter listings. Document parsing and chunking must adapt to content with no fixed format, and avoid splitting that breaks the contextual association of deployment steps and equipment parameters. Single on-site collected documents vary widely in size. Parsing configurations must adapt to this range, and avoid parsing timeouts for large documents. The requirement for per-order or per-phase updates means parsing results must quickly link to the deployment nodes of the corresponding project. Configurations must also support chunking logic for different document types, to prevent inconsistent chunking results caused by format differences.
How to set configurations
The following configurations apply to FastGPT v4.17.0, and meet the document parsing and chunking requirements for on-site deployment scenarios.
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
chunkSize | 800–1200 characters | Adapts to the length of parameter paragraphs in documents such as deployment manuals and configuration lists, and avoids splitting critical configuration items |
chunkSplitter | Split by paragraph + semantics | Adapts to the mixed structure of linear processes and scattered parameters in on-site documents, and preserves contextual association |
PARSE_FILE_TIMEOUT_SECONDS | 300–600 seconds | Adapts to the scenario where single on-site documents vary widely in size, and avoids parsing timeouts for large topology-style documents |
customPdfParse | Enabled | Supports enhanced parsing of PDF format documents such as deployment manuals, and extracts image annotations and topology connection information |
UPLOAD_FILE_MAX_SIZE | 1000 MB | Meets the upload requirements for large software package logs and network topology documents collected on-site |
chunkSettingMode | Manual configuration | Adapts to format differences across project documents, and allows flexible adjustment of chunking rules |
Three common mistakes
- Phenomenon: Files uploaded via the file collection API have inconsistent chunking results compared to files uploaded directly on the platform. Cause: The API upload request does not specify the
chunkSettingModeparameter, and the default chunking rules do not match the configuration used for manual platform uploads. - Phenomenon: Documents configured with custom URLs return no data after parsing. Cause: The
customPdfParseconfiguration is not enabled, or the custom URL does not have public access permissions, preventing the parsing node from pulling the content. - Phenomenon: The model cannot read uploaded parsed files. Cause:
PARSE_FILE_TIMEOUT_SECONDSis set too short, causing large documents to be marked as failed before parsing completes, orchunkSizeis set too small, splitting critical deployment steps.
How to confirm configurations are properly set
- Upload a single typical on-site document, compare the chunking results of direct upload and API upload, and confirm the chunking rules are consistent.
- Test parsing a large network topology document, and confirm the parsing time does not exceed the preset
PARSE_FILE_TIMEOUT_SECONDSthreshold. - Check the parsing logs, and confirm that documents from custom URLs can be pulled normally and chunking results are generated.
- Verify the uploaded file size, and confirm it does not exceed the
UPLOAD_FILE_MAX_SIZElimit.
The values provided are common starting points, and should be evaluated 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 2026-10-19.