What data for this category includes
This category of data originates from central state-owned enterprise procurement platforms, public resource trading platforms, supplier databases, contract management systems, and audit issue databases. Updates occur daily. Processed documents include procurement plans, requirement specifications, bidding documents, bid evaluation records, contracts, acceptance receipts, and supplier complaint materials. Document fields cover procurement methods, budget basis, scoring methods, supplier qualification requirements, performance clauses, and more. Document formats vary widely. Some documents include tables, attachments, and structured data.
What constraints do these characteristics impose on document parsing and chunking
The characteristics of this category of data impose multiple constraints on the document parsing and chunking process. Mixed-format data from multiple sources includes structured tables, plain text, and non-standardized bid evaluation and complaint materials. Parsing rules must adapt to different document types. Daily updated high-frequency incremental data demands sufficient processing efficiency from parsing workflows. Documents contain core compliance fields such as procurement methods, budget basis, and scoring methods. Chunking processes must retain contextual connections to avoid breaking critical compliance logic. Some documents include cross-page clause associations. Chunking processes must not interrupt complete review basis chains.
Reference only; verify against the target FastGPT release and deployment documentation.
How to set configurations
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
chunkSize | 800–1200 characters | Tendering and procurement documents are mostly long texts containing compliance clauses and structured data. This range preserves the integrity of a single compliance review logic. |
chunkSplitter | Split by paragraph + custom rules | Tendering and procurement documents include fixed-format clauses and non-standardized bid evaluation records. Combining paragraph splitting and custom rules adapts to different document structures. |
customPdfParse | Enabled | Bidding documents, bid evaluation records, and similar materials are mostly in PDF format. Enabling custom parsing preserves associations between tables and cross-page clauses. |
PARSE_FILE_TIMEOUT_SECONDS | 600 seconds | Large bidding documents and batch-uploaded supplier materials may take a long time to process. This duration covers conventional processing requirements. |
UPLOAD_FILE_MAX_SIZE | 1000 MB | Some bidding project attachment packages contain multiple documents. This upper limit meets batch upload requirements. |
chunkSettingMode | Manual configuration | Different types of tendering and procurement documents require differentiated chunking parameters. Manual configuration allows flexible adjustments. |
Three common misconfigurations
- The symptom is that after uploading a bidding document or bid evaluation record containing images, no image-related content appears in knowledge base search results. Large language model outputs do not reference image text. The cause is that image text extraction configuration is not enabled. Only plain text content in the document is extracted.
- The symptom is that after uploading a bidding document and performing a search test, the interface shows parsing succeeded but no matching entries are found, or an error prompt is returned directly. The cause is that contextual connections of core compliance fields such as procurement methods and budget basis are not retained during chunking, or chunk length settings are improper, leading to critical clauses being split.
- The symptom is that after uploading a batch of bidding attachment packages, some documents fail to parse. The cause is that
UPLOAD_FILE_MAX_SIZEis set too low, exceeding the allowed upper limit for a single uploaded file.
How to confirm configurations are properly set
- Upload a standard bidding document, review the parsed chunk content, and confirm that core compliance fields such as procurement methods and budget basis are not split.
- Perform a knowledge base search test, input core keywords such as procurement methods and scoring methods, and confirm that corresponding chunk content is matched.
- Upload a single large PDF bidding document, review the processing time of the parsing task, and confirm that the
PARSE_FILE_TIMEOUT_SECONDSsetting covers the processing duration. - Upload a Word-format bid evaluation record containing tables, and confirm that the parsed chunks retain the complete structure and content of the tables.
The values provided are common starting points and should be evaluated against relevant 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.