What the data for this category looks like
Coal transportation and sales business data comes from multiple sources. These include sales contracts, weigh tickets, coal quality test reports, and transportation plans. Data is updated for each batch of transportation and sales tasks. Each batch links to multiple document types. Sales contracts are multi-page structured text. They contain batch numbers, transaction party information, and transaction terms. Weigh tickets are single-page forms. They include vehicle information and weighing data. Coal quality test reports are fixed-format documents. They contain batch-associated numbers and test parameters. Common fields include batch number, business number, quantity, and test parameters. Units include tons, yuan per ton, and similar units.
What constraints do these characteristics impose on the source citation and traceability link
Coal transportation and sales business data is managed by batch. Different documents must use the same batch number to complete traceability. The citation link must support aggregating and retrieving content by batch. A single batch links to many documents. Some documents have short content, others have longer length. Unified chunking and matching rules are needed to ensure retrieval accuracy. Business fields contain professional parameters. These parameters must match specific business scenarios to generate accurate document content. Data updates frequently. Real-time synchronization of the latest documents to the knowledge base is required. This prevents expired data from being used in generated content.
How to Configure
Reference only; verify against the target FastGPT release and deployment documentation.
| Configuration Item | Recommended Setting | Rationale for This Setting |
|---|---|---|
| Maximum Reference Token Count | 8000–12000 token | A single batch of coal transportation and sales is associated with many documents, so full business information must be covered to avoid context truncation |
| Return Data Format | Only return document numbers and core business fields | Document drafting only requires traceability basis, no full redundant content |
| Chunk Aggregation Rule | Aggregate by transportation and sales batch before chunking | Coal transportation and sales is managed by batch, and aggregation enables accurate association with corresponding business scenarios |
| Citation Retrieval Limit | Top 20 matching documents | The volume of documents for a single batch of transportation and sales is controllable; excessive retrieval increases context load |
| Matching Threshold | Calibrate based on actual testing | Matching requirements vary across business scenarios, and adjustments must be made based on actual business logs |
| Traceability Display Switch | Enable batch association display | Facilitate subsequent verification of business links and comply with compliance requirements for coal transportation and sales |
Three Common Misconfigurations
- When the maximum reference token count is set above 12000 token, the interface shows context failed to load, and the large language model does not return knowledge base citation content. Cause: FastGPT has a limit on the total token count of single-round context. When the threshold is exceeded, the knowledge base citation link is automatically skipped.
- After uploading raw data returned by HTTP interfaces as a knowledge base source, the citation function returns no matching results. Cause: Data was not converted to the title + body structured format required by FastGPT knowledge base, so it cannot be indexed correctly.
- After configuring knowledge base citation, the generated document does not include professional business content from the knowledge base. Cause: The matching threshold was not set correctly, leading to retrieved content that is irrelevant to generation requirements, or retrieved content being truncated.
How to Verify Proper Configuration
- Upload a single coal transportation and sales contract and corresponding weigh ticket, trigger the document drafting process, and check whether the generated content marks the corresponding document numbers.
- View the knowledge base citation logs to confirm that the number of retrieved documents matches the set citation retrieval limit.
- Adjust the maximum reference token count, observe whether the context loading status is normal, with no truncation or skip prompts.
- Test generating documents containing professional business parameters, and confirm that the retrieved content includes the corresponding business fields.
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.