What the Data for This Category Looks Like
Coal transportation and sales declaration materials and file review data mainly come from business documents. These include monthly or quarterly capacity declaration forms, weighing settlement slips, coal quality test reports, transportation and sales contracts, and compliance review files. The data update rhythm has two modes: regular archive updates and temporary incremental synchronization. The system enters materials from a single declaration cycle after review is completed. The document structure is fixed. It includes business fields such as coal type, shipment volume (unit: ton), calorific value (unit: large calorie/kilogram), departure location, destination, carrier qualification number, and review status. File documents add metadata fields including archive number, reviewer, and review time.
What Constraints These Characteristics Impose on the Knowledge Base Retrieval and Recall Link
Professional fields and fixed units require matching the relationship between numerical values and units during retrieval. This avoids incorrect recall caused by only matching numerical values. For example, do not mistakenly match "5000 large calories/kilogram" to "5000 tons". The fixed document structure requires retaining field associations when chunking. Do not disrupt the correspondence between metadata and business content. Otherwise, review logic will break. The combined update rhythm of regular and incremental updates requires the knowledge base to support incremental updates and regular full verification. This avoids missing the latest review status changes. The associated attributes of declaration materials and files require retrieving all associated documents under the same declaration batch during retrieval. This ensures all materials required for review are fully covered.
How to Set the Configurations
| Configuration Item | Recommended Value | Basis for This Setting |
|---|---|---|
similarity | 0.75-0.85 | Coal transportation and sales business has many professional terms and fixed units. A higher similarity threshold filters irrelevant results while covering professional semantic matching |
limit | Top 8-12 entries | The number of associated documents for a single declaration batch is usually under 10. Limiting returned entries avoids exceeding context length while covering all necessary review materials |
searchMode | keyword+vector hybrid mode | Retrieval must match both professional keywords (such as "calorific value", "shipment volume") and semantic review logic. The hybrid mode balances precise matching and semantic recall |
usingReRank | Enabled | Coal transportation and sales documents contain many numerical values and associated fields. Reranking optimizes result sorting and improves the display priority of highly relevant documents |
datasetSearchUsingExtensionQuery | Enabled | Declaration review questions often have implicit conditions, such as "meeting the national standard for thermal coal". Extended queries cover these implicit needs and improve recall coverage |
collectionIds | Bind the exclusive dataset ID for coal transportation and sales declaration files | This avoids recalling documents from unrelated categories, and ensures the retrieval scope only covers current business declaration and file data |
Three Common Mistakes
- The symptom is empty retrieval results. The cause is failure to correctly bind the exclusive coal transportation and sales dataset
collectionIds. A general business dataset is mistakenly called, so the retrieval scope does not include target declaration and file documents. - The symptom is failure to import Notion-format declaration material documents. The cause is failure to enable Notion link import adaptation in the data source configuration. Only general web links are supported, so the system cannot parse Notion's exclusive document block structure.
- The symptom is failure to select the locally deployed ChatGLM2 model. The cause is failure to correctly associate the OneAPI model access path in the FastGPT configuration. Local model adaptation is not enabled in the dataset configuration, so the model list fails to load.
How to Confirm the Configuration Is Complete
Reference only; verify against the target FastGPT release and deployment documentation.
- Check the dataset management interface. Confirm the bound
collectionIdsis the exclusive ID for coal transportation and sales declaration files, with no other unrelated datasets associated. - Test retrieval of professional business terms, such as "Q3 thermal coal shipment volume". Confirm returned results include documents with corresponding coal type and shipment volume fields.
- View the retrieval log. Confirm
usingReRankanddatasetSearchUsingExtensionQueryhave taken effect, with no error returns. - Adjust the
similaritythreshold and compare retrieval results. Confirm the number of results fluctuates as expected with changes to the threshold.
The values given 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.