What this category of data looks like
Coal transportation and sales business data mainly comes from four types of core documents: transportation and sales contracts, weigh tickets, settlement sheets, and transportation plan sheets. The data sources are the enterprise's internal ERP system, on-site weighing equipment, and contract management module. Data updates are triggered in real time with business processing, and synchronization is completed within 1 hour after a single business is completed. The document structure includes two categories: structured table fields and unstructured text. Structured fields include departure station, arrival station, coal type, billing tonnage, freight rate standard, etc. Units include ton, yuan/ton, kilometer, train number, etc. Unstructured content is mostly contract clauses and business explanations.
What constraints do these characteristics impose on the knowledge base retrieval and recall link
Coal transportation and sales data has a high proportion of structured content and highly specialized fields. Retrieval is required to match both keywords and field attributes to avoid interference from generalized matching. Real-time updated business documents require the knowledge base to support incremental synchronization, preventing expired transportation and sales plans from being retrieved. Document lengths vary significantly: short tickets are only tens of characters, while long contracts can reach thousands of characters, so recall rules need to adapt to different lengths. Business terms have synonyms, such as "clean coal" and "washed clean coal", so term expansion support is needed to cover all business scenarios.
How to set the configurations
Reference only; verify against the target FastGPT release and deployment documentation.
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
similarity | 0.72–0.80 | Coal transportation and sales business has high requirements for keyword accuracy; this range can filter low-relevance non-business documents |
limit | Top 6–8 results | Business processing requires linking multiple related documents; too many results will increase the context processing load |
searchMode | Hybrid retrieval | Supports both structured field matching and unstructured text retrieval, covering both contract and ticket data |
usingReRank | Enabled | Business keywords for coal transportation and sales are scattered; re-ranking can improve the sorting priority of key results |
datasetSearchUsingExtensionQuery | Enabled | Covers matching requirements for business synonyms, such as "shipping volume" and "shipping tonnage" |
collectionIds | Bind three collections: transportation and sales contracts, weigh tickets, settlement sheets | Covers all core document types required for complete business processing |
Three common mistakes
- Phenomenon: Retrieval time exceeds 10 seconds, and the interface displays a loading timeout prompt. Cause: Retrieval rules are not optimized for structured data of coal transportation and sales, and documents from too many irrelevant collections are loaded at the same time, resulting in excessive load on vector retrieval and re-ranking links.
- Phenomenon: Transportation and sales documents of non-specified coal types appear in retrieval results. Cause: A reasonable
similaritythreshold is not set, orusingReRankre-ranking is not enabled, causing low-relevance results to appear first and interfere with business judgment. - Phenomenon: A single knowledge base cannot import more than the preset number of transportation and sales documents. Cause: The storage space upper limit configuration of the corresponding FastGPT collection is not adjusted, or the quota is not allocated according to the actual average size of business documents.
How to confirm the configuration is correct
- Upload a single coal transportation and sales contract and weigh ticket, initiate a query asking "What is the settlement process for XX coal type", and check whether the returned results include the key fields of the corresponding documents.
- View the knowledge base storage space usage statistics to confirm that the imported transportation and sales documents do not exceed the preset quota.
- Simulate high-frequency retrieval requests to check whether the retrieval time meets business expectations and there are no obvious timeouts.
- Test using business synonyms for queries, such as replacing "weigh ticket" with "weighing ticket", and confirm that the retrieval results can still match the correct documents.
The values provided here 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.