What this category of data looks like
Coal transportation and sales policy and regulatory basis data mainly comes from official industry policies released by the National Energy Administration, provincial and municipal energy regulatory departments, and the China Coal Industry Association, as well as internal enterprise transportation and sales management systems, contract templates, and compliance documents. Update rhythm is dynamically adjusted with policy releases. Synchronization is completed within 1 to 3 working days after a new policy is issued. Internal documents are updated quarterly. Document structure includes policy document number, issuing authority, effective date, and clause text. Some files have attachments such as traffic volume calculation forms and tax and fee standard charts. Fields cover applicable regions, implementation period, tax and fee rates (unit: yuan/ton), transportation and sales process node requirements, and other content.
What constraints these characteristics impose on the knowledge base retrieval and recall link
Official sources of policy data require precise matching of document numbers and applicable scenarios during retrieval, to avoid non-industry policies being included due to generalized searches. Dynamic update rhythm requires configuring incremental update trigger rules to avoid resource waste from full updates. Chart attachments in documents cannot be directly recognized by semantic retrieval. OCR must be used to extract text from charts to fully cover the retrieval scope. Numeric values with units such as tax and fee rates and transportation volumes require unit matching logic during retrieval to avoid ambiguity. Effective and expiration dates of policies require automatic filtering of expired documents during retrieval to ensure the timeliness of returned results.
How to set the configurations
Reference only; verify against the target FastGPT release and deployment documentation.
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
similarity | 0.75–0.85 | Matches the precision requirement for coal transportation and sales policies, prevents low-relevance non-industry policies from appearing in results |
limit | Top 8 entries | Covers multi-dimensional associations of policy clauses, while keeping the context length within the limits of large model input |
searchMode | hybrid | Combines keyword and semantic retrieval, adapts to mixed retrieval needs for policy document numbers and clause content |
usingReRank | Enabled | Re-ranks initial recall results, corrects local deviations from semantic retrieval, matches the hierarchical structure of policies |
datasetSearchUsingExtensionQuery | Enabled | Automatically expands synonyms related to policies. For example, "coal transportation and sales" is expanded to "coal purchasing and sales" and "coal transportation and dispatch" to cover differences in policy wording |
collectionIds | Specify exclusive collection IDs for coal policies | Isolates knowledge base data from other categories, prevents mixed retrieval results |
Three common errors
- Phenomenon: Retrieval results do not include key data such as tax and fee rates and traffic volume calculation forms in documents. Cause: The image OCR text extraction configuration of the knowledge base is not enabled, and only plain text content is extracted.
- Phenomenon: Non-coal transportation and sales policy documents are mixed in retrieval results. Cause: Exclusive
collectionIdsare not configured, causing cross-category knowledge base data to be recalled at the same time. - Phenomenon: Knowledge base data cannot be accessed after the renewal grace period ends. Cause: The platform's renewal rules are not paid attention to, and the renewal operation is not completed within the grace period.
How to confirm the configuration is complete
- Upload a test document containing a policy document number and a tax and fee rate table, and check whether the retrieved results include the text content extracted from the chart.
- Enter a search query containing a policy document number, and confirm that the corresponding policy document is returned first in the retrieval results.
- View the knowledge base configuration page, and confirm that the exclusive
collectionIdshave been correctly bound to the current dataset. - Enter a search query involving expired policies, and confirm that the results are not recalled.
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.