What this category of data looks like
Coal transportation and sales policy and regulatory basis data mainly comes from official documents released by the National Energy Administration, provincial energy authorities, and coal industry associations, as well as industry mandatory standard documents. Data updates are released irregularly in line with adjustments to national energy policies, and are synchronized when new policies are issued or standards are revised. Most documents use standard official formats, including structured fields such as document number, issuing authority, release date, effective date, clause number, and specific content. Clauses are organized by article, paragraph, and sub-paragraph, and dates uniformly use the YYYY-MM-DD format.
What constraints these characteristics impose on the citation sources and traceability process
The clear document numbers, clause numbers and other structured fields in coal transportation and sales policy data require that the traceability process must accurately match a specific clause of an individual document, rather than only returning vague document names. The standard official document format and fixed field structure require that configurations retain metadata to ensure traceability information is verifiable. The irregular update frequency of data requires that the knowledge base synchronization mechanism adapt to non-fixed cycle update rhythms, to avoid referencing outdated policies. The long paragraph content of clauses requires that the chunking process retain contextual connections, to prevent loss of logical links between clauses after splitting, which affects traceability accuracy.
How to set the configurations
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
quoteMaxToken | 800–1200 token | Matches the average length of a single clause in coal policies, to avoid truncating key traceability information |
responseData | Includes document number, release date, clause number | Matches the structured fields of policy documents, to ensure complete and traceable traceability information |
detail | Enabled | Retains contextual connections between clauses, to avoid loss of clause logic after chunking |
| Single search limit | Top 6 results | Covers multi-policy citation needs in a single scenario, to avoid missing relevant clauses |
| Chunk overlap length | 50–80 characters | Retains connecting content between adjacent clauses, to ensure that the complete clause system can still be linked after chunking |
Three common errors
- Phenomenon: The citation source returned by knowledge base Q&A only shows 1 entry, and the workspace prompts "Knowledge Base Citation (1 entry)". Cause: The single search limit is configured too low, failing to cover the actual demand for multi-policy citations.
- Phenomenon: Citation sources do not display metadata such as policy document numbers or release dates, only returning vague document names. Cause: The
responseDataconfiguration does not specify extracting document metadata fields, and metadata synchronization is not enabled. - Phenomenon: Chunked content cannot be linked to the original clause number, and traceability results only show paragraph content without specific clause identifiers. Cause: Clause number identifiers are not retained during chunking, or the
detailconfiguration is not enabled to retain context.
How to confirm the configuration is properly set
- Upload a single coal transportation and sales policy document, launch a Q&A targeting a specific clause, and check whether the returned results include complete metadata such as document number and release date.
- Launch a Q&A that requires referencing multiple policies, and check whether the number of citation sources in the returned results matches the configured single search limit.
- View the knowledge base chunk preview, confirm that each chunk retains identifiers such as clause number and issuing authority, and no key traceability information is lost.
- Access the system log in FastGPT v4.17.0, confirm that configuration parameters such as
quoteMaxToken,responseData, anddetailhave been correctly loaded and take effect.
The values listed above are common starting points and should be evaluated 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.