What This Category of Data Looks Like
Coal-electric integration policy and regulatory basis data comes primarily from official documents and industry standard files issued by the National Energy Administration, the State-owned Assets Supervision and Administration Commission of the State Council, and provincial and municipal energy authorities. It also includes policy interpretation documents released by industry associations.
Update cycles adjust dynamically with policy revisions. Regular policies are updated in quarterly batches. Temporary special policies are synchronized within 24 hours of release.
Each single document contains five core fields: policy document number, issuing authority, release date, applicable scope, main text clauses, and supporting interpretations. Field units include document number, year/month/day, and applicable regional scope. No unified fixed numeric units exist.
Constraints Imposed on Knowledge Base Retrieval and Recall
The formal nature of coal-electric integration policy documents requires retrieval results to accurately match policy document numbers and implementation timelines. Relying solely on main text keyword recall will lead to invalid cited policy basis.
Documents include multi-field metadata. Retrieval must match both metadata and main text content to accurately associate with the corresponding policy document.
Single documents may have lengthy main text that exceeds conventional file size limits, increasing upload and chunking difficulty.
The rapid update requirement for temporary policies means knowledge base retrieval configuration must support incremental synchronization. Delayed data will otherwise reduce Q&A accuracy.
When a large number of policy documents exist, unconstrained retrieval will generate redundant recall results, increasing token consumption and response delay.
Configuration Settings
This configuration scheme requires FastGPT v4.17.0 or higher.
| Configuration Item | Recommended Value Range | Rationale |
|---|---|---|
similarity | 0.75–0.85 | Coal-electric policy texts are highly formal. Strict matching of core policy expressions is needed to avoid low-relevance recall. |
limit | Top 3–5 results | Policy Q&A requires precise citations. Excessive results increase token consumption, aligning with the precise needs of public sector and state-owned enterprise policy documents. |
searchMode | semantic + metadata | Combine semantic and metadata retrieval. Prioritize matching policy numbers and applicable scope fields to ensure accuracy of policy basis. |
usingReRank | Enabled | Perform secondary sorting on recall results by policy timeliness and relevance. Prioritize returning newly released policy documents. |
datasetSearchUsingExtensionQuery | Enabled | Perform extended queries for policy terminology. Expand "coal-electric integration" to "coal-electric joint operation" and "coal-electric coordinated development" to cover synonymous expressions in policies. |
collectionIds | Bind dedicated coal-electric policy knowledge base collection | Avoid retrieving policy documents from other categories, ensuring precise recall scope. |
Three Common Configuration Mistakes
- Phenomenon: Only 1 cited result is returned per query in basic workspace knowledge base applications. Multiple matching documents cannot be recalled. Cause: The
limitparameter is not configured correctly, or the retrieval mode only matches single-document metadata without enabling multi-document recall logic. - Phenomenon: Q&A response times out, and token consumption exceeds thresholds. Cause: Chunking configuration is not adjusted according to the long text characteristics of coal-electric policy documents, or the
limitvalue is set too high, leading to recall of too many long text fragments.
Reference only; verify against the target FastGPT release and deployment documentation.
How to Verify Successful Configuration
- Enter the knowledge base retrieval test interface, input coal-electric policy-related questions, and check whether the number of cited results returned matches the configured
limitvalue. - View retrieval logs to confirm that recall results include both policy metadata (such as document number, issuing authority) and main text keyword matches.
- Upload a single coal-electric policy document exceeding 100 MB, and confirm that the system can perform chunked upload normally without triggering file size limit errors.
- Simulate incremental update of temporary policies by uploading newly released special policy documents, and confirm that the knowledge base can complete synchronization within 1 hour.
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.