What the Data for This Category Looks Like
Data for this category comes primarily from full-process documents of coal-electric integration projects, including coal mining scheduling records, thermal power unit operation logs, safety production regulations, energy consumption accounting reports, contract templates, and compliance review documents. There are two update cycles for the data: production scheduling documents are updated in real time alongside operations, while compliance and template documents are revised and released on a quarterly or annual basis. Most documents are structured long texts, with fields such as operation number, unit number, coal type indicators, and power generation load. Units include tons, megawatts, hours, and others.
What Constraints Do These Characteristics Impose on the Knowledge Base Retrieval and Recall Link
A high proportion of full-process long texts and specialized fields require retrieval to match precise fields such as operation number and unit number, to avoid retrieving irrelevant cross-scenario documents. Real-time updated production scheduling data requires the knowledge base to support incremental synchronization. Without this support, retrieval results will lag behind on-site operations. Many specialized terms with exclusive units require prevention of misjudgment by general semantic models, and adaptation to the chunking logic of long documents. This ensures retrieved content covers the context of complete operation links. Fixed-format contracts and contract templates need to support filtering retrieval results by document type, to improve reference accuracy during document drafting.
How to Configure
Reference only; verify against the target FastGPT release and deployment documentation.
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
similarity | 0.75–0.85 | Semantic similarity threshold for matching coal-electric domain specialized terms, prevents retrieving irrelevant documents if set too low, and avoids missing precise specialized content if set too high |
limit | Top 10 entries | Covers the typical chunk count of a single coal-electric document, ensures sufficient compliance and operation reference content is available during document drafting |
searchMode | semantic | Adapts to semantic retrieval requirements for specialized terms, matches the specialized expression habits of the coal-electric domain |
usingReRank | Enabled | Performs secondary ranking on retrieved results to filter irrelevant content across units or operation links |
datasetSearchUsingExtensionQuery | Enabled | Automatically supplements coal-electric domain specialized keywords based on user-provided document drafting requirements, improving retrieval accuracy |
collectionIds | Fill in according to the exclusive coal-electric knowledge base ID | Limits the retrieval scope to only full-process documents for coal mining and power generation, avoiding mixing irrelevant data from other categories |
Three Common Misconfigurations
- Phenomenon: Knowledge base documents contain keywords such as coal type parameters and unit number, but no results are returned after semantic retrieval. Cause:
datasetSearchUsingExtensionQueryis not enabled, and coal-electric domain specialized keywords are not supplemented. This causes the general semantic model to fail to match the semantic association of specialized terms. - Phenomenon: When using a strict question-answering template, documents containing image addresses are retrieved, and the system prompts that no answer is found. Cause: The strict question-answering template only supports direct extraction of text content, and cannot parse non-text information corresponding to image addresses. Switching the template will return relevant content normally.
- Phenomenon: After passing knowledge base-related parameters during API calls, the system still prompts that the knowledge base is not configured. Cause: The format of the
collectionIdsparameter is not specified correctly, or the global variable is not bound to the current retrieval link. This causes the system to fail to recognize the specified knowledge base scope.
How to Confirm the Configuration Is Correct
- Submit a retrieval request for coal-electric specialized terms, and verify that retrieved results include document content corresponding to the target operation scenario.
- Check the similarity score of retrieval results, confirm that the score falls within the preset
similarityrange. - Check the
collectionIdsfield in API return results, confirm that only IDs of exclusive coal-electric knowledge bases are included. - Test retrieval of documents containing image addresses, confirm that return results meet expectations under different templates.
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.