What the Data for This Category Looks Like
Reference only; verify against the target FastGPT release and deployment documentation.
Constraints Imposed by These Characteristics on Knowledge Base Retrieval and Recall
Gas industry structured inspection data includes fixed fields and standard units. Retrieval must prioritize matching field completeness and unit accuracy to avoid returning results with mismatched parameters. For unstructured long documents such as annual safety training materials, chunking must adapt to the semantic boundaries of professional terms to avoid term splitting. Real-time synced work order data requires the retrieval pipeline to support incremental updates, while expired work orders must be filtered. Gas industry professional terms have synonymous expressions, so extended queries must be used to cover these, otherwise recall accuracy will decrease.
How to Configure the Settings
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
similarity | 0.72–0.78 | Balances recall accuracy and coverage for gas industry professional terms, avoids overly filtering valid results |
limit | Top 8 entries | Covers the recall needs of three core business data types: inspection points, safety regulations, and work order templates, controls context length |
searchMode | hybrid | Combines keyword retrieval and semantic retrieval, adapts to mixed data scenarios of structured forms and unstructured documents |
usingReRank | Enabled | Performs secondary ranking on initial recall results, prioritizes content containing gas industry professional parameters and terms |
datasetSearchUsingExtensionQuery | Enabled | Adds gas industry synonym expansion, covers synonymous expressions such as "pressure regulating box" and "pressure regulator", "pipe network pressure drop" and "pipeline pressure decrease" |
collectionIds | Bound by business group | Only allows inspection knowledge base, safety regulation knowledge base, and work order knowledge base to participate in retrieval, avoids cross-library recall of irrelevant documents |
Three Common Configuration Mistakes
- A large number of non-gas industry equipment documents appear in retrieval results. The cause is that
collectionIdsis not configured to bind dedicated business knowledge bases, leading to cross-business library recall of irrelevant data. - Uploaded gas safety regulation files cannot be retrieved. The cause is that
datasetSearchUsingExtensionQueryis not enabled, so professional terms cannot be correctly expanded and matched. - The number of single retrieval results far exceeds expectations. The cause is that the
limitparameter is not set, leading to excessive recall of long documents causing context overflow.
How to Confirm the Configuration Is Complete
- Upload a structured form containing inspection point numbers and pressure values, retrieve "XX point pressure abnormality", and verify whether the recall results include the inspection specification content for that point.
- After enabling
usingReRank, compare the ranking of results for the retrieval query "gas pipeline network maintenance", and confirm that documents containing professional terms are ranked at the top. - Check the binding status of
collectionIds, and confirm that only the specified gas business knowledge bases are included in the retrieval scope, with no cross-library irrelevant results. - Test the retrieval query "latest gas safety regulations", and confirm that the update time of the recalled results matches the preset monthly or quarterly sync cycle.
The values given 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.