What this category of data looks like
Data for shipbuilding declaration material and file review primarily comes from ship design drawings, construction inspection reports, classification society certification documents, project approval files, and similar sources. Data updates are triggered per individual construction project or vessel type batch. A single update includes mixed-format documents. Documents contain fixed fields such as vessel type number, construction batch, inspection item number, and approval status. Units include engineering professional units such as meters, tons, and hours. Some long documents include nested drawing annotations and form attachments.
What constraints these characteristics impose on knowledge base retrieval and recall
The mixed-format document structure requires the retrieval system to support cross-format parsing and unified indexing. This prevents missing audit materials such as drawing notes and approval forms. Professional fields like vessel type number and inspection item number require precise matching. This avoids generic searches returning irrelevant files. The batch-based update rhythm requires the knowledge base to support incremental synchronization. This avoids redundant overhead from full retrieval. Nested content in long documents needs proper chunking. This ensures core audit information is not truncated.
How to configure these settings
Reference only; verify against the target FastGPT release and deployment documentation.
| Configuration Item | Recommended Value Range | Rationale |
|---|---|---|
similarity | 0.75–0.85 | Meets the precision requirements for matching shipbuilding terminology, filters irrelevant files with low similarity scores |
limit | Top 6–10 results | The number of files for a single batch of shipbuilding declaration materials typically ranges from 5 to 10, covering core nodes such as construction inspection and classification society review |
searchMode | Full-text retrieval | Supports cross-field matching for multiple document types including drawing notes and approval forms |
usingReRank | Enabled | Performs secondary sorting on search results for shipbuilding terminology, improves the recall priority of core audit items |
datasetSearchUsingExtensionQuery | Enabled | Automatically expands the search scope for professional fields such as vessel type number and inspection item number, covers hidden associated fields |
collectionIds | Specify the collection ID of the corresponding shipbuilding knowledge base | Limits the retrieval scope, prevents mixing irrelevant documents from other categories |
Three common configuration mistakes
- A parameter validation error pop-up appears during retrieval, indicating the
similarityvalue is outside the system's validation range. This occurs when the threshold is set above 0.9, which does not align with the precision requirements for professional matching in the shipbuilding scenario. - Retrieval returns only 2 results, which is far fewer than the expected number of audit files. This occurs when
limitis set to 3 anddatasetSearchUsingExtensionQueryis not enabled, so hidden fields in declaration materials cannot be matched. - Retrieval results include a large number of general office documents with no connection to shipbuilding professional content. This occurs when
collectionIdsis not specified, so the retrieval scope includes non-shipbuilding knowledge base data from other categories.
How to verify configurations are correct
- Upload a single shipbuilding construction inspection report, run a keyword search that includes the vessel type number, and confirm the returned results include the core audit items of this report.
- Adjust the
similarityvalue, compare the number of retrieval results across different thresholds, and confirm the results meet the precision requirements for business audits. - Check the system retrieval logs, and confirm that
usingReRankanddatasetSearchUsingExtensionQueryhave been triggered normally. - Switch search keywords using different vessel type numbers, and confirm the returned results only include declaration files for the corresponding vessel type.
The values provided here 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.