What this category of data consists of
Policy and regulatory basis data for the shipbuilding sector is sourced primarily from official red-letter documents, industry standards, and regulatory guidelines published by the Equipment Industry Division 1 of the Ministry of Industry and Information Technology, the Maritime Safety Administration of the Ministry of Transport, and the China Shipbuilding Industry Association. Data updates trigger when new policies are issued or standards are revised, with no fixed update cycle. Most documents are structured official papers, containing fields such as document number, effective date, applicable scenarios, and clause details. Some standard documents include technical parameter tables, using shipbuilding-specific units including millimeters, tons, and kilowatts.
What constraints these characteristics impose on knowledge base retrieval and recall
Shipbuilding policy data characteristics impose multiple constraints on retrieval and recall. The multi-field structured nature of official papers requires retrieval to match both content and metadata such as document number and effective date, to avoid retrieving expired or non-applicable policies. Technical parameter tables included in standard documents require that table structures are preserved during text segmentation, to prevent loss of specialized metering parameters. For data sources with no fixed update cycle, incremental sync logic must be configured to ensure retrieved content matches the latest policies. Dense specialized terminology in texts requires adapting the searchMode exact match logic, while enabling usingReRank secondary reranking and datasetSearchUsingExtensionQuery term expansion to improve retrieval accuracy.
How to configure the settings
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
| Similarity Threshold | 0.75–0.85 | Shipbuilding policy texts have high professionality, requiring filtering of low-match results. This range balances recall coverage and accuracy |
| Single Search Limit | Top 8 entries | Most policy documents are long texts. Retrieving 8 entries at a time covers core clauses and associated supplementary documents |
| Search Mode | Exact Match | Policy retrieval requires strict matching of clause content and metadata. Exact match mode avoids irrelevant retrievals |
| Reranking Switch | Enabled | Texts with many specialized terms require secondary reranking to raise the ranking priority of high-match results |
| Query Optimization Switch | Enabled | Can expand industry-specific terms based on user queries. For example, expand "shipbuilding specifications" to "China Shipbuilding Quality Standards" |
| Knowledge Base Collection ID | Fill in according to the actually bound shipbuilding policy dataset | Ensures only policy data from the specified domain is retrieved, preventing unrelated documents from being included |
Three common configuration mistakes
- Fails to load a self-built Qdrant vector database as a knowledge base. The cause is failure to configure custom vector database connection parameters in FastGPT, and only the platform's default vector storage service is used.
- Receives an authorization failed prompt when calling the API. The cause is failure to bind the correct knowledge base permissions in the application configuration, or the API key has expired.
- Loses the table of contents structure of Word documents after segmentation. The cause is failure to enable the document structure-based segmentation setting, and instead splitting directly by fixed character length, leading to separation of table of contents and content.
How to confirm the configuration is properly set
- Navigate to the knowledge base management page of FastGPT v4.17.0, check if the bound dataset includes shipbuilding policy-related documents, and confirm that the collection ID matches the configured settings.
- Submit a test query, such as "CCS certification requirements for shipbuilding", and check if the retrieved results include matching policy clauses and have expected sorting.
- Check the API call logs to confirm that the API key carried in the request is valid, and no permission-related error messages appear.
- View the segmented document preview to confirm that long documents retain core clauses and metadata, with no content breaks or structural loss.
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.