What data for this category looks like
Core data for welded structural component manufacturing comes from design drawings, material quality certificates, welding equipment parameters, filler metal inventory, non-destructive testing systems, and personnel qualification ledgers. Daily documents include welding procedure qualifications, welding process specifications, filler metal usage records, welder qualification records, weld inspection reports, and repair orders. Fields cover weld numbers, base material grades, filler metal models, welder qualification levels, inspection parameters, and more. Data updates follow the pace of individual welds, batches, or shifts. Single-batch production data is added or modified in real time as production processes advance.
Constraints imposed by these characteristics on knowledge base retrieval and recall
The data characteristics of this category create multiple constraints for knowledge base retrieval and recall. High-frequency real-time updated data requires the retrieval link to support incremental synchronization. This avoids retrieving expired or unupdated process and inspection documents. Documents with multiple types and rigid verification fields need differentiated recall rules configured by document attribute. This ensures key content such as welder qualifications and procedure qualifications is retrieved first, preventing verification errors. Heterogeneous multi-source data sources require retrieval to support cross-system associated recall. This covers parameter content from dispersed data sources such as filler metal inventory and non-destructive testing systems. The number of associated documents linked to a single weld is limited. The upper limit of recall results must be controlled to avoid exceeding the context processing range.
How to set configurations
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
similarity | 0.75–0.85 | Matches the precision requirements for welding process and qualification verification. The interval covers the effective recall threshold for core documents, preventing accidental retrieval of mismatched base material or filler metal documents |
limit | Top 10 entries | Covers multiple types of documents related to a single weld, including processes, filler metals, qualifications, and inspections. Prevents excessive results from exceeding context processing limits |
searchMode | mixedRecall | Balances precise matching of welding professional terminology and fuzzy semantic understanding, adapting to retrieval needs for process documents and avoiding retrieval failures caused by terminology differences |
usingReRank | Enabled | Reorders recall results by relevance, prioritizing key documents for rigid verification such as welder qualifications and procedure qualifications |
datasetSearchUsingExtensionQuery | Enabled | Expands the retrieval scope to cover documents from external associated data sources such as filler metal inventory and non-destructive testing systems |
collectionIds | Configured grouped by document type | Distinguishes between process, qualification, and inspection documents, filtering irrelevant data to improve retrieval accuracy |
Three Common Misconfigurations
- Phenomenon: Custom-split welded document chunks stored in the knowledge base have duplicate content automatically removed, resulting in disordered custom index order. Cause: Automatic deduplication configuration for the knowledge base was not disabled. The system uses text hashing for deduplication by default, which disrupts the custom chunk order.
- Phenomenon: Retrieving welding process documents from external websites returns empty results, with no target content obtained. Cause: Access adaptation rules for external data sources were not configured, and the extended retrieval function of
datasetSearchUsingExtensionQuerywas not enabled. - Phenomenon: After a source-deployed FastGPT connects to a Docker-deployed database, uploaded welded structural component documents do not sync to the knowledge base, and an empty list is displayed. Cause: The service was not restarted to load new database connection configurations, or database incremental sync trigger rules were not configured.
How to Confirm Configurations Are Set Correctly
- Upload one welding process specification document, run a query that includes "welding sequence for a specific weld model", and verify the returned results include the corresponding process content.
- View the document chunk list in the knowledge base, confirm that custom-split chunks have not been automatically deduplicated, and retain the initial index order.
- Configure
collectionIdsto the welder qualification document group, run a query for "welder qualification verification", and confirm only welder qualification type documents are returned. - Confirm that FastGPT v4.17.0 is compatible with the current configuration items, and check that retrieval logs have no parameter errors.
The values given are common starting points and should be measured against the reader's own samples.