Reference only; verify against the target FastGPT release and deployment documentation.
What this category's data looks like
Core data for this category comes from the road transport administration system, the national road transport vehicle dynamic information public service platform, and the vehicle and practitioner database. The update frequency is daily. Daily processed documents include business license applications, vehicle technical files, driver qualification materials, safety production management systems, and road transport business licenses. Document fields include applicant information, vehicle license plate numbers, driver qualification certificate numbers, safety production management system clauses, and license validity periods. Typical users include transportation bureau approval departments, road transport management agencies, and road transport enterprises.
What constraints do these characteristics impose on the knowledge base retrieval and recall link
Daily updated multi-source data requires the retrieval system to support incremental synchronization and incremental recall, to avoid reprocessing historical data. Multi-type document structures require retrieval configurations to balance retrieval accuracy for structured qualification fields and unstructured management texts. High requirements for verification accuracy in approval processes mean retrieval results must filter irrelevant content, and prioritize core materials related to vehicles, personnel, and safety systems. Differences in document lengths require chunking configurations to adapt to varying text lengths, avoiding over-truncation of long texts or loss of key information in short texts.
How to set the configurations
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
similarity | 0.75–0.85 | Must accurately recall content related to vehicle and driver qualifications, and avoid irrelevant documents interfering with approval workflows |
limit | Top 8–12 results | Approval requires simultaneous verification of multiple types of materials: vehicles, personnel, safety systems. An appropriate number of results covers core information |
searchMode | vector+keyword | Business documents include both structured qualification fields and unstructured management texts. Hybrid retrieval balances accuracy and recall rate |
usingReRank | Enabled | Multi-source retrieval results must be sorted by relevance to ensure approval personnel first access the most relevant license applications and vehicle file materials |
datasetSearchUsingExtensionQuery | Enabled | Approval personnel’s queries may omit full names of qualification fields. Extended queries can cover synonyms or abbreviations |
collectionIds | Specify the knowledge base collection ID corresponding to road transport business licensing | Isolate documents from other categories, and avoid mixing irrelevant traffic law enforcement or passenger transport management materials in retrieval results |
Three common mistakes
- Phenomenon: Garbled characters appear after importing Chinese road transport business license application materials into the knowledge base, while English materials work normally. Cause: Document encoding was not converted to UTF-8. Documents exported from older versions of road transport administration systems use GBK encoding.
- Phenomenon: Searching for "vehicle technical files" returns insufficient results, failing to cover all required file content. Cause: The
similaritythreshold is set too high, filtering out some relevant archive texts with slightly lower similarity. - Phenomenon: FastGPT knowledge base disk usage continues to grow, with no clear source of usage found. Cause: No expiration cleanup rules are configured for split text chunks and embedding vectors. Daily new data uses full synchronization instead of incremental synchronization, leading to repeated storage of historical documents.
How to confirm the configuration is correct
- Upload a test driver qualification material, search for "driver qualification", and confirm that the retrieval results include the core field content of this material.
- View the FastGPT disk usage statistics interface, and confirm that the statistics cover three parts: original files, split text chunks, and embedding vectors.
- Trigger an incremental synchronization task, and confirm that only newly added vehicle or personnel data from the current day is updated, with no historical documents reprocessed.
- Compare the sorting of retrieval results before and after enabling
usingReRank, and confirm that highly relevant core materials appear at the top of the results.
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.