What the data for this category looks like
Data sources for urban public transit service inquiries include official operation manuals, route adjustment announcements, fare policy documents, customer service response standard script libraries, and station service description documents. Update frequency varies: route adjustment documents are updated in real time alongside route optimizations, fare policy documents are updated quarterly, and customer service script libraries are iterated quarterly. Most documents are structured entries, containing fields such as route number, starting and ending stations, first and last departure times (format: HH:MM), fare (unit: yuan), list of passing stations, and accessibility facility descriptions. Service guide documents include process steps and application material lists.
What constraints these characteristics impose on the knowledge base retrieval and recall link
The structured fields and real-time update characteristics of this category create multiple constraints for retrieval and recall. Structured fields like route numbers and station names require precise matching in retrieval to avoid mixing information from different routes and stations. Real-time updated route adjustment documents require the recall link to automatically filter expired documents and prioritize the latest version content. Station names may have duplicate names across regions, so the recall scope must be narrowed using context such as the route and area included in the user's question. The standardized content of the customer service response script library requires that when retrieving and matching similar questions, the standard response document that most closely matches the semantics of the user's question is prioritized.
How to set the configurations
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
similarity | 0.75–0.85 | Most urban public transit inquiry questions are short texts. This range filters irrelevant results while retaining sufficiently matched route and station information |
limit | Top 10 entries | In service inquiry scenarios, users typically need 1 to 3 core pieces of information. 10 entries covers supplementary explanation needs and avoids excessive results that interfere with reading |
searchMode | semantic | Most public transit inquiry questions are semantic-based, such as "last bus time". Semantic search matches documents with similar phrasing |
usingReRank | Enabled | There are many similar phrasing for public transit station and route information. Reranking improves the sorting priority of high-match results |
datasetSearchUsingExtensionQuery | Enabled | Automatically expands user questions. For example, "bus card replacement" is expanded to "bus card loss replacement process" to match more comprehensive service guide documents |
collectionIds | Fill in according to the dedicated collection for public transit service inquiries | Avoid mixing in knowledge base documents from other categories, ensuring recall results only target content related to urban public transit |
Reference only; verify against the target FastGPT release and deployment documentation.
Three common mistakes
- Phenomenon: When configuring retrieval parameters, no optional values appear in the reference variable dropdown menu after selecting the target knowledge base. Cause: Urban public transit business documents have not been uploaded to the specified dataset collection, or uploaded documents do not generate recognizable structured variable fields.
- Phenomenon: After calling the retrieval interface, the returned results contain a large number of non-public transit irrelevant documents. Cause: The
collectionIdsparameter is not configured correctly, causing the recall scope to cover knowledge base content from other categories. - Phenomenon: The number of retrieval returned results is much lower than expected, or the route information requested by the user cannot be matched. Cause: The
similarityparameter is set too high, filtering out documents that meet the matching requirements but have similar semantics, ordatasetSearchUsingExtensionQueryis not enabled, so questions cannot be expanded to match relevant content.
How to confirm the configuration is complete
- Upload a test public transit route document, and check if fields such as route number and station name appear in the reference variable dropdown menu on the retrieval configuration page.
- Submit a test inquiry such as "last bus time for Route 10", and confirm that the returned results only include content related to urban public transit with no content from other categories.
- Adjust the
similarityparameter to 0.8, submit a test inquiry, and check if the matching degree of the returned results meets expectations with no excessive irrelevant content. - Call the document insertion API to upload a public transit service guide document, and confirm that the document is correctly synchronized to the specified dataset collection.
The values provided here 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.