Reference only; verify against the target FastGPT release and deployment documentation.
What the data for this category looks like
Civil aviation airport business data mainly comes from official operation manuals, terminal guidance documents, flight dynamic ledgers, special passenger service specifications, security check and check-in process announcements, and similar materials. Update cadence: flight plan data is updated weekly, process specification documents are updated monthly based on new industry policies or internal airport adjustments, and temporary notice data is synced in real time. Most documents follow a step-by-step entry structure, including fields such as flight number, terminal number, gate number, service duration, contact number, and others. Units include minutes, hours, terminal numbers, gate numbers, and similar units. Some documents include cross-departmental process coordination instructions.
Constraints imposed by these characteristics on the knowledge base retrieval and recall link
Airport business data is scattered across static process documents and dynamic real-time ledgers. The data source scope must be clearly distinguished during retrieval to avoid confusion between static processes and real-time flight data. Fields include precise identifiers such as flight number and terminal, so specific fields must be matched during retrieval to filter irrelevant results and improve content relevance. Real-time data with fast update frequency requires an incremental sync mechanism to ensure recalled content matches the airport's current operational status. Long-text process documents require reasonable chunking while retaining step continuity to avoid loss of complete process logic after splitting.
How to set the configurations
| Configuration Item | Recommended Value | Rationale |
|---|---|---|
similarity | 0.75–0.85 | Civil aviation airport process documents are mostly step-by-step text. A balance between retrieval precision and coverage must be maintained. Values below 0.75 may mix in irrelevant non-process documents, while values above 0.85 may miss similar special process descriptions. |
limit | Top 8 entries | Airport business processes cover multiple scenarios including check-in, security check, boarding, special services, and others. Recalling 8 entries at a time can cover complete process branches, avoiding context overflow caused by excessive results. |
searchMode | semantic_search | Process documents are mostly structured steps. Semantic search can accurately match process keywords in user questions, which performs better than the rigid matching of keyword search. |
usingReRank | Enabled | A large number of similar expressions exist in airport processes, such as "check-in closes 45 minutes in advance" and "check-in deadline is 45 minutes before flight departure". Reranking filters semantically duplicate recalled results and improves sorting rationality. |
datasetSearchUsingExtensionQuery | Enabled | Users often omit preconditions such as "gate". Extended queries supplement contextual fields including terminals and flight numbers to improve retrieval precision. |
collectionIds | Bind the exclusive knowledge base collection ID of the airport | Distinguish the airport's own process documents from general civil aviation knowledge, avoid recalling irrelevant industry-general content, and ensure results only come from the current airport's official data. |
Three common misconfigurations
- Empty results are returned when calling the knowledge base search node, and the specified file content cannot be obtained. The cause is failure to bind the corresponding airport knowledge base collection via
collectionIds, or failure to enable access permissions for the dataset. - The content format referenced by the knowledge base does not match expectations, or the model does not correctly use the recalled knowledge base content. The cause is confusion between the configuration location and scope of the reference content template and the reference template prompt word.
- After uploading a CSV-format airport process document, the text block still contains cross-line content. The cause is failure to use the CSV-specific line separator, or the custom separator does not match the actual separator in the file.
How to confirm the configuration is successfully applied
- Upload a test airport process document, configure the corresponding
collectionIds, then initiate a test query to confirm that the returned results include the specified content in the test document. - Adjust the
similarityparameter value, observe changes in the number and precision of recalled results to verify that the parameter configuration has taken effect. - Enable
usingReRank, then compare the recalled result sorting before and after enabling to confirm that the reranking function operates correctly. - Configure authentication rules, then initiate a query using an unauthorized account to confirm that a no-permission prompt or empty result is returned.
The values provided are common starting points and should be measured against applicable deployment-specific 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.