What the data for this category looks like
Civil aviation airport document data primarily comes from official operation manuals, flight support procedures, air traffic control coordination documents, passenger service guidelines, and emergency response plans. Staff update data monthly to align with new civil aviation authority regulations, new route launches, or airport expansion and renovation projects. Most documents use a chapter-based structure, including fields such as document number, effective date, applicable positions, and execution standards. Common units include passenger trips, flight sorties, and support duration. The length of individual documents ranges from hundreds to tens of thousands of characters.
What constraints these characteristics impose on knowledge base retrieval and recall
The chapter-based structure of civil aviation airport documents requires retrieval to match the applicable scenarios of corresponding positions, to avoid retrieving redundant content across positions. The monthly update frequency requires the knowledge base to support incremental synchronization mechanisms, reducing resource consumption from full refreshes. Specialized units are tied to specific scenarios, so retrieval must associate the queried position or business type when recalling, to filter mismatched documents. The wide range of individual document lengths requires limiting the total character count of recalled content, to avoid exceeding the context window limit.
How to set configurations
Reference only; verify against the target FastGPT release and deployment documentation.
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
similarity | 0.75–0.85 | Civil aviation airport documents have dense specialized terminology. This range balances retrieval precision and coverage, filtering irrelevant content with low matching degrees |
limit | Top 6–10 results | Airport documents have large differences in length. Limiting the number of recalled results controls total token consumption and avoids context overflow |
searchMode | hybrid | Combines semantic and keyword matching, adapting to retrieval needs for specialized terminology and fixed identifiers such as flight numbers and document numbers |
usingReRank | Enabled | Reorders initial recall results by relevance, correcting deviations from semantic matching and ensuring core business content is returned first |
datasetSearchUsingExtensionQuery | Disabled | Most airport documents are fixed professional content, so no query extension is needed, avoiding introduction of irrelevant general information |
collectionIds | Bind the current airport's exclusive business collection ID | Only recall operation documents from this airport, eliminating redundant data interference from other airports |
Three common errors
- Phenomenon: The knowledge base returns no recall results, but still generates irrelevant casual replies unrelated to airport business. Cause: Only relies on knowledge base recall results to judge query relevance, without configuring pre-filtering rules for business scenarios, and cannot identify queries with no clear business purpose.
- Phenomenon: The total character count of recalled documents in a single query exceeds the context window, causing replies to be truncated or interface errors. Cause: No reasonable
limitparameter is set, or the number of recalled results is not adjusted based on the length differences of airport documents, and total token consumption is not controlled. - Phenomenon: Replies include citations from documents not in the current business collection, or citations from irrelevant airport documents. Cause: No exclusive
collectionIdsare bound, resulting in recall of redundant documents from other airports or non-business categories.
How to confirm configurations are set correctly
- Submit a query containing airport-specific business terminology, and check whether the content of the corresponding documents is included in the recall results.
- Submit a casual query with no clear business purpose, and check whether irrelevant replies are returned or the preset filtering logic is triggered.
- Check the knowledge base synchronization records, confirm that only updated airport documents have been synchronized, and no full refresh operations have been performed.
- Check the generated reply, confirm that all cited documents come from the exclusive business collection configured currently, and no redundant content across airports appears.
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.