What the data for this category looks like
Matter consultation data for rail transit equipment comes from original factory technical manuals, in-service equipment maintenance ledgers, and industry operation and maintenance standard documents. The update rhythm is monthly sync for maintenance ledgers, and quarterly updates for technical standards. Documents are structured by equipment components such as bogies, traction systems, and brake systems. Fields include component number, fault code, processing procedure, spare part model, and more. Most units are engineering units such as millimeters, newtons, and operating hours. Document length varies widely, from hundreds of characters of fault alerts to tens of thousands of characters of complete equipment operation and maintenance manuals.
Constraints imposed on forms and interactions by these characteristics
The multi-component classification attributes of rail transit equipment require forms to split interaction entrances by module. This prevents users from entering redundant non-target information. Fields with multiple engineering units require forms to automatically bind unit prompts. This stops input values from mismatching required units. The wide range of document lengths requires form interactions to support filtering retrieved knowledge base content by component type. This avoids interference from irrelevant information. The monthly updated maintenance ledgers require the knowledge base associated with the form to match the update rhythm. This ensures returned content aligns with the latest specifications for current in-service equipment.
How to set the configuration
Reference only; verify against the target FastGPT release and deployment documentation.
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
| Chunk Length | 800–1200 characters | Adapt to the document length of single component modules for rail transit equipment |
userSelect | Preset options categorized by equipment component | Match the interaction constraints of multi-component classification |
| Knowledge Base Sync Cycle | Once per month | Match the monthly update rhythm of maintenance ledgers |
| Result Retrieval Count | Top 3–5 entries | Avoid context overflow caused by too many long documents |
| Unit Validation Switch | Enabled | Adapt to the field requirements of multiple engineering units |
formInput | Only used for free text supplementary items | Limit the scope of free input and reduce invalid content |
Three common configuration errors
- Phenomenon: After specifying a model version in configuration, calls still use the default version. Cause: The model version parameter was not bound in the associated configuration of
formInputoruserSelect, causing the system to call the default model. - Phenomenon: Knowledge base retrieval results do not match the equipment component entered by the user. Cause: The filtering rule based on component classification was not enabled, and the retrieval scope covers all documents, interfering with effective information extraction.
- Phenomenon: After text extraction fails, the specified reply does not trigger a wait for user input. Cause: No wait-for-user-input node was configured in the workflow, only the specified reply action was executed, and the workflow ended directly.
How to verify a successful configuration
- Enter the form configuration page, check if the options for
userSelectare preset categorized by equipment component. - Trigger form submission, verify that the returned knowledge base content matches the currently selected equipment component type.
- Simulate input with engineering units, check if unit validation prompts are triggered.
- Manually trigger knowledge base synchronization, confirm that the sync cycle matches the preset once per month.
- Check the configuration scope of
formInput, confirm that it is only used for free text supplementary items. - Check the form timeout setting, confirm that the value adapts to the input thinking time required for complex equipment consultations.
The values provided here are common starting points, and should be evaluated 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.