What the data for this category looks like
Data sources for this category include design BOM, silicon steel sheet and insulating material inventory logs, vacuum drying equipment logs, insulating oil test reports, and high-voltage test station records. Data is updated per single product or batch material. Daily documents include winding and core assembly records, vacuum drying records, insulating oil reports, partial discharge and withstand voltage test reports, and nameplate parameter tables. Single product documents include fields such as insulation structure parameters, lead connection positions, tap switch gears, and test data. Some fields use industry-standard units.
What Constraints These Characteristics Impose on Multi-Turn Dialogue and Prompt Engineering
The feature that data is updated per single product or batch requires multi-turn dialogue to bind the corresponding product ID or batch identifier. This prevents mixing of data across products. Differences in fields across multiple document types require prompts to clearly distinguish verification logic for different document types. For example, insulating oil reports need to verify performance parameters. Test reports need to match test data ranges. The traceability requirement for single products requires dialogue context to retain all associated document information for that product. This avoids incomplete verification due to context truncation. The specialized nature of high-voltage test data requires prompts to limit verification scope to industry-standard process requirements. This prevents invalid deductions beyond the content of the documents.
Configuration Settings
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
| Maximum context length | 8000–12000 token | Adapts to the context limit of FastGPT v4.17.0, matches the total character count of multiple associated documents for a single product, and avoids context truncation |
messages field length limit | No more than 1800 characters per turn | Matches the typical length of a single professional document, and ensures completeness of single-turn verification |
variables binding parameters | Bind product ID, batch number | Corresponds to the feature that data is updated per single product or batch, and enables accurate data association |
chatId binding rule | Bind the unique identifier of a single product | Prevents context confusion in cross-product dialogues, and ensures traceability accuracy |
| Single recall document limit | Top 5 entries | Matches the number of associated documents for a single product, and avoids interference from redundant information |
| Question optimization switch | Enabled | Addresses verification needs for professional test data, and optimizes the precision of queries |
Three Common Errors
- Calling the dialogue record deletion interface returns a success status, but the dialogue corresponding to
chatIdstill remains. The cause is that thechatIdparameter of the target dialogue is not correctly bound, so the system cannot match the corresponding session. - After uploading or reading an insulating oil report or withstand voltage test report, the parsing result is empty or an error occurs. The cause is that an adapted parsing logic is not set for professional engineering documents, resulting in failure to recognize industry-specific fields.
- After enabling the question optimization function, no verification questions matching the document content are generated. The cause is that the prompt does not clearly limit the verification scope to the process parameters of transformers and high-voltage electrical equipment, or does not adapt to the query logic of professional documents, resulting in no valid verification questions being generated.
How to Verify Successful Configuration
- Initiate a multi-turn dialogue bound to a product ID. Check whether the
variablesparameter carries the identification information of the corresponding product, and confirm that data association is correct. - Upload or read a test report in Feishu online document format. Check whether the parsing result includes the core fields in the document, and confirm that the data source adapts to the category's document structure.
- Test deleting a dialogue with a specified
chatId. Confirm that the interface return matches the actual session status, and confirm that the session binding rule is effective. - Initiate a verification query after enabling the question optimization function. Check whether the generated reply focuses on category-specific content such as insulation parameters and test data, and confirm that the prompt configuration meets the requirements.
The values provided are common starting points and should be measured against the reader's own samples.