What Data Looks Like for This Category
Data for this category comes from design BOMs, silicon steel sheet and insulating material procurement records, vacuum drying equipment logs, insulating oil test reports, and high-voltage test station records. Documents include winding and core assembly records, vacuum drying records, insulating oil test reports, partial discharge and withstand voltage test reports, and product nameplate parameter tables. Data is updated per single product or per batch of materials. Fields include insulation resistance value (unit: MΩ), partial discharge volume (unit: pC), test date, batch number, lead connection position, and other details. Requirements must match the product’s unique insulation distance and cleanliness standards.
Constraints Imposed on Model Access and Configuration
Single-product-specific documents and batch-updated data require the model context to match the full set of test, assembly, and material data for a single product. This prevents cross-batch data confusion. High-precision unit requirements for insulation parameters require the model to accurately identify and retain professional units such as MΩ and pC, avoiding unit conversion errors. Frequent demand for drafting test anomaly reports requires the model’s response speed to fit the time window of the test process, and the ability to associate corresponding batch material data. Additionally, full product lifecycle traceability requirements mean documents generated by the model must be bound to a unique batch number and nameplate parameters. Configuration must ensure the accuracy of data association logic.
The following configuration is written for FastGPT v4.17.0.
Configuration Settings
| Configuration Item | Recommended Range | Rationale |
|---|---|---|
maxContext | 8000–12000 token | Fits the full test, assembly, and material data for a single product, avoiding truncation of critical insulation parameters |
maxTokens | 3000–4000 token | Meets the length requirements for a single process document or anomaly report, and supports detailed test report descriptions |
quoteMaxToken | 1500–2000 token | Prioritizes referencing core data from insulating oil reports and partial discharge test reports, ensuring complete referenced content |
temperature | 0.2–0.4 | Reduces generation randomness, ensuring accuracy of test data and parameters, and meeting the strict requirements of manufacturing scenarios |
AIPROXY_API_ENDPOINT | Fill in the endpoint provided by the accessed model service provider | Adapts to proxy configurations for OneAPI or Ollama, ensuring normal model call links |
AIPROXY_API_TOKEN | Exclusive token obtained from the service provider | Verifies model call permissions, avoiding unauthorized access or over-limit calls |
Three Common Configuration Errors
- Phenomenon: The prompt "No available channels for model whisper-1 under current group default" appears when calling the model, and logs show the request failed. Cause: No Whisper-related model is configured in the model group, or a model not included in Ollama is used, and supported models are not added via OneAPI.
- Phenomenon: A configured TTS model generates results with no corresponding voice, or configuration items are invalid. Cause: The
voicesparameter is not filled according to model requirements, and does not match the voice list supported by the model, leading to parameter parsing failure. - Phenomenon: Large numbers of unauthorized model call requests appear at night, and account balance is rapidly depleted. Cause: No access permission restrictions are configured for the API token, or the token has been leaked, and no call frequency or time period restrictions are set, leading to malicious or accidental calls.
How to Confirm Configuration Is Correct
- Initiate a process document generation test for a single transformer, and check whether the generated content includes core parameters such as insulation resistance and partial discharge volume, with accurate units.
- View model call logs, and confirm that the configuration items used in the request match the actual filled values, with no missing or incorrect parameters.
- Test the anomaly report drafting scenario, input partial discharge abnormal test data, and check whether the generated content associates corresponding batch material data.
- Verify API call permission restrictions, simulate an unauthorized request, and confirm it is blocked to avoid over-limit calls.
The values provided are common starting points and should be measured against the reader’s own samples.