What the data for this category looks like
Production ledger data for this category comes from design BOMs, silicon steel sheet and insulating material procurement data, vacuum drying equipment operation logs, insulating oil test reports, partial discharge and withstand voltage test records from high-voltage test stations, and product nameplate parameter tables. Daily documents are categorized by production node: winding and core assembly records, vacuum drying records, insulating oil test reports, partial discharge and withstand voltage test reports, and others. Each field corresponds to professional parameters such as insulation distance, water content, partial discharge amplitude, and withstand voltage value, with units of millimeters, percentage, picocoulombs, and kilovolts respectively. Data is updated once per product or per batch of raw materials, and ledger data for a single product is associated with multiple documents and test results.
What constraints these characteristics impose on deployment and upgrade
Production ledger data for this category has three core characteristics: multi-source heterogeneity, frequent batch updates, and high precision requirements for professional fields. These characteristics impose multiple constraints on deployment and upgrade. Multi-source data must connect to formats including device logs, test reports, and test records. Deployment must support access configuration for multiple data source types. Single product or batch updates happen frequently. Upgrades must retain dual-mode configuration for incremental indexing and full updates. This avoids re-indexing full datasets. Professional parameters such as insulation distance and partial discharge values require accurate extraction. The document parsing engine must adapt to industry-specific document formats. Field verification rules must be configured to prevent erroneous data from entering the aggregation process. Factory production scenarios often have restricted network access. Deployment must support offline operation and local model docking configuration.
How to set configurations
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
DB_MAX_LINK | 200–400 concurrent connections | Adapt to batch processing demands for ledger data of multiple products in a single batch, avoid database connection overflow |
STORAGE_S3_ENDPOINT | http://192.168.xx.xx:9000 (local object storage address) | Match local storage deployment scenarios commonly used in factory intranets, ensure offline document storage is available |
REDIS_URL | redis://:custom password@127.0.0.1:6379/0 | Provide caching services to speed up retrieval and aggregation of multi-batch ledger data |
| Document chunking length | 800–1200 characters | Adapt to segmented parsing of long documents such as insulating oil reports and test reports, retain integrity of professional fields, support document parsing enhancement functions for FastGPT v4.9.0 and above |
| Offline deployment switch | Enabled | Adapt to network restricted requirements in factory production scenarios, support running all core functions locally without public network dependencies |
| Model docking method | Local ollama deployment access | Meet large model call demands in environments without public networks, comply with security requirements of industry production scenarios |
Three common misconfigurations
- Testing errors occur after adding the DeepSeek model deployed via local ollama. The cause is failure to configure the factory intranet network whitelist and failing to open the default port of the ollama service. This prevents the platform from accessing the local model service.
- Document parsing fails when using the v2 version of the marker image during offline deployment. The cause is failing to use the marker image version compatible with FastGPT v4.17.0, leading to dependency package incompatibility with core platform functions.
- The insulating oil water content field in the production ledger is empty in the aggregated report. The cause is failing to enable professional field extraction configuration and failing to match specific format tags for insulating oil reports. This results in failure to correctly extract key parameters.
How to confirm successful configuration
- Check connectivity for the
REDIS_URLconfiguration item. Use the platform's built-in connection test tool to verify the caching service is running normally. - Upload an insulating oil test report document. View the parsed field list to confirm inclusion of professional parameters such as water content and breakdown voltage.
- Initiate a ledger aggregation task for a single transformer. Check the task execution log for no error messages such as database connection timeouts or document parsing failures.
- After configuring local model docking, initiate a test query for insulation parameter verification. Confirm the model returns results using industry standard terminology.
The values provided are common starting points and should be measured against the reader's own samples.