What the Data for This Category Looks Like
Business data sources include concentrate and auxiliary material analysis reports, furnace condition and cell voltage/current data collected by DCS systems, composition test results issued by the laboratory, real-time operation records of electrolytic cell groups, and energy and environmental monitoring ledgers. Data update frequency changes with business nodes. Each batch of raw materials, each smelting furnace, and each electrolytic cell generates independent records. Key operating parameters are synchronized in real time. Daily documents include batching lists, smelting logs, electrolytic cell condition tables, anode and cathode quality reports, environmental protection ledgers, and others. Fields include batch numbers, furnace numbers, cell numbers, composition test values, voltage values, current values, and others. Some documents are structured tables, others are unstructured text.
Constraints Imposed by These Characteristics on Deployment and Upgrades
Multiple heterogeneous data sources require adaptation to different data source access formats during deployment. This avoids missing core business data such as concentrate analysis and cell condition monitoring. The real-time update feature of the business requires the connection stability between the vector database and business database to meet standards. Real-time data synchronization cannot be interrupted during upgrades. Otherwise, the quality traceability link will be broken. Diverse document structures require the system to support multi-format parsing and field mapping. This ensures that different types of business documents such as batching lists and smelting logs can be correctly indexed. This category has high requirements for impurity control and metal recovery accuracy. Retrieval rules must be optimized for core business fields during deployment. Upgrades must ensure compatibility with existing indexes, to avoid affecting launched quality analysis processes.
Configuration Settings
These configurations apply to FastGPT v4.17.0.
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
PG_URL | postgresql://{user}:{pass}@{host}:5432/fastgpt?schema=metal_smelt | Core business data of this category is mostly stored in PostgreSQL. A dedicated business schema is required to isolate data |
REDIS_URL | redis://:{pass}@{host}:6379/0 | Used to cache real-time electrolytic cell condition and batch data, to reduce repeated query overhead |
DB_MAX_LINK | 30–50 | This category has frequent real-time data writes, to avoid exhausting the database connection pool |
STORAGE_S3_ENDPOINT | Configured according to the intranet endpoints of domestic cloud service providers | Batch upload large files such as test reports and smelting records, to reduce external network transmission latency |
| Chunk length | 800–1200 characters | Matches the average length of documents such as batching lists and smelting records, to avoid losing context during chunking |
| Single search limit | Top 10 entries | Covers associated data from multiple cell positions and furnace heats, to ensure retrieval coverage |
Three Common Mistakes During Deployment and Upgrade
- Phenomenon: Pulling the official DockerHub image fails during deployment, and the FastGPT container cannot be started. Cause: No domestic mirror source is configured, and the domestic network cannot directly access the DockerHub public repository.
- Phenomenon: Workflow calls return empty values, and knowledge base retrieval has no matching results. Cause:
PG_URLdoes not point to the business database storing smelting records and test reports, or the vector database connection parameters are configured incorrectly. - Phenomenon: Vector retrieval results have insufficient matching accuracy, and defect analysis accuracy decreases. Cause: pgvector 0.7.4 or higher compatible with PostgreSQL 15 is not used, resulting in abnormalities in vector storage and retrieval.
How to Verify Correct Configuration
- Run the
docker pscommand, confirm that FastGPT-related containers are in running status, with no abnormal exit records. - Log in to the FastGPT backend, enter the knowledge base configuration page, upload an electrolytic cell condition record document, confirm that the document is parsed successfully and field mapping meets business requirements.
- Call the test API, pass a query instruction containing cell numbers and furnace heats, confirm that non-empty retrieval results and associated business data are returned.
- View the database connection monitoring panel, confirm that the connection count does not reach the
DB_MAX_LINKconfiguration limit, and the connection pool is operating normally.
The values given are common starting points and should be measured against the reader's own samples.