What This Category of Data Looks Like
Power generation equipment manufacturing internal office and employee service data mainly comes from device operation logs, maintenance work orders, employee attendance systems, and official technical documents. Device operation logs are updated at 5-minute intervals. Maintenance work orders are updated in real time as faults are submitted. Technical manuals are updated quarterly. Data includes two categories: structured and unstructured. Structured data includes fields such as 12-digit device numbers, maintenance duration (unit: hours), and work order priority. Unstructured data includes device maintenance PDF manuals and on-site operation videos, with a maximum size of 1000 MB per document.
Constraints Imposed During Deployment and Upgrade
Real-time device logs and high-frequency maintenance work orders require sufficient database connection counts during deployment to avoid concurrent write timeouts. Long technical manuals and video files require adapted large-capacity storage configurations to ensure stable upload and loading. The integrity requirements for historical maintenance data and device archives mandate retaining the original database structure during upgrades to prevent field loss. Device maintenance data involves production-sensitive information, so locally deployed models require deployment environments isolated from external networks, while also being compatible with port configurations for local model calls. Index mappings for device manuals must be updated synchronously during upgrades to ensure retrieval results match the latest document versions.
How to Set Configurations
Reference only; verify against the target FastGPT release and deployment documentation.
| Configuration Item | Suggested Value | Rationale |
|---|---|---|
DB_MAX_LINK | 20–30 | Adapts to the high-frequency concurrent write demands of device logs and maintenance work orders, avoiding connection pool exhaustion |
MONGODB_URI | mongodb://[username]:[password]@[host]:[port]/fastgpt?authSource=admin | Stores structured work orders and device archives, supporting multi-instance concurrent read and write |
STORAGE_S3_ENDPOINT | http://[internal-storage-ip]:9000 | Stores long-document PDFs and operation videos, adapting to large-capacity unstructured data storage demands |
REDIS_URL | redis://:[password]@[host]:[port]/0 | Caches frequently accessed device manual retrieval results, reducing database load |
FILE_TOKEN_KEY | Generated according to enterprise internal key specifications | Used to generate temporary access tokens for device manuals, ensuring controllable document access permissions |
ROOT_KEY | Generated according to enterprise internal security specifications | Used for system initialization and super permission control, ensuring security for internal deployments |
Three Common Mistakes
- Phenomenon: A
TypeError: Cannot read properties of undefinederror appears, and the knowledge base page fails to load. Cause: The originalMONGODB_URIconfiguration was not retained during upgrade, leading to database connection failure and failure to load existing device archives and work order indexes. - Phenomenon: Historical maintenance work orders cannot be viewed normally after a docker deployment upgrade. Cause: The original database version was not retained during upgrade, resulting in incompatible structured data fields and inability to read existing work order information.
- Phenomenon: A channel test error occurs when calling the deepseek model via a local ollama deployment. Cause: The
REDIS_URLcache for model call context was not configured, leading to concurrent requests exceeding limits and causing call timeouts.
How to Confirm Configurations Are Correct
- Run the
docker pscommand to confirm all FastGPT-related containers are running with no abnormal exit records. - Upload a 1000 MB device maintenance PDF document to confirm normal upload progress and successful access after storage completion.
- Initiate simulated maintenance work order creation and query requests to confirm no timeouts occur during database write and read operations, and that the
DB_MAX_LINKconfiguration does not trigger connection pool exhaustion. - Call the local ollama model interface, initiate a device fault troubleshooting query, confirm that structured replies can be returned normally with no channel errors.
The values provided are common starting points and should be measured 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.