What the data for this category looks like
Data sources for process and standard data in petroleum refining scenarios include DCS/APC control systems, LIMS laboratory information management systems, tank area metering systems, crude oil evaluation reports, equipment management systems, and energy management systems. Data updates follow three frequencies: real-time, shift-based, and daily. Document types include operating procedures, production plans, material balance sheets, distillate outlet quality reports, interlock ledgers, shift handover records, and change management documents. Data fields include unit load, distillation range, sulfur content, tank inventory, and more. Corresponding units include tons per hour, ℃, wt%, m³, and others.
What constraints these characteristics impose during deployment and upgrade
This document is written for FastGPT v4.17.0. Since data sources cover multiple heterogeneous industrial systems, deployment requires support for batch import and parsing of multiple document formats. Real-time and high-frequency updated data requires vector index refresh strategies aligned with business rhythms. This avoids index lag that harms Q&A accuracy. Process data from multiple coupled devices has strong relational connections. Knowledge base context recall must adapt to cross-device logical associations. Complex document structures require chunking strategies that balance field integrity and retrieval accuracy. This prevents splitting critical process parameters.
How to set the configurations
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
DB_MAX_LINK | 10–15 | Matches the high-frequency index generation and query concurrent requests in petroleum refining scenarios |
MONGODB_URI | Cluster connection address with read-write separation | Stores massive process documents and vector data to ensure read and write performance |
REDIS_URL | Cluster address with RDB persistence | Caches real-time updated process parameters and session information to reduce database load |
FE_DOMAIN | Legal domain address for WeChat projects | Configures cross-domain access permissions to adapt to access requirements of WeChat projects |
STORAGE_S3_ENDPOINT | Endpoint address adapted to the enterprise's internal object storage | Stores large files such as crude oil evaluation reports and interlock ledgers |
ROOT_KEY | Exclusive key generated according to the enterprise's permission system | Manages the highest access permissions for the knowledge base, complying with industrial data security requirements |
FILE_TOKEN_KEY | 32-bit randomly generated string | Encrypts access tokens for file upload and download to ensure document data security |
Three common mistakes
- Symptom: Index creation steps stall, and the interface shows index generation timeout. Cause: The
DB_MAX_LINKconfiguration was not adjusted to match the concurrent requirements of refining scenarios. Insufficient concurrent connections cause index write blocking. - Symptom: After adding an One API channel, it displays as unavailable. The test passes, but calls return an invalid token error. Cause: The
ROOT_KEYand token verification logic were not configured correctly, and the One API permission verification rules were not matched. - Symptom: Container-deployed FastGPT cannot connect to a locally deployed WeChat project. Interface calls return a cross-domain error. Cause: The legal domain address of the WeChat project was not added to
FE_DOMAIN, so cross-domain requests are blocked.
How to confirm the configuration is complete
- Run FastGPT's built-in database connectivity test to verify that the connection status of
MONGODB_URIandREDIS_URLis normal. - Upload a typical petroleum refining operating procedure document to trigger index generation. Check whether the index progress completes within the preset time.
- After configuring the One API channel, initiate a test call. Confirm that the returned results match the content of the uploaded process documents.
- Access the domain address configured in
FE_DOMAINto verify that the FastGPT interface loads normally and no permission error messages appear.
The provided values are common starting points and should be evaluated against the applicable samples of the deployment environment.