What Data for This Category Looks Like
Data related to refinery production inquiries and responses mainly comes from production logs, equipment inspection records, emergency response manuals, and standard operating procedure documents. Inspection records are daily updated structured data, including fields such as equipment ID, inspection time, operating parameters, and units including hours, degrees Celsius, MPa, and others. Logs and manuals are unstructured documents revised as needed, mostly in long text format, including step instructions, troubleshooting workflows, and similar content. Most data storage uses a combination of internal file servers and structured databases. Individual document lengths vary widely, ranging from hundreds of characters for inspection entries to tens of thousands of characters for emergency manuals.
What Constraints These Characteristics Impose on Deployment and Upgrade
Refinery production inquiry data has both structured and unstructured attributes, with clearly different update frequencies. This creates multiple constraints for the deployment and upgrade process. Structured inspection data requires support for complex queries, so databases must support multi-field index configuration. Long-text emergency manuals and procedure documents require support for large file parsing and chunked storage. Sufficient storage space and a reasonable chunking strategy must be configured. Different data sources have different update rhythms. Flexible switching between incremental and full updates is required to avoid reloading full datasets repeatedly. In addition, internal data access must follow state-owned enterprise and public sector cybersecurity norms. Dedicated storage ports and access permissions must be configured, and adaptation to internal domain names and authentication rules is required.
How to Set Configurations
| Configuration Item | Recommended Value | Basis for This Setting |
|---|---|---|
STORAGE_S3_ENDPOINT | http://{internal S3 service IP}:9000 | Refinery production inquiry documents are mostly PDFs, inspection reports, and similar content, which need to be stored via an S3 service. Port 9000 is the officially recommended internal S3 service port |
REDIS_URL | redis://:{cache password}@{internal host address}:6379/0 | Cache inquiry sessions and index results to ensure internal user access response time does not exceed 2 seconds |
MONGODB_URI | mongodb://:{database password}@{internal host address}:27017/fastgpt?authSource=admin | Store knowledge base entries and user session data, adapting to multi-field structured query requirements |
PG_URL | postgresql://:{database password}@{internal host address}:5432/fastgpt | Store structured inquiry procedures and equipment log data, supporting complex conditional retrieval |
FE_DOMAIN | https://{internal enterprise inquiry domain} | Adapt to state-owned enterprise internal domain specifications, ensuring internal users can normally access the FastGPT frontend interface |
DB_MAX_LINK | 10–20 entries | Control the number of knowledge base recall entries for inquiries, adapting to the volume of refinery production data documents |
Three Common Mistakes
- Phenomenon: A 504 timeout error is returned when calling the enhanced PDF parsing function, and the interface displays parsing failed. Cause: The local port of the MinerU service is not opened in the internal firewall, and the correct MinerU API address is not specified in the FastGPT configuration.
- Phenomenon: Knowledge base index construction fails, and the log shows an S3 connection timeout. Cause: Only port 9000 is opened, but
STORAGE_S3_ENDPOINTis not correctly configured to the actual address of the internal S3 service, resulting in inability to access storage resources. - Phenomenon: The number of recalled results for inquiries is insufficient, failing to cover the required inquiry procedure content. Cause: The
DB_MAX_LINKparameter is not adjusted to a reasonable range, resulting in too few recalled knowledge base entries.
How to Confirm Configuration is Complete
- Run a curl command to access the address configured in
STORAGE_S3_ENDPOINT. If a valid list of stored files is returned, the storage configuration is correct. - Log in to the FastGPT v4.17.0 backend, enter knowledge base settings, enable the enhanced PDF parsing function, upload a refinery production inspection PDF document, and check whether the parsing status shows completed after 5 minutes.
- Initiate a simulated inquiry, such as "What is the daily inspection procedure for model XX pumps?", and check whether the corresponding knowledge base document can be recalled and a qualified response is generated.
- Check internal firewall rules, confirm that the port corresponding to
STORAGE_S3_ENDPOINTis opened, and that the internal network can normally access the service corresponding to this port.
The configuration values listed above are common starting points. They should be evaluated against the user’s own specific samples and operating environment.
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.