Reference only; verify against the target FastGPT release and deployment documentation.
What this category of data looks like
Consulting and response data for survey and design work originates primarily from project survey reports, structural design drawing texts, on-site visa forms, design change documents, and similar materials. Data updates trigger irregularly alongside project progress, such as when on-site parameters change or design plans are adjusted. Most document structures include fields like project number, survey/design phase, professional chapter, and standard reference number. Some documents include professional indicators with units such as geotechnical parameters and load values. Overall text lengths vary widely, with individual documents reaching tens of thousands of characters.
What constraints do these characteristics impose on deployment and upgrade
The characteristics of survey and design data—long text volumes, numerous professional fields, and updates that fluctuate with project cycles—impose three core constraints on deployment and upgrade. First, long text requires adaptation to large context storage, and storage media for structured metadata and unstructured documents must be separated. Second, units and standard reference numbers of professional fields must be retained during knowledge base chunking to avoid loss of key information during retrieval. Third, the high-frequency updates during project cycles require support for incremental synchronization during upgrades, to prevent service interruptions caused by full reindexing. In addition, the consulting and response scenario needs to connect to internal office systems, so deployment must adapt to internal domain names and permission rules.
How to configure the settings
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
DB_MAX_LINK | 1000–1500 connections | Survey and design projects have large document volumes. This parameter limits the upper limit of database connection reuse, matching business concurrency requirements |
MONGODB_URI | Dedicated cluster address for storing unstructured documents | Meets the unstructured storage needs of long texts such as survey reports and design drawings |
PG_URL | Structured database address for storing project metadata | Supports relational queries for structured fields such as project numbers and standard reference numbers |
REDIS_URL | Local cache cluster address | Caches query results of high-frequency consulting requests to reduce pressure on backend databases |
FE_DOMAIN | Enterprise internal office domain name | Restricts front-end access sources, adapting to the security requirements of internal deployment for survey and design institutions |
Three common mistakes
- Issue: Text content extraction components cannot extract content from knowledge base references, and no corresponding option appears in the parameter list. Cause: Knowledge base-associated survey and design documents are not configured as extractable data sources, and only the general knowledge base index is enabled.
- Issue: Switching applications still triggers validation of required global variables from the previous application, causing an error. Cause: Session storage cache in Redis is not cleared during deployment, and session state is not reset after upgrade.
- Issue: Configuration modifications take no effect after Docker startup, and default values are restored after container restart. Cause: Local configuration volumes are not mounted, and configuration files inside the container are modified directly, leading to non-persistent configurations.
How to confirm the configuration is complete
- Access the enterprise internal domain name configured via
FE_DOMAIN, log in to the system, and view the list of knowledge base-associated survey and design documents to confirm that metadata loads normally. - Submit a consulting query for survey report parameters, check whether the returned results include corresponding professional text fragments in the knowledge base, and verify that the text extraction function works correctly.
- Switch between two test applications with different configurations, confirm that there are no conflicts in required variable validation, and verify that the global variable processing logic works correctly.
- View the database monitoring panel, confirm that the cluster connection counts corresponding to
MONGODB_URIandPG_URLare within a stable range, and there are no connection error alerts.
The values provided here are common starting points, and should be evaluated against the user’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.