Reference only; verify against the target FastGPT release and deployment documentation.
What the data for this category looks like
The business data for this category mainly comes from the national land spatial planning unified map, project approval and supervision platform, current land use database, and ecological protection redline data. Daily processed documents include land pre-examination and site selection opinions, construction land planning permits, construction engineering planning permits, redline diagrams, and planning conditions. The data includes fields such as project site selection coordinates, land scale, control boundaries, and approval requirement texts. Updates follow no fixed schedule, and are only triggered when a project advances or planning control rules are adjusted. Supported document formats include vector files, PDF approval documents, and structured forms.
Constraints imposed on deployment and upgrade
Most data sources come from government intranet systems. Deployment must adapt to intranet network isolation and data security specifications, and public network storage and database addresses cannot be used directly. Document formats include vector files and large-volume PDFs, so deployed parsing components must support multi-format parsing and large-file chunk processing. Data updates have no fixed cycle, so incremental synchronization and manually triggered full updates must be supported to avoid resource waste from fixed scheduled tasks. The data involves spatial control information, so the knowledge base’s vector recall must adapt to spatial feature matching logic. The vector model’s adaptation configuration must be updated synchronously during upgrades.
Configuration Settings
| Configuration Item | Recommended Setting | Rationale |
|---|---|---|
MONGODB_URI | Compliant replica set address for government intranets, with read-write separation configured | Business approval data is stored on the government intranet, which requires ensuring data security and read-write performance, and avoiding public network access risks |
PG_URL | Government intranet PostgreSQL address for storing approval process metadata | Approval nodes and requirement submission records for land pre-examination must be synchronized with government approval systems, so an exclusive intranet database must be used |
STORAGE_S3_ENDPOINT | Access address for intranet object storage services, with local caching prioritized for files under 1000 MB | File sizes such as redline diagrams and planning conditions vary widely; intranet storage ensures file reading speed, and local caching reduces intranet bandwidth usage |
REDIS_URL | Exclusive government intranet Redis instance address, with a 600-second session expiration time configured | Internal office sessions require session security on the government intranet; a 600-second expiration time meets the session validity requirements of government systems |
DB_MAX_LINK | 20–30 | The connection quota for government intranet databases is limited; this value avoids service exceptions caused by exceeding the quota |
FE_DOMAIN | Intranet domain name of the internal office system | Services are only open to intranet users, so a compliant government intranet domain name must be bound |
Three Common Mistakes
- Issue: After local deployment, uploading a PDF file related to land pre-examination results in empty data in the processing link. Cause:
STORAGE_S3_ENDPOINTis not configured to point to the intranet storage path. After file parsing, the file cannot be written to the knowledge base storage location normally, so retrievable vector data cannot be generated. - Issue: After building and running a Docker image, all vector index scores are identical. Cause:
REDIS_URLis not correctly configured to point to an exclusive intranet instance. Multiple containers share a public Redis cache, and vector recall results are contaminated by old cache, leading to abnormally uniform scores. - Issue: When calling a local large model, the first response works normally, but subsequent conversations with multi-round context trigger errors. Cause: A valid key for
FILE_TOKEN_KEYis not configured. Historical conversation context token verification fails, so multi-round conversation context data cannot be loaded, and interface errors are triggered.
How to Confirm Proper Configuration
- Upload a standard land pre-examination redline diagram PDF. Check the data processing logs to confirm that text fragments generated after file parsing are created normally and stored in the knowledge base.
- Start a conversation with multiple rounds of project site selection questions. Verify that vector recall results include the corresponding planning condition data, and that no abnormally uniform scores appear.
- Check database connection count monitoring to confirm that the current connection count stays within the range set by
DB_MAX_LINK. - Access the intranet domain name
FE_DOMAINto verify that the internal office interface loads uploaded approval requirement documents normally.
The values provided here are common starting points, and should be evaluated 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.