What this category of data looks like
This category of data is typically sourced from object storage, data warehouses, annotation platforms, and customer acceptance feedback. Updates happen once per batch or per version. Daily output documents include dataset descriptions, version records, delivery lists, and acceptance reports. Common data fields include version number, dataset unique identifier, delivery scope, and permission binding fields. No additional statistical percentage parameters are included.
Constraints imposed by these characteristics on permissions and data isolation processes
This category of data has dispersed sources, involving multiple storage carriers such as object storage and data warehouses. Permission configurations must match the access rules of different storage sources to avoid cross-source data mixing. Data updates follow a per-batch or per-version schedule. Each version iteration requires synchronous updates of permission binding relationships to prevent expired permissions for old versions. Daily documents have fixed-structure dataset descriptions and delivery lists. Permission configurations must align with the identification fields in these documents to avoid configuration mismatches. Delivery scenarios involve multiple parties. Strict isolation must be maintained between customer-owned data and platform public data to prevent data leaks.
How to set configurations
| Configuration Item | Recommended Approach | Rationale |
|---|---|---|
teamId binding strategy | Bind a unique teamId to each delivery batch. Prohibit cross-batch sharing | Matches the delivery subject identifier of the dataset, isolates datasets from different customers. This configuration can be set via the permission management page in FastGPT v4.17.0 |
tmbId access scope | Restrict to the storage directory of the current delivery package, formatted as /dataset/v[version number]/ | Limits data access only to the delivery package of the corresponding version, prevents unauthorized access to cross-version data |
| Login-free link validity period | 72 hours | Balances access convenience and data security, avoids long-term exposure risks of links |
| Login-free link quantity limit | Configured according to the subscription plan, can be adjusted to 10–100 | Matches the actual usage needs of customers, supports modifying the default limit of 10 as needed |
| Version record permissions | Only authorize data project managers to edit version records | Matches the job scope of data project managers, prevents unauthorized modification of version information |
Three common configuration mistakes
- Symptom: Login-free links can access datasets from unbound batches. Cause: The
teamIdis not bound to the permission rules of the login-free link, so the link is not associated with the permissions of the exclusive delivery batch. - Symptom: The login-free link quantity limit cannot be adjusted to 100. Cause: The link quantity limit parameter is not modified in the permission configuration of the corresponding subscription plan, and the default limit of 10 is retained.
- Symptom: Old version datasets can be accessed via new permission rules. Cause: The
tmbIdaccess scope configuration is not updated synchronously, so the storage directory of the old version has not had its permission restrictions removed.
How to confirm correct configuration
- In the permission management panel of FastGPT v4.17.0, check the
teamIdbinding status of each delivery batch, confirm there are noteamIdconfigurations that allow cross-batch sharing. - Generate a login-free link, verify that only users corresponding to the bound
teamIdcan access the corresponding dataset, and unauthorized users cannot open the link. - Adjust the login-free link quantity limit, verify that after the configuration takes effect, the number of created links matches the modified value.
- Attempt to edit version records, confirm that only authorized roles can perform edit operations, and unauthorized roles cannot modify version information.
The values given 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 2026-10-19.