What data for this category looks like
Core data for text annotation projects comes from customer corpora, samples produced by annotation platforms, custom vocabularies, and quality inspection sampling results. Daily circulating documents include annotation specifications, batch sample packages, quality inspection reports, and problem feedback forms. Data updates follow a daily incremental update schedule or updates per annotation batch. The size of single-batch sample packages varies based on business requirements. Data fields include sample text, label hierarchy, quality inspection results, annotator information, and others. There are no unified fixed unit requirements, and adjustments must be made based on the specific annotation task type.
What constraints do these characteristics impose on deployment and upgrade
Text annotation project data has scattered sources and high update frequency, and requires synchronization of business documents such as annotation specifications and quality inspection rules. During deployment, access configuration for multiple data sources must be adapted to ensure unified circulation of sample data from different sources. Due to frequent data updates, upgrade processes should minimize downtime to avoid disrupting ongoing annotation and ticket processing workflows. Errors in this category of data will directly lead to model training bias, rework, or delayed delivery. During deployment, complete backup and verification mechanisms must be configured to ensure data consistency before and after upgrades. Additionally, core logic such as text annotation semantic boundaries and entity recognition relies on configuration parameters. During upgrades, the matching degree between business rules and system parameters must be verified synchronously.
How to set the configurations
| Configuration Item | Recommended Approach | Rationale |
|---|---|---|
PG_URL | postgresql://postgres:[custom password]@[database address]:5432/fastgpt?sslmode=disable | Metadata and ticket records for text annotation projects must be stored in PostgreSQL. The default postgres role is the officially recommended initialization role |
REDIS_URL | redis://:[custom password]@[cache address]:6379/0 | High-frequency sample retrieval and ticket status synchronization require low-latency cache support. This port is the default Redis service port |
DB_MAX_LINK | 20–30 | Text annotation projects handle a large number of sample package uploads and quality inspection sampling requests daily. Excessive connection counts will cause database overload |
STORAGE_S3_ENDPOINT | Configure per the endpoint provided by the object storage service provider | Large files such as annotation sample packages and quality inspection reports require unified storage. This configuration specifies the access path for file storage |
ROOT_KEY | Random 32-bit string generated per business permission rules | Permission control for internal tickets relies on this key to prevent unauthorized access to core configuration of annotation projects |
FE_DOMAIN | Fill in the access domain name of the FastGPT front-end service | Used to generate jump links for internal tickets, ensuring annotators can access relevant pages normally |
Three common mistakes
- Symptom: The PostgreSQL container displays "FATAL: role "postgres" does not exist" after startup. Cause: The default role named postgres was not created in the target database instance in advance. Metadata storage for text annotation projects relies on this role for initialization.
- Symptom: Authentication failure or timeout errors occur when connecting to an externally deployed large model service. Cause: The
DB_MAX_LINKparameter was not adjusted based on the business requirements of the text annotation project, leading to database connection exhaustion and subsequent failure to call model services normally. - Symptom: The internal ticket file upload function fails after upgrading to FastGPT v4.17.0. Cause: The
STORAGE_S3_ENDPOINTconfiguration was not updated synchronously, causing the upgraded system to fail to access the original annotation sample storage path.
How to confirm configurations are correct
- Run the official database connectivity test script to verify whether the
PG_URLandREDIS_URLconfigurations are normal, and confirm that the database role and cache service can connect normally. - Upload a small test annotation sample package to verify whether the
STORAGE_S3_ENDPOINTconfiguration can correctly store files and generate accessible links. - Log in to the internal ticket system using the permission account corresponding to
ROOT_KEY, submit a test ticket and complete its workflow, and verify that permission control and ticket processes are working properly. - Check container runtime logs to confirm there are no abnormal messages such as database connection timeouts or file storage errors, and confirm that the upgraded system is running stably.
The values provided are common starting points and should be evaluated against local deployment 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.