What data for this category looks like
Image annotation project data includes six core document types: annotation specifications, bounding box examples, quality inspection reports, rework lists, version update instructions for single-batch annotation tasks, and aggregated operational report data. Data sources are image samples, annotation platforms, category dictionaries, and spot check results. Updates occur daily or per single-batch annotation task. Core fields include image identifiers, bounding box coordinates, category labels, quality inspection results, rework requirements, and operational statistics metrics. Unlike text annotation projects, this category requires processing position and category information for visual targets. The most time-consuming steps focus on defining target bounding boxes, judging occluded samples, and communicating rework requests. Errors may cause recognition errors, elevated rework rates, or client rejection.
What constraints do these characteristics impose on the database and operations workflow
The data characteristics of image annotation projects impose multiple constraints on the database and operations workflow. First, structured data volumes for image samples and annotation results are large. Deploy an independent database instance to isolate business data and avoid conflicts with data from other platforms. Second, the high-frequency update requirement (daily or per batch) requires the database connection pool to adapt to batch write scenarios and prevent connection exhaustion. Third, business requirements for release notes and aggregated operational reports require the database to support fast aggregate queries by time range and task batch. Create joint indexes for statistical fields. Frequently caching category dictionaries and bounding box examples can reduce real-time database pressure and minimize read-write conflicts.
How to set configurations
| Configuration Item | Recommended Value | Rationale |
|---|---|---|
MONGODB_URI | mongodb://anno_user:anno_pass@db-host:27017/image_anno?authSource=admin | Stores structured data such as bounding box coordinates and category labels for image annotations, and binds a dedicated database to isolate business data |
PG_URL | postgresql://anno_admin:anno_pwd@pg-host:5432/anno_docs?sslmode=require | Stores text-based documents such as annotation specifications, quality inspection reports, and rework lists, and supports multi-condition associated queries |
DB_MAX_LINK | 20–30 | Handles batch write requests for 100 to 1000 image samples per batch, and prevents database connection exhaustion |
REDIS_URL | redis://redis-host:6379/0?password=anno_redis_pwd | Caches category dictionaries and frequently accessed bounding box example data, and reduces real-time query pressure on the database |
| Database read-write timeout | 300 seconds | Adapts to batch annotation data writes and aggregate queries for operational reports, and avoids long-term blocking of requests |
Three common misconfigurations
- Symptom: A startup failure prompts a MongoDB connection error, with logs showing
Connection refusedorauth failed. Cause: The username, password, or database name in theMONGODB_URIconfiguration is incorrect, or the MongoDB service is not running properly. - Symptom: The database connection configuration page does not allow input of a port number, or saved port settings do not take effect. Cause: FastGPT v4.17.0 or later is not used, or there is a compatibility issue with the front-end configuration component.
- Symptom: Some fields are empty after batch annotation data is written, or operational report query results are incomplete. Cause:
DB_MAX_LINKis not set appropriately, leading to insufficient connection pool capacity to handle batch write requests, or no joint indexes are created for statistical fields, causing query timeouts.
How to confirm configurations are properly set
- Log in to the FastGPT backend database configuration page, and check if the configured values for
MONGODB_URI,PG_URL,DB_MAX_LINK, andREDIS_URLmatch the preset values. - Initiate a single image annotation data write request, and check the database monitoring panel to confirm that the number of connections does not exceed the
DB_MAX_LINKsetting. - Call the FastGPT annotation data query interface, and verify that the returned results include expected fields such as bounding boxes and category labels.
- Check the Redis monitoring panel, and confirm that the hit count of frequently cached data aligns with business expectations.
The values provided 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.