Jens
faeb58ef6d
GeoIntel release gates / Compile, test, contracts and builds (push) Successful in 1m49s
GeoIntel release gates / Python and npm vulnerability policy (push) Successful in 21s
GeoIntel release gates / Production AI image, SBOM and container scan (push) Successful in 5m39s
GeoIntel release gates / Deploy exact gated revision to Unraid (push) Failing after 58m43s
53 lines
863 B
Markdown
53 lines
863 B
Markdown
# Model Registry Specification
|
|
|
|
## Purpose
|
|
|
|
The model registry tracks which AI models are available, what they can detect or segment, and how outputs should be interpreted.
|
|
|
|
## V1 model record
|
|
|
|
```yaml
|
|
id: yolov8n-building-demo
|
|
name: YOLO Building Demo
|
|
framework: ultralytics
|
|
runtime: pytorch
|
|
task: object_detection
|
|
version: 0.1.0
|
|
classes:
|
|
- building
|
|
input:
|
|
tile_size: 640
|
|
channels: 3
|
|
outputs:
|
|
type: bounding_boxes
|
|
metrics:
|
|
mAP50: null
|
|
precision: null
|
|
recall: null
|
|
status: experimental
|
|
```
|
|
|
|
## Required metadata
|
|
|
|
- id
|
|
- name
|
|
- framework
|
|
- task
|
|
- version
|
|
- classes
|
|
- artifact_path or provider
|
|
- input constraints
|
|
- output schema
|
|
- status
|
|
|
|
## Status values
|
|
|
|
- experimental
|
|
- validated
|
|
- deprecated
|
|
- disabled
|
|
|
|
## Rule
|
|
|
|
Detection and segmentation services must request model info through the registry. They must not hardcode model paths in API routes.
|