Services
Requesting Predictions
The AI services each have a route to the Seldon service. To create a prediction a POST
request to the route under the path /api/v0.1/predictions
. For instance, the flake analysis route on a CRC cluster in the Project `odh’ will look something like:
https://regrclassifier-odh.apps-crc.testing/api/v0.1/predictions
Models vary according to each service, but often, data is passed in through a field strData
and parsed by the model accordingly. Within this strData
field, most models require the a list of key value pairs separated by commas. For this reason, strData must have any other commas stripped out. Two common keys are model
and data
. A simple JSON body would look like:
{"strData":"model=healthpredictor/model.pkl, data=66:150"}
Putting that together in a curl command:
curl 'https://regrclassifier-odh.apps-crc.testing/api/v0.1/predictions' \
-X POST \
-H 'Content-Type: application/json' \
-H 'Accept: application/json' \
-d '{"strData":"model=healthpredictor/model.pkl, data=66:150"}'
Sample Output:
{
"meta": {
"puid": "k8h9uir10qdbsmv1qigp2qnk5l",
"tags": {},
"routing": {},
"requestPath": {
"c-regrclassifier": "quay.io/opendatahub/regression_predict"
},
"metrics": []
},
"data": {
"names": [],
"tensor": {
"shape": [
1
],
"values": [
2.2885771963212247
]
}
}
}
Service Documentation
As each service varies, please visit the documentation for the service and model.
- Anomaly Detection
- Association Rule Learning
- Correlation Analysis
- Duplicate Bug Prediction
- Flake Analysis
- Fraud Detection
- Matrix Factorization
- Regression Analysis
- Sentiment Analysis
- Topic Modeling