52 lines
1.5 KiB
JSON
52 lines
1.5 KiB
JSON
{
|
|
"$schema": "https://json-schema.org/draft/2020-12/schema",
|
|
"$comment": "SAP AI Core Deployment Configuration Template",
|
|
|
|
"foundation_model_configuration": {
|
|
"name": "my-model-deployment-config",
|
|
"executableId": "azure-openai",
|
|
"scenarioId": "foundation-models",
|
|
"parameterBindings": [
|
|
{"key": "modelName", "value": "gpt-4o"},
|
|
{"key": "modelVersion", "value": "latest"}
|
|
]
|
|
},
|
|
|
|
"orchestration_configuration": {
|
|
"name": "my-orchestration-config",
|
|
"executableId": "orchestration",
|
|
"scenarioId": "orchestration",
|
|
"parameterBindings": []
|
|
},
|
|
|
|
"deployment_request": {
|
|
"configurationId": "<configuration-id-from-above>",
|
|
"ttl": "24h"
|
|
},
|
|
|
|
"deployment_with_replicas": {
|
|
"configurationId": "<configuration-id>",
|
|
"ttl": "7d",
|
|
"minReplicas": 1,
|
|
"maxReplicas": 3
|
|
},
|
|
|
|
"_documentation": {
|
|
"ttl_format": "Natural numbers with units: m (minutes), h (hours), d (days)",
|
|
"ttl_examples": ["5m", "2h", "7d", "30d"],
|
|
"executable_ids": {
|
|
"azure-openai": "Azure OpenAI models (GPT-4o, GPT-4, GPT-3.5)",
|
|
"aicore-opensource": "SAP-hosted open source (Llama, Mistral, Falcon)",
|
|
"gcp-vertexai": "Google Vertex AI (Gemini, PaLM)",
|
|
"aws-bedrock": "AWS Bedrock (Claude, Titan)",
|
|
"aicore-mistralai": "Mistral AI models",
|
|
"aicore-ibm": "IBM Granite models",
|
|
"orchestration": "Orchestration service"
|
|
},
|
|
"model_version_options": {
|
|
"latest": "Auto-upgrade to newest version",
|
|
"specific": "Pin to specific version (e.g., '2024-05-13')"
|
|
}
|
|
}
|
|
}
|