Class AIModelRunner

AIModelRunner — Lightweight AI model execution tracker for non-LLM model types.

Creates AIPromptRun records for embedding calls (and future: image, audio, video) with full token/cost tracking. Uses the same model selection and vendor failover infrastructure as AIPromptRunner but without template rendering or validation.

Architecture:

• @memberjunction/ai (no MJ infra): BaseEmbeddings, ModelUsage
• @memberjunction/ai-core-plus (this class): AIModelRunner — creates AIPromptRun records
• @memberjunction/ai-prompts: AIPromptRunner — LLM-specific (template rendering, validation)

Usage:

const runner = new AIModelRunner();
const result = await runner.RunEmbedding({
    Texts: ['Hello world', 'How are you'],
    ContextUser: currentUser,
    Description: 'Content item vectorization'
});
// result.PromptRunID links to the AIPromptRun record with token/cost data
Copy