vai-workflow-cost-optimizer

Voyage AI's shared embedding space enables ~83% cost reduction by embedding with voyage-4-large and querying with voyage-4-lite. But developers want to verify the quality trade-off is acceptable for their specific data before committing.

Install

vai workflow install vai-workflow-cost-optimizer

How It Works

1. Parallel search — Query with both voyage-4-large and voyage-4-lite
2. Cost estimates — Get cost data for both models
3. Compare — Measure similarity between result sets
4. Report — Generate cost optimization analysis with recommendations

Execution Plan

Layer 1 (parallel):  search_large | search_lite | cost_large | cost_lite
Layer 2:             compare_quality
Layer 3:             optimization_report

Example Usage

vai workflow run vai-workflow-cost-optimizer \
  --input query="Explain the process for handling customer refunds" \
  --input collection="support_docs"

What This Teaches

• This directly quantifies the ~83% cost savings from asymmetric retrieval
• Four steps in parallel gather all data needed for comparison
• similarity between result sets measures overall retrieval agreement
• The generate prompt explicitly teaches about the shared embedding space

License

MIT © 2026 Michael Lynn