Concepts

Understand the technology behind dish-embed. These pages explain the core ideas without requiring ML expertise.

• How Food Embeddings Work - What embeddings are and why food needs specialized ones
• Matryoshka Dimensions - Trade-offs between embedding size, quality, and cost
• Two-Stage Retrieval - How bi-encoder + reranker achieves production-grade accuracy
• Built-in Preprocessing - Noise stripping and normalization you get for free
• Dietary Detection - How veg/non-veg signals are identified and enforced