Package com.tnsai.evaluation.rag.embedding

package com.tnsai.evaluation.rag.embedding

Opt-in multilingual embedding retrieval bench (TAN-2998).

Default mvn test never talks to a model host. The measured mE5 vs BGE-M3 comparison runs only under -Pembedding-bench.

Since:

0.14.0

Class	Description
EmbeddingBenchCorpus	Fixed TR/EN retrieval fixture shipped with the evaluation module.
EmbeddingBenchModel	Named embedder under comparison.
EmbeddingBenchModelScore	Per-model scorecard across every slice.
EmbeddingBenchPair	One query and its gold document inside a EmbeddingBenchSlice.
EmbeddingBenchRecommendation	Picks a default candidate and a quality override from measured scores.
EmbeddingBenchRecommendation.Decision
EmbeddingBenchReport	Renders a markdown table for the measured model × slice matrix.
EmbeddingBenchRunner	Runs the bilingual fixture against one or more EmbeddingBenchModels.
EmbeddingBenchSlice	Language slice of the bilingual retrieval fixture.
EmbeddingBenchSliceScore	Retrieval metrics for one language slice.
OllamaEmbeddingFunction	EmbeddingFunction over an Ollama /api/embed endpoint.
TokenBagEmbeddingFunction	Deterministic bag-of-tokens embedding used as an offline bench baseline.