Metadata-Version: 2.4
Name: inferencebench-embeddings
Version: 0.0.2
Summary: Embeddings retrieval plugin for InferenceBench Suite (deterministic ranking skeleton; real embedding-model invocation deferred)
Project-URL: Homepage, https://github.com/yobitelcomm/bench
Author-email: Yobitel Communications <bench@yobitel.com>
License: Apache-2.0
Keywords: ai,benchmark,embeddings,ml,ndcg,recall,retrieval
Classifier: Development Status :: 2 - Pre-Alpha
Classifier: Intended Audience :: Developers
Classifier: Intended Audience :: Science/Research
Classifier: License :: OSI Approved :: Apache Software License
Classifier: Programming Language :: Python :: 3
Classifier: Programming Language :: Python :: 3.12
Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
Requires-Python: >=3.12
Requires-Dist: inferencebench-envelope
Requires-Dist: inferencebench-harness
Requires-Dist: pydantic~=2.9
Requires-Dist: pyyaml~=6.0
Description-Content-Type: text/markdown

# inferencebench-embeddings

Embeddings retrieval plugin for the InferenceBench Suite.

Phase-2-quality skeleton: produces signed envelopes via deterministic
hash-based rankings, with placeholders for real embedding-model invocation
that future revisions wire to TEI / OpenAI / Cohere.

Suite ID: `embeddings.retrieval`

Bundled benchmarks:

- `embeddings.retrieval.beir-mini` — 5 queries × 20-doc corpus, recall@5.
- `embeddings.retrieval.long-doc` — 3 queries with longer documents, nDCG@10.

The skeleton does NOT actually embed any text. For each query it ranks the
corpus by `sha256(query + doc_id)`, then scores the top-k against the
fixture's relevant set. This produces a real, well-defined retrieval metric
in [0, 1] without external dependencies — future revisions replace the
hash rank with a real vector search.
