Rankify v0.1.4

Highlights

🛠️ Fixed numerous bugs across indexing and retrievers

📏 Integrated RAGAS evaluation metrics

🧩 Added new RAG pipelines/configs

💻 Introduced CLI for indexing: rankify-index

Bug fixes (selected)

• Lucene (BM25): stable JsonCollection wiring, index dir layout, robust load_index.

• Contriever: fixed JSONL/TSV mismatch; chunked embedding generation; float32 normalization; safer serialization & cleanup.

• BGE: correct CLS pooling + L2 normalization; cosine via IndexFlatIP; chunk merge validation.

• ColBERT: deterministic collection.tsv with sequential IDs; original↔sequential ID mappings; TSV verification & better diagnostics; loader-based load.

• ANCE: robust doc-id extraction across fields; consistent FAISS↔docid mapping; safer metadata writer.

• DPR: reliable Pyserini encode/index orchestration; mapping & metadata persisted.

Rankify v0.1.3

features:

• "🔧 Fixed UPR bugs to enhance retrieval and ranking stability."
• "📦 Removed 7z dependency and added native Python extraction for datasets."
• "🚀 Added Hyde Retriever (arXiv:2212.10496) for improved retrieval using hypothetical document embeddings."
• "📂 Expanded support for pre-retrieved datasets, adding more benchmark datasets."

Rankify v0.1.2

We have updated the pyproject.toml file to adjust the default installation behavior. With this release, users can install the Rankify library without including vllm by default. vllm is now part of the optional reranking dependencies group.

Rankify v0.1.0.post4

Edit pyproject.toml to include .cpp files for ColBERT-2

Rankify v0.1.0.post3

fix pip install

Rankify v0.1.0.post1

Enable automatic PyPI upload on release

v0.1.0

Full Changelog: https://github.com/DataScienceUIBK/Rankify/commits/v0.1.0

Rankify v0.1.0.post2

Bump version to 0.1.0.post2