SignalRank-RAG v0.2.0 — Shipped Retrieval System

• Public Azure deployment with automated CI/CD
• Guided live retrieval demo
• BM25, dense, and hybrid retrieval
• Adversarial retrieval benchmark
• Per-result relevance feedback
• API service-token protection and request limits
• 57 automated tests across the CI matrix

SignalRank-RAG v0.1.0 — Retrieval Foundation

SignalRank-RAG v0.1.0

First tagged release of SignalRank-RAG.

v0.1.0 establishes the reproducible retrieval foundation for the project, with deterministic document processing, lexical and semantic retrieval, API and UI layers, observability, testing, and containerized execution.

Highlights

• Recursive TXT and Markdown ingestion
• Structured DOCX, PPTX, and XLSX ingestion
• Deterministic, content-sensitive document IDs
• Deterministic chunking with configurable overlap
• Document and element provenance preservation
• SentenceTransformer embeddings
• BM25 lexical retrieval
• Dense semantic retrieval
• Qdrant vector store integration
• FastAPI retrieval service
• Streamlit retrieval interface
• Distributed tracing with Logfire
• Docker and Docker Compose support
• Reproducible dependency management with uv
• 35 automated tests
• CI across Python 3.11–3.13 on Ubuntu and Python 3.11 on Windows

Running locally

The complete application can be started with:

docker compose up --build

Then open:

Streamlit UI: http://localhost:8501
FastAPI docs: http://localhost:8000/docs

Next

Future releases will extend this foundation with additional retrieval, ranking, and evaluation capabilities.