← All projectsIndependent repository
Webflyx RAG Studio
A Windows-first retrieval lab that combines keyword, semantic and image search with source-grounded question answering over a movie corpus.
PythonSentenceTransformersCLIP
Problem
Explore how retrieval choices affect the context supplied to a language model.
Architecture
BM25 and embedding search feed reciprocal-rank fusion, with optional cross-encoder reranking. CLIP supports image-to-text retrieval; model embeddings are cached locally.
Tradeoffs & constraints
Retrieval works without an LLM key. Generative features are optional and send selected context to OpenRouter only when invoked; model availability and loading remain explicit constraints.
Architecture summary based on the repository documentation (new tab). Metadata verified 2026-09-17; repository availability and functionality may change.