Mouez-Yazidi/WhisperMesh

Implements a full RAG pipeline using Haystack with Qdrant vector storage, Cohere embeddings, and Groq-hosted speech-to-text (Whisper-large-v3) and generative models (gemma2-9b-it), enabling PDF document indexing and retrieval-augmented responses. Built on Streamlit with dual deployment options (local Docker or Streamlit Cloud), it processes user queries through semantic vector matching against indexed documents to ground LLM outputs in retrieved content. The modular architecture separates configuration for local versus cloud environments, allowing flexible infrastructure choices without code changes.

No commits in the last 6 months.