add support for pinecone retrieval and change backend

add logic to accumulate retrieved chunks in user session

fix: update routing decision logic to avoid storing in history and adjust citation format in retrieval

refactor: extract llm-as-judge logic to reusable module

refactor: update retrieval module to support configurable chunking strategy

refactor: streamline LLM call and maintain internal conversation history

update retrieval layout to be more verbose

Deploying Chainlit app to HF Spaces

add interface for generation and retrieval

initial commit