feat: add conversation history support to chat service and update Gradio UI for interactive sessions

feat: implement modular RAG architecture, query routing, metadata-filtered retrieval, and a comprehensive offline testing suite.

refactor: move title from launch configuration to Blocks constructor in Gradio interface

refactor: migrate chat and ingestion logic into dedicated services and consolidate database initialization

feat: implement AI query router and metadata-based filtering for RAG retrieval with updated UI controls

feat: implement RAG pipeline with Qdrant vector store, Groq LLM integration, and structured prompt management

feat: migrate LLM from local Flan-T5 to Groq API and add environment configuration files

feat: refine system prompt, set max sequence length, and adjust generation parameters for improved response quality

refactor: replace transformers pipeline with manual model loading and add Gradio UI interface

refactor: restrict retrieval to the single most relevant document and remove Gradio UI implementation

refactor: remove code comments, optimize retrieval logic, and refine LLM prompting and output formatting

refactor: reduce retrieval counts, improve document filtering logic, and enforce stricter prompt constraints for the LLM assistant

fix: update pipeline task type from text2text-generation to text-generation

refactor: optimize RAG pipeline with Flant-T5, centralized retrieval, and UI refinements

feat: update Qdrant indexing, add payload filtering, and switch to Mistral-7B model for RAG pipeline

refactor: remove unused imports, standardize naming, and update Qdrant filters to use FieldCondition objects

feat: integrate local LLM pipeline and structured RAG retrieval for policies and products

refactor: remove redundant section headers, inline helper functions, and simplify retrieval and chat logic

refactor: optimize data ingestion with type metadata, improve retrieval filtering, and clean up codebase structure

refactor: implement UUID generation for ingestion and replace local transformer pipeline with placeholder LLM function

refactor: swap Flan-T5 for GPT-2 causal language model and clean up ingestion and UI code

refactor: clean up imports, update LLM to flan-t5-small, and refine Gradio UI labels

refactor: modernize data ingestion, update LLM pipeline implementation, and improve Qdrant configuration handling

refactor: migrate to local LLM inference using flan-t5 and remove HuggingFace API dependency

refactor: improve Qdrant initialization, enhance data ingestion, and add robust error handling to RAG pipeline

feat: implement automated Qdrant collection recreation and improve LLM response handling and UI presentation

refactor: migrate from langchain-community to langchain-qdrant for vector store integration

refactor: simplify code structure, refine system prompt, and migrate Gradio interface to Blocks API

feat: implement ingest_to_qdrant function to upload policy and product data to vector store

feat: add load_files utility to parse policy and product JSON data

Create app.py