Release 1.5

Changelog

Release 1.5

Features

• Package restructure — flattened main.py split into core/ package (config_loader.py, ingestion.py, rag_engine.py, logger.py)
• Streamlit frontend — new app.py with session management and chat UI
• Hybrid search — semantic retrieval (ChromaDB + MMR) combined with BM25 keyword search via weighted fusion (configurable weights in config JSON)
• Config-driven architecture — config/default_config.json and config/apple_mlx_config.json control all runtime parameters
• Conversation history — full chat history passed to RAG engine with an improved prompt that references prior turns
• Heading-aware PDF ingestion — switched to fitz/PyMuPDF for robust, structure-preserving chunking

Changes

• PDF parser: unstructured → fitz (PyMuPDF)
• Dockerfile & docker-compose rewritten for the new module structure
• run.sh ported to macOS compatibility
• README architecture diagram: static PNG → interactive Mermaid flowchart
• Search strategy: single retriever (k=2) → hybrid ensemble (top_k=6, MMR with fetch_k=18)

Removals

• main.py (replaced by app.py + core/)
• media/wiki/Flowchart_v1.3.png
• media/HRBuddy-Flowchart.png
• unstructured dependency

Infrastructure

• .gitignore: added mongo_data/ and .DS_Store
• requirements.txt: added rank_bm25

Release 1.4

Changelog:

Added Docker Support - rewritten the complete code to support running as docker containers.

Release 1.3

Release 1.3:

Changelog:

1. Added password support for login.
2. Added streamlit web app interface

Release 1.1

Release 1.1

Changelog:

• Added memory storage for the model with MongoDB with multi user support for seperating conversations.

Release 1.0

Initial Release 1.0

Features:
Basic Llama3.2 + Nomic combination for a Retrieval-augmented generation (RAG) functionality based HR Chatbot.