How amitmanchella/realestatellm works
Project type and purpose
• This is a local LLM-based real-estate document analysis tool and chatbot, with an accompanying browser extension.
• It ingests PDF guides into a vector store, generates embeddings, provides an interactive CLI/chatbot, supports fine-tuning, and exposes an API for the extension.
File structure
Top-level directories and files:
• RealEstateLLM-Extension/ ← Browser extension UI & background scripts
• api/ ← HTTP server (likely Flask or FastAPI) exposing search/chat endpoints
• data/ ← Raw real-estate PDF documents
• data_processors.py ← Preprocessing pipeline for text extraction
• enhanced_rag.py ← Retrieval-Augmented-Generation orchestration
• fine_tuning.py ← Prepares data for LLM fine-tuning
• main.py ← CLI entrypoint (process, embeddings, chat, finetune)
• real_estate_chatbot.py ← Chatbot logic and loop implementation
• requirements.txt ← Python dependencies
README for additional insights
• Workflow steps: venv setup → document processing → embedding generation → chat launch → optional fine-tuning.
• The extension likely calls the API server (api/server.py) to deliver answers inside the browser.
Detail and coverage
• Two workflows: (1) preprocessing→embedding→chat, (2) preprocessing→embedding→finetune.
• That both CLI and extension share the same API/RAG layer.