How sunnysavita10/custmor_support_system works
Project type and purpose
• The backend is a Python web server (FastAPI run via Uvicorn), the frontend is an HTML/CSS chat UI, and there are modules for data ingestion, retrieval, prompt engineering, and model loading.
File structure
Top-level:
• Dockerfile, setup.py, requirements.txt, GitHub workflows → packaging, containerization, CI/CD
• main.py → entry point for the web server (FastAPI/Uvicorn)
• templates/chat.html + static/style.css → Jinja2-style chat UI
• data_ingestion/ingestion_pipeline.py → batch pipeline to import and preprocess flipkart_product_review.csv
• retriever/retrieval.py → code to lookup or search preprocessed data (e.g. similarity search)
• prompt_library/prompt.py → functions to assemble LLM prompts
• utils/config_loader.py, utils/model_loader.py → load YAML configs and initialize the LLM
• config/config.yaml → configuration (e.g. API keys, model names, database URIs)
• notebook/ → exploratory Jupyter notebook for prototyping
• test.py → smoke tests or simple end-to-end checks
README for additional insights
• Shows how to create/activate a conda environment and start the server:
uvicorn main:app --reload --port 8001
• Mentions CI/CD is configured (via GitHub Actions).