AI engineeringCompleted
E-commerce Customer-Service RAG Q&A System
RAG-powered intelligent Q&A for e-commerce customer service, FastAPI + Vue 3 + Docker
- Period
- Jan 2025 — May 2025
PythonFastAPIVue 3RAGLLMDocker
What I did
- Full RAG pipeline: multi-format parsing → chunking → embedding → retrieval → generation
- Dual backend — local open models or cloud APIs — switched by config
- Answers carry citations, checkable and traceable; hallucination risk stays low
- Docker-packaged delivery, one command to start
Background
E-commerce customer service faces high-volume repetitive queries during peak periods; manual replies are costly and inconsistent. RAG lets LLMs answer based on the company’s own knowledge base, combining natural-language understanding with accurate, traceable answers.
Contributions
- Implemented multi-format knowledge document parsing (PDF / Word / web pages) and text chunking
- Built the complete RAG pipeline with vectorization and similarity retrieval
- Supported dual modes: local open-source models (e.g., Ollama) and cloud APIs
- Built the Q&A interface with Vue 3, including citation/source tracing
- Containerized the whole stack with Docker for one-click deployment
Highlights
- End-to-end RAG pipeline: parse → chunk → embed → retrieve → generate
- Local/cloud model compatibility adapts to different compute and budget scenarios
- Dockerized deployment for consistent environments and simple ops