How yonkosama/llmops_series-sunny-savita works
Project type and purpose
• From the README and file tree, this is a full-stack Retrieval-Augmented Generation (RAG) chat application built with FastAPI.
• Purpose: Let users upload documents, build per-session FAISS vector indexes, then chat with those documents via an LLM.
File structure
• Top-level scripts and config:
• main.py – FastAPI app entry point
• Dockerfile, docker-compose.jenkins.yml, Jenkinsfile.test – containerization & CI/CD
• .github/workflows – GitHub Actions for AWS deployment and CI
• Core application under multi_doc_chat/:
• config (config.yaml) & utils (config_loader, file_io, document_ops, model_loader)
• exception, logger modules
• model/models.py – data models (Pydantic, FAISS index metadata)
• prompts/prompt_library.py – prompt definitions
• src/document_ingestion – ingestion pipeline (file split, embedding, FAISS index creation)
• src/document_chat – retrieval and chat logic (load index, perform similarity search, call LLM)
• Frontend assets:
• templates/index.html
• static/styles.css
• Evaluations & notebooks:
• run_evaluations.py, notebook/*.ipynb – LangSmith evaluation harness
• Tests: unit and integration under tests/
README for additional insights
• Endpoints: /upload, /chat, /health, / (UI)
• Storage: data/<session_id>/ for raw splits, faiss_index/<session_id>/ for vector stores
• Session management: in-memory per-session chat history
• External dependencies: OpenAI (embeddings, chat) via ModelLoader, LangSmith/Google APIs for evaluation