How ibrahimbagislar/chefgpt works
ChefGPT combines recipe-data acquisition and preparation, model training, and a local FAISS/Ollama RAG assistant with a web interface. The documented data workflow runs from scraping through Q&A generation and training, while the runtime workflow retrieves recipe context before generating an answer. Some stages are documented but their implementations were not sampled; those relationships are cited to the README, and the web backend is not present in the repository.
Main entry points: Program.cs Recipe scraper · build_faiss.py FAISS index builder · chefgpt_server_faiss.py RAG server · QAGenerator.cs Q&A generator
Components 18
Recipe collection
- Recipe scraper
scraper/Program.cs - Recipe CSV
Data preparation
- Q&A generator
scraper/QAGenerator.cs - Q&A dataset
Model training
- Model fine-tuning
training/ChefGPTModelEgitim.ipynb - GGUF model
Local RAG runtime
- FAISS index builder
local/build_faiss.py - Sentence embeddings
local/build_faiss.py - FAISS index
- Recipe metadata
- RAG server
local/chefgpt_server_faiss.py - Recipe retrieval
local/chefgpt_server_faiss.py - Context prompting
local/chefgpt_server_faiss.py
Web interface
- Chat interface
website/app.js
Other components
- User
- Recipe source
- Ollama service
- Chat API backend
Connections 22
- User → Recipe scraper: requests recipes
README.md - Recipe scraper → Recipe source: fetches pages
scraper/Program.cs - Recipe scraper → Recipe CSV: writes recipes
README.md - Recipe scraper → Q&A generator: invokes Q&A mode
scraper/Program.cs - Q&A generator → Recipe CSV: reads recipes
README.md - Q&A generator → Q&A dataset: generates pairs
README.md - Q&A dataset → Model fine-tuning: trains on data
README.md - Model fine-tuning → GGUF model: exports model
README.md - GGUF model → Ollama service: runs locally
README.md - Recipe CSV → FAISS index builder: provides recipes
README.md - FAISS index builder → Sentence embeddings: encodes recipes
local/build_faiss.py - FAISS index builder → FAISS index: writes vectors
local/build_faiss.py - FAISS index builder → Recipe metadata: writes metadata
local/build_faiss.py - FAISS index → RAG server: loads index
local/chefgpt_server_faiss.py - Recipe metadata → RAG server: loads metadata
local/chefgpt_server_faiss.py - RAG server → Recipe retrieval: runs search
local/chefgpt_server_faiss.py - Recipe retrieval → Context prompting: builds context
local/chefgpt_server_faiss.py - Context prompting → Ollama service: requests completion
local/chefgpt_server_faiss.py - User → Chat interface: asks question
README.md - Chat interface → Chat API backend: posts chat
website/app.js - Chat API backend → RAG server: forwards request
- Chat API backend → Chat interface: streams response
website/app.js