This project utilizes the GPT-3 model from OpenAI to create a customer support chatbot that can answer user's questions. It uses the text-embedding-ada-002 model for embedding and the gpt-3.5-turbo model for question answering. The text is indexed using FAISS for efficient retrieval.
Dataset The dataset used for training the customer support chatbot is the Customer Support on Twitter dataset. It provides a collection of customer support interactions on Twitter.
OpenAI API key: You will need an OpenAI API key to access the GPT-3 model and make API calls.
- company: The customer support ID on Twitter. You can find a list of customer support IDs in the
data/companies.csv
file. - gpt_model: The name of the GPT model to use. Refer to the OpenAI model documentation for available options. The default
- is gpt-3.5-turbo.
- embedding_model: The name of the embedding model to use. The default is
text-embedding-ada-002
. - data_path: The path to the data directory.
- index_path: The path to the pickle file that stores the FAISS index.
- openai_api_key: Your OpenAI API key.
- Install packages
pip install requiremetns.txt
- Export your OpenAI API key
export OPENAI_API_KEY='sk-...'
- Run GPT
python main.py --company AppleSupport --openai-api-key [OPENAI_API_KEY]
- Type your question in terminal
What do you want to ask the customer support? (Type 'stop' to quit)
Question: [Your question]