https://github.com/adirthaborgohain/gpt-qa-chat
A Question Answering chatbot powered by GPT-3 and sentence-transformers that can answer questions based on your own data
https://github.com/adirthaborgohain/gpt-qa-chat
bot chatbot gpt-3 natural-language-processing nlp question-answering
Last synced: over 1 year ago
JSON representation
A Question Answering chatbot powered by GPT-3 and sentence-transformers that can answer questions based on your own data
- Host: GitHub
- URL: https://github.com/adirthaborgohain/gpt-qa-chat
- Owner: AdirthaBorgohain
- License: mit
- Created: 2023-02-23T15:48:23.000Z (over 3 years ago)
- Default Branch: main
- Last Pushed: 2023-02-27T05:38:06.000Z (over 3 years ago)
- Last Synced: 2025-02-05T13:18:11.004Z (over 1 year ago)
- Topics: bot, chatbot, gpt-3, natural-language-processing, nlp, question-answering
- Language: Jupyter Notebook
- Homepage: https://huggingface.co/spaces/adirtha1903/CapedConversations
- Size: 15.7 MB
- Stars: 1
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# GPT-QA-Chat
A Question Answering chatbot powered by GPT-3 answer synthesis and sentence-transformers sentence embeddings that can answer questions based on your own data. Here, we use a PDF file about superheroes and their evolution and influence on pop-culture. The knowledge from the PDF is fed into the bot so that the bot can later answer questions about it when asked.
Currently works with [FAISS](https://github.com/facebookresearch/faiss) but can easily be used with other vector DBs like qdrant, chroma, pinecone, etc for better scalability.
A notebook on how to build such a bot is given. Make sure all modules and libraries mentioned in `requirements.txt` is installed before starting! In this method, rather than using OpenAI api for generating embeddings for the text, we use open-source models which brings down the cost by a lot while maintaining the same level of quality. (Thanks to excellent models from [sentence-transformers](https://www.sbert.net)) Only during answer synthesis i.e. to generate the final answers we use the OpenAI api.
A Demo app can be found [here](https://huggingface.co/spaces/adirtha1903/CapedConversations)
## Screenshot of how the demo works:
