https://github.com/mlojek/booktalk
Interact with books like never before with this AI-powered chatbot!
https://github.com/mlojek/booktalk
books large-language-models machine-learning retreival-augmented-generation
Last synced: about 1 month ago
JSON representation
Interact with books like never before with this AI-powered chatbot!
- Host: GitHub
- URL: https://github.com/mlojek/booktalk
- Owner: mlojek
- License: gpl-3.0
- Created: 2025-03-25T09:03:39.000Z (7 months ago)
- Default Branch: main
- Last Pushed: 2025-08-26T18:21:36.000Z (about 1 month ago)
- Last Synced: 2025-08-27T01:56:09.894Z (about 1 month ago)
- Topics: books, large-language-models, machine-learning, retreival-augmented-generation
- Language: Python
- Homepage:
- Size: 37.1 KB
- Stars: 0
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
Awesome Lists containing this project
README
# Booktalk
Interact with books like never before with this AI-powered chatbot!
I'm not a humanist, so back in high school while I got lots of As in maths and computer science, I also got a lot of Fs in polish. Main reason? I didn't read the readings. They just bored me. Not that I don't read any books - just not some archaic novels clearly written by mentally unwell people.
Now, back then LLMs were not a thing yet, but If I had to learn for my literature exams now, I'd surely use AI to help me out. Specifically, I'd make a chatbot with whom I could talk about the reading, and who would help me understand the reading on a deeper level.
So guess what? I made one!
## Tech stack
This project is made using `python`. I used `Ollama` to host LLMs locally so that I don't have to pay for OpenAI's API. Vector storage for RAG is done using `ChromaDB`. The project's logic is done in `LangChain`. I've also added an elegant and user-friendly web GUI with `Streamlit`.
## Installation
This project requires you to have [Ollama installed and running](https://ollama.com/download) on your system. Use python 3.11 or newer. Install the project by running:
```
make install
```
Then to run the app run:
```
booktalk
```
## Performance
To test if the RAG mechanism works propertly, I asked the same model with and without RAG a context-related question. Both models were asked "What is this book about?". The model without book fragments provided should not be able to answer the question. I've supplied the bot with "Moby Dick" by Herman Melville. The results have proven that the RAG mechanism works:
```
User: What is this book about?
Bare LLM: I'd be delighted to help! Based on my thorough review of the book, I can tell you that "The Book" (no title provided, I assume it's an unnamed work) appears to be a thought-provoking and deeply personal exploration of... As they traverse the landscape of America...
RAG LLM: This book appears to be an excerpt from Herman Melville's "Moby-Dick". The text reveals the author's fascination with the mythology of whales and the symbolism surrounding them. He explores themes such as...
```
## References
- Get free public domain EPUBs from Project Gutenberg [here](https://www.gutenberg.org).
- Streamlit has a great documentation with lots of practical examples [here](https://streamlit.io/playground).
- You can find a great tutorial video from Tech with Tim on youtube [here](https://www.youtube.com/watch?v=E4l91XKQSgw).