factsChat

factsChat lets you ask questions about content provided by you. It uses the OpenAI API.

Requirements

You need at least Python 3.11 installed on your system and virtualenv. To use this tool you'll need an OpenAI API-Key: https://platform.openai.com/account/api-keys

Installation

git clone https://github.com/maczarr/factschat.git
cd factschat
virtualenv ./venv
source venv/bin/activate
pip install -r requirements.txt

For some installation advices on macOS go to the end of this page

Configuration

Copy the .env.example and adjust the content: cp .env.example .env

name	description
OPENAI_API_KEY	This is the only required entry. Get your API-Key at https://platform.openai.com/account/api-keys
OPENAI_MODEL	Name the model you want to use and have access to. Default is gpt-3.5-turbo
OPENAI_TEMPERATURE	Controls how deterministic the answer are. Has to be a floating number, default is 0.0.
ANSWER_LANGUAGE	Tries to fix the language of the answer, despite the given context. It's more a hack in the system prompt, but works in most cases.
DB_FOLDER	Folder where the FAISS databases are stored. Default is db.
CONTENT_FOLDER	Folder from where the files are being read. Default is content.
URLS_FILENAME	Name of the text file from where to read in URLs if you want to make a database of website content. Default is urls.txt.

Usage

Assuming you are in the factschat folder and have the virtualenv activated.

Retrieve data...

...from files like a PDF

Put your PDF in the content folder. Afterwards:

python create_embeddings.py

Type 2 and press enter. In the end provide a database name and you're ready.

...from websites

Create a urls.txt file in the root directory of the repository. Put one URL per line in the file, nothing more. It can be as much from one line to dozens. Afterwards:

python create_embeddings.py

Type 1 and press enter. In the end provide a database name and you're ready.

Ask questions

python chat.py

At first you have to choose to which database you want to chat. Choose by answering with the number in front of the file name (e.g. 1).

Now you can state your question. Depending on how busy the API is at the moment, the answer may take some time.

To leave the conversation, simply type "exit" or "bye".

Known issues

Sometimes there are no sources provided with the answer. There is an open issue on the langchain project.

Disclaimer

This repository isn't "a product", it's just a bit of fiddling around with the API and try out what's possible with LLMs.

There are no tests for the software and not much failsafes, it will likely fall apart if it's not used as intended.

Installation advices for macOS

The default Python version on macOS is a bit old. You can install a newer one with Homebrew:

brew install python
# open a new tab and check the version
python3 --version

Install virtualenv:

python3 -m pip install --user virtualenv

If your terminal tells you there is no virtualenv chances are high your Python3 bin-folder isn't in the PATH. Add an entry at the very end of your .zshrc:

export PATH="/Users/YOUR_USERNAME/Library/Python/3.11/bin:$PATH"

(the path to your Python3 bin-folder may be different)

Error message: libmagic is missing

If you'll get an error message saying libmagic is missing when running create_embeddings.py the follwing steps helped me:

brew install libmagic

and in the activated venv:

pip install python-magic