wch / doc_query

Documentation query app

Documentation query app

This is a Shiny for Python application which allows you to ask questions about a set of documents. It uses the chatstream package to query the OpenAI API.

First, install some Python packages:

pip install -r requirements.txt

Then load at least one set of text/markdown/PDF files into a vector database, using doc_db.py script. This will save the documents in a database in a directory named doc_db/.

The syntax is:

./doc_db.py add-dir <collection name> <directory> <file pattern>

For example:

./doc_db.py add-dir admin-guide ../docs/admin "*.qmd"
./doc_db.py add-dir user-guide ../docs/user "*.qmd"

After the documents are loaded into the vector database, they can be queried using the Shiny app.

To use the Shiny app, you need an API key from OpenAI. After you have it, you can set an environment variable named OPENAI_API_KEY:

export OPENAI_API_KEY="<your_openai_api_key>"

Then you can run the app with:

shiny run app.py --launch-browser