askymyfiles Python App

This app creates a local database of the current directory and utilizes it in conjunction with ChatGPT for answering questions.

Data is stored in a ChromaDB database in .vectordatadb

Install

You need an OpenAI API account. You figure that out and get a key. Make sure the environment variable OPENAI_API_KEY is set.

# You've got python3 installed, right?
# Clone this repo somewhere convenient. git pull if you want new stuff
cd ~/Code/ && git clone https://github.com/codeprimate/askymyfiles.git
cd askmyfiles
pip install -r requirements.txt

# Make it easy to use.
ln -sf /path/to/askmyfiles.py ~/bin/askmyfiles
chmod u+x ~/bin/askmyfiles

Usage

Always run askmyfiles at the root of your project folder!!!

askmyfiles looks for new and updated information.

Add a list of files or directories to ignore in .askignore (Like a .gitignore). Add a list of hints/instructions for chat in .askmyfileshints

To update or add a directory, file, or single webpage to the local database, then "add":

~/bin/askymyfiles add the/path/to/file.txt
~/bin/askymyfiles add /really/the/path/to/file.txt
~/bin/askymyfiles add the/path/
~/bin/askymyfiles add /really/the/path/
~/bin/askymyfiles add https://www.example.com/file.html

# if there is a failure/interruption touch the file and retry
touch the/path/to/file.txt
~/bin/askymyfiles add the/path/to/file.txt

To ask a question using the gpt-3.5-turbo-16k model, then "ask":

~/bin/askmyfiles ask "Your question."
~/bin/askmyfiles "Your question"

You can also pipe queries to askmyfiles:

echo "What is the name of my app" | ~/bin/askmyfiles

Note

• Results are not perfect or deterministic.
• Hallucinations can and will occur so ask the question more than once
• Do some prompt engineering as needed to tease the information you want out of your data.

To list all entries in the database, then "list":

~/bin/askmyfiles list

To remove a file or URL from the database, then "remove":

# Specify a single resource to remove 
# Also useful when adding fails
~/bin/askymyfiles remove the/path/to/file.txt
~/bin/askymyfiles remove /really/the/path/to/file.txt
~/bin/askymyfiles remove the/path
~/bin/askymyfiles remove "https://www.example.com/example.html"

Once the file is loaded into the database, you don't NEED it in the project directory anymore.

Back it Up

tar cvf my_db.tgz .vectordatadb

All file paths are relative. Drop your database anywhere.

TODO

• oobabooga integration!
• GitHub Issues integration (suggest root cause or solutions)