I'd like to share my open-source project "Incognito Pilot". It is similar to ChatGPT Code Interpreter, but the interpreter runs locally and it can use open-source models like Llama 2.
Would be great to hear your thoughts! Thanks a lot.
When ChatGPT code interpreter came out, I was amazed about it's capabilities. But at the same time, in my company, I often had the problem that I couldn't upload some sensitive data, so I couldn't use it for stuff where I would have been much faster than writing a small script.
That's why I built a tool which tries to solve this: It's not an easy trade-off between privacy and capabilities and I would be very interested in your judgement whether you think this is a useful solution.
How does it work?
- It uses OpenAI functions to connect the remote GPT-4 model with a local Python code interpreter
- The interpreter runs in a local docker container which has access to a folder with your (potentially sensitive) data
- There is a UI in which you approve inputs and outputs of the interpreter
The difficult tradeoff is not leaking your data, but allowing the model to know enough about it that it can work with it. For an image that's easy for example, but for a spreadsheet much harder.
If you have the time, I would be very interested in what you think about it.
Hi all, just wanted to share my new open-source project: It's basically ChatGPT Code Interpreter, but the Python interpreter runs locally on your machine.
The idea is that you can use it for sensitive data (or any data you don't want to upload to their services). There is an approval process to not run code you don't want to and not send results to OpenAI you find too sensitive.
After releasing a Python config management library (https://github.com/Zuehlke/ConfZ), I now wrote an article on how to use it together with FastAPI and SQLModel (all three tools are based on pydantic in the end) to build highly configurable and powerful APIs with ease.
Would be very interested in your thoughts / comments / ideas (or even pull requests ;) )!
I tried to build a python library for configuration management. It can automatically load all config (db credentials, API ports, crawler URLs, ...) you might have in your code from different sources like config files, environment variables or command line arguments. It then uses pydantic to parse and verify it. At last, it provides the values to the remaining part of your application through class attributes, using a singleton pattern with lazy loading.
Would be great to hear your thoughts about it! It's also already on PyPI and documented on readthedocs.
I'd like to share my open-source project "Incognito Pilot". It is similar to ChatGPT Code Interpreter, but the interpreter runs locally and it can use open-source models like Llama 2.
Would be great to hear your thoughts! Thanks a lot.