Show HN: Func Runner – Managed Function Calling for OpenAI Assistants(chat.funcrunner.com)
chat.funcrunner.com
Show HN: Func Runner – Managed Function Calling for OpenAI Assistants
https://chat.funcrunner.com
1 comments
It would be nice to have any information other than a signup / login form greet you. What the hell am I getting myself into?
Hi,HN link posts are confusing I know but read my initial discussion post. One of the first items there is the demo credentials
Username: hackernews Password: ilovehn
A Note On Demo Credentials
I have provided some credentials above for everyone to use and will keep the demo up and running as long as I fiscally can. I am not rich so if my Azure or OpenAI bill gets out of hand I will have to temporarily disable the demo. I wanted to make this easy for everyone to see so I am eating the cloud cost right now so please play nice :).
Why I Built Func Runner
I built Func Runner, which is a managed function calling provider for OpenAI Assistants that allows anyone to quickly build function calling enabled assistants using Python. Although function calling is not incredibly difficult to implement on your own, I felt that this pattern was something I had done frequently enough to warrant toying with the idea of providing a service built around it. My target audience would likely be from one of two camps. Camp A would be hobbyist Python programmers looking to build their own assistants who also lack the technical chops to build a full-fledged application. Camp B would potentially be professional developers looking to abstract away their assistant code execution so they can worry about other problems.
What Func Runner Does
In a nutshell, I built a system that deploys a dedicated serverless Durable Function app service to Microsoft Azure for each of your assistants. The Python functions you write are deployed as executable activities inside your app service. To enable this, I have effectively built a proxy service around the OpenAI API mixed with my own secret sauce to handle function executions on your behalf. The proxy is 100% compatible with all existing OpenAI libraries and requires minimal setup to adopt. Whenever I see a run execution with the state of “requires_action”, func runner steps in and orchestrates the execution and return of results back to your OpenAI assistant.
Key Features Right Now
- Custom Python code execution in a serverless environment dedicated to your assistant.
- Python dependency management and secret storage for sensitive keys.
- A ChatGPT chat experience powered by an OpenAI proxy service with specialized function calling handling.
Although this project is still in its infancy, it has been a great source of fun building it out over the past few months.
Please let me know what you think, ask any questions you may like and please excuse any bugs :)