Hey there! We're Riza and we make it safe and easy to run untrusted code generated by LLMs or humans, in our cloud or on your infra.
We're excited to mark some long-standing items off of our product roadmap!
The Riza Code Interpreter API now accepts input files whereas previously you had to shunt all input into stdin. You can tell the API to make files available to your code at whatever paths you want, and Riza writes the file contents to those paths before execution.
This makes it a lot easier to have code operate on multiple inputs. See the files documentation for details.
The Riza Code Interpreter API also now allows you to configure credentials that we attach to outbound HTTP requests outside of the code execution environment, so you can authenticate HTTP API requests from untrusted code without sharing this sensitive data directly. Initially we support the HTTP Basic and Bearer authentication schemes, but we’ll be adding more.
See the HTTP documentation for details.
And just in case you haven’t heard, you can run the Riza Code Interpreter API yourself via our docker container. Get started with the Riza self-hosted quickstart.