As discussed in this video https://youtu.be/_IK18goX4X8?t=719 a common approach while you are tuning your specs/prompts is to do "once" style approach where you do a single Ralph loop in an interactive session. This would help you ensure that your Ralph loops are on the right track. It would be nice to have support added for this.
Next, running the Ralph loop in a Docker sandbox should make it safer as it runs in an isolated environment. It would be nice to have instructions and support for this.
Let me know if you'd like me to open some PRs.
Thanks!
As discussed in this video https://youtu.be/_IK18goX4X8?t=719 a common approach while you are tuning your specs/prompts is to do "once" style approach where you do a single Ralph loop in an interactive session. This would help you ensure that your Ralph loops are on the right track. It would be nice to have support added for this.
Next, running the Ralph loop in a Docker sandbox should make it safer as it runs in an isolated environment. It would be nice to have instructions and support for this.
Let me know if you'd like me to open some PRs.
Thanks!