#Human-in-the-loop
1 messages ยท Page 1 of 1 (latest)
Ok this thread is ruined, let me make another one ๐ญ
regarding an agent that generated files (like a test) it would be awesome for it to prompt you "These tests pass, does this look good to you? Should I place the new test in your local repo? kind of flow
(yeah, there is no emoji that can accurately describe that image @marble zodiac)
Here's core thinking: if we could somehow add the Dagger Cloud a human feedback system (eg. ๐๐) that enriches the otel traces that could be very very very powerful, because it opens the possibility of compounding improvements
is the idea for the enginee to ingest that and auto-optimize? or could we also use it to provide feedback to the end user on code or prompt changes they should make?
HIL from dagger cloud is a really good idea but how my app will be served continuously in pipeline like is it possible now in dagger?
is it related to agent reflections?
not sure. but this would set us up to be able to deliver both over time. All the way to fine-tuning models for you one day
I don't know exactly, let's find out together ๐
Native HITL/HIL support in Dagger could enable agent building and orchestration without external frameworks like LangGraph.
Just a thought...
And yes dagger cloud can help in this becuase in other ci system like github action or anywhere we wont be able to achive this easily...