My harsh feedback would be that the way to communicate what your library does is almost entirely defensive and aims to justify its existence but without telling you in basic terms what it does. It feels like it was written off the back of 100 conversations with someone who said "oh okay, but can't you just [...]".
Humans tend to read in a Z pattern, so I've read 3 blocks of mostly irrelevant text that should be buried in an FAQ somewhere before you actually introduce your core primitives.
I think if you took a few hours to do the following and resubmit you'll get a much warmer reception:
1. Drastically reduce the text. Action Engine provides a set of utilities to help developers build multimodal, streaming, stateful, long-running, agentic applications. Actions are bla bla, async nodes are bla bla. Save the philosophy for the docs.
2. Include a middle step before you jump straight from a oneliner to working program in RestOfTheFuckingOwl style. Here's how these abstractions can be combined in simple way to do surprisingly complex thing.
3. Add a complete, highly intuitive code example or if not possible, a neat diagram. Bonus points for comparing to an existing popular framework you want to replace.
I think all of it makes sense—I've been putting off writing the steadier, more easy going walkthroughs, but they should really make a difference.
helenapankov•3mo ago
It's not something currently backed broadly, but a flexible, quick-feedback way to navigate the landscape of AI infrastructure demands—an experiment from me and early internal supporters at Google DeepMind.
Current frameworks solidified around abstractions that are growing increasingly inadequate for the new demands of multimodal, streaming, stateful, long-running applications. Action Engine aims to be a common building block for these new kinds of applications, without imposing rigid abstractions or heavy dependencies, adapting to applications.