When will asynchronous support be available

[sue2xlh]

HI, Thank you for your work. I would like to inquire about when will asynchronous support be available

[yyiilluu]

thank you for your interest in this. this can be up for debate. the original purpose of autochain is creating something simple that people have very easy time to go through and experiment. adding async request would be more towards productionizing this repo, which could make custom exploration more challenging

[hooman-bayer]

@yyiilluu This is a great library and props to you for releasing it. Here are my two cents on this. I would like to challenge what you mentioned. If the idea is only for exploration:

1. I would personally say LangChain Agents are already good enough for that purpose. LangChain agents provide both sync and async options which provides the flexibility to some extent. Now I understand that AutoChain is easier to get started with and add long term memory
2. Agents take a long time to run, both async and streaming options enable continous feedback from each tool which is essential.
3. What is the purpose of Agent evaluation in this library then? It more resembles unit tests that are often helpful for production settings.
4. I think it is extremely easy to build Agent prototypes that do not support async and streaming (here is a repo creating one in around 100 lines of code). I think the trick is how to make a simple modular library for agents that support async and streaming.