In a very general sense, this isn't different from any other open vs walled garden debate: the hackable, open project will always have more functionality at the cost of configuration and ease of use; the pretty walled garden will always be easier to use and probably be better at its smaller scope, at the cost of flexibility, customizability, and transparency.

Yep, if you look carefully a lot of the demos don't actually work because the LLM hallucinates tool answers and the framework is not hardened against this.

In general there is not a thoughtful distinction between "control plane" and "data plane".

On the other hand, tons of useful "parts" and ideas in there, so still useful.

Yeah I primarily like Langchain as an aggregator of stuff, so I can keep up with literature

I had the exact same impression. Is anyone working on similar projects and planning to open source it soon? If not, I'm gonna start building one myself.

Yeah I wrote my own plunkylib (which I don't have great docs for yet) which is more about having the LLM and prompts in (nestable) yaml/txt rather than how so many people hard code those in their source. I do like some of the features in langchain, but it doesn't really fit my coding style.

Pretty sure there will be a thousand great libraries for this soon.

Same impression here. Rolling my own to learn more in the process.

I've found it extremely useful but also you are not wrong at all. It feels like it wants to do too much and the API is not intuitive at all. Also I've found out the docs are already outdated (at least for LangChainJS). Any good alternatives? Especially interested in JS libs.