bright needle
Hi team, quick architecture question about WhatsApp reactions.
Right now we have a custom WhatsApp flow that receives messages, stores them, analyzes them, creates metadata, and then sends reactions. For reactions, we’re currently reaching into OpenClaw internals by dynamically loading the dist export that exposes sendReactionWhatsApp, instead of going through a cleaner public surface.
What we’d like to do is structure this as separate steps:
receive and persist the incoming message
analyze it and attach metadata
send the reaction from that metadata later
Question: what’s the best supported way to send a WhatsApp reaction from OpenClaw for that kind of staged pipeline?
Specifically:
Is there a public internal API we should be using instead of dynamically loading library exports from dist?
Is there a CLI-supported way to send WhatsApp reactions that would be safer/more stable for deferred processing?
If the recommended path is hook-based, what is the cleanest way to split ingest, analysis, and reaction into separate phases without relying on dynamic imports?
We want to use the already-configured OpenClaw models and stay on supported surfaces as much as possible, rather than manually wiring internal exports and making curl commands to LLM api servers
Thanks.
hybrid flicker