n8n talking to vapi · AI Automation Society

n8n talking to vapi

been doing a lot of work with vapi lately. its a good for extracting away a lot of the voice orchestration but ive never cared much for the rest of its black box stuff. we almost everytime pass the conversation through vapis custom llm option and handle the chat completions ourselves. the accuracy/reliability is much better that way. previously we had to use a flask server to maintain an open http connection to consume the sse sent through vapi but ive started messing around with n8n as the backend. we use the server events sent by vapi to trigger a webhook instead of the actual sse. ive spun up a poc. all it does so far is launch a transient assistant for an inbound call, grabs the control url and then tells vapi to say something programatically. next step is to wire up an n8n agent and send its output to vapi. but conceptually, this first part was the most problematic. once i get the agent wired up, i will release that, too. if anybody is interested, i can also do a walk through of what weve done so far

0 comments