Knowledge Base in Local Machine
# supportc
Cerebrain
02/19/2025, 11:30 PMI am integrating and demoing some voice enabled solutions to clients so far with much success, even though I am not highly technical I have managed to work my way connecting Vapi with Google Calendar, Cal, Make, etc. However I am facing a technical challenge well above my weight: An important prospect loved the solution however he strictly needs, for policy reasons, the Knowledge Base to reside in a local machine. Is there a way to do that? if so how? is there any documentation I could refer to? thanks!!
v
Vapi
02/20/2025, 10:42 PMYou'll need to set up a local server that can receive and respond to knowledge base requests.
VAPI will send requests to your server with the conversation context
Your server needs to return either:
a) Relevant documents with similarity scores
b) A direct response for the assistant to speak
Vapi
02/20/2025, 10:44 PMImplementation Steps:
a. Create a local HTTP server that accepts POST requests at an endpoint
b. Configure the server to expect requests in this format:
Copy code
{
"message": {
"type": "knowledge-base-request",
"messages": [
{
"role": "user",
"content": "What are your business hours?"
}
]
}
} Copy code
{
"documents": [
{
"content": "Our business hours are 9am-5pm Monday through Friday",
"similarity": 0.95
}
]
} Copy code
{
"message": {
"role": "assistant",
"content": "Your direct response here"
}
}Vapi
02/20/2025, 10:44 PM Copy code
{
"model": {
"knowledgeBase": {
"provider": "custom-knowledge-base",
"server": {
"url": "https://your-local-server.com/knowledge-base"
}
}
}
}j
JustSeth
02/22/2025, 1:35 AMCouldn’t this be made with a ChatGPT module in make.com and webhook module?
JustSeth
02/22/2025, 1:36 AMping when u respond ❤️
s
Shubham Bajaj
02/22/2025, 6:42 PMI don't know how it works. so cannot make a comment on that part.
j
JustSeth
02/22/2025, 8:03 PMOk