Participate in our State of AI Development Surveyfor a chance to win a MacBook M4 Pro!Take 4-min Survey →
SaaS
Customer Support Q&A

Create a bot that searches your documentation in a vector database and integrates with Zapier or Slack to answer customer questions.

Input Variables
user_question
how do I pass inputs into an API node?
Run the prompt to see the model response...

Perform a vector search in the "Help Desk" vector database based on a user's query across multiple data sources.

Summarize the output in a conversational format, and answer the user in the preferred channel (e.g. Slack).

Perform a vector search in the "Q&A bank" vector database that stores all previously asked user queries and answers.

It looks like you're asking about passing inputs into an API node. I recommend checking out our Help Doc on that topic here: [API Node Tips](https://docs.vellum.ai/help-center/workflows/node-types#api-node-tips). If you have any more questions, feel free to ask!

How it works
step
1
Search Context from previous questions

Perform a vector search in the "Q&A bank" vector database that stores all previously asked user queries and answers.

step
2
Search context

Perform a vector search in the "Help Desk" vector database based on a user's query across multiple data sources.

step
3
Summarize

Summarize the output in a conversational format, and answer the user in the preferred channel (e.g. Slack).

Tools used
Tools
Data Extraction
RAG
Integration
Memory
Parser

Customize this workflow

1/ Integrate multiple vector dbs

2/ Try different models

3/ Integrate business logic

4/ Perform evaluations

5/ Debug and test in production

talk with aI Expert
See what else you can build
View More

Experiment, Evaluate, Deploy, Repeat.

AI development doesn’t end once you've defined your system. Learn how Vellum helps you manage the entire AI development lifecycle.

Talk with us
const embed = document.getElementById('app-embed'); const loader = document.querySelector('.lottie-loader'); embed.addEventListener('load', function() { loader.style.display = 'none'; });