Snowflake
Our application is available on Snowflake Marketplace.
Here some technical and guide details to help you to use us through Snowflake.
Technical requirements
Large language model
Currently we use these models :
Claude-3-7-sonnet
openai-gpt-4.1
snowflake-artic-embed-1-v2.0
You need to activate these models before installing KAI app.
Warehouse
We use multi-cluster warehouses to optimize speed of our application. To activate it, you need to have an enterprise edition snowflake account.
External Access Integration
When you deploy KAI app, before activate it, you need to add a reference to an external access integration. On the references part of the app, accept the creation of an external access integration.
This external access integration will allow to the KAI app to call your knowledge base through API.
Deploy KAI app
Grant all privileges and references
You should grant all privileges and references, here why we asking these privileges and references :
BIND SERVICE ENDPOINT : Allow us to expose our API (Ingress)
CREATE COMPUTE POOL : Allow us to create machines to host our application
CREATE WAREHOUSE : Need to generate related datas (semantic graphs, audit, ...) for each KAI instances you create.
CREATE DATABASE : All datas generated will be stored into a specific database on your Snowflake account.
IMPORTED PRIVILEGES ON SNOWFLAKE DB: Allow us to use Cortex API (LLM call).
(USAGE) EXTERNAL ACCESS INTEGRATION: Allow us to call your knowledge base once you configured it (egress).
Activate the application
When you click on activate the KAI application, the deployment script will :
create a new database "KAI_APP"
create a warehouse
create 2 compute pools
deploy the admin UI interface
Once the admin UI interface is deployed, you can go on it and deploy your first KAI Instance.
Deploy your first instance
Create an KAI Instance
Create a new KAI instance on your admin UI interface
Connect a knowledge base to your KAI instance
Generally, steps to connect a knowledge base to KAI app is :
prepare credentials needed by KAI to connect the knowledge base. Take a look here to know how to retrieve these credentials following the type of knowledge base you want to connect.
Fill these credentials into KAI admin UI when you configure the knowledge base.
(Optional) Fill the search goal, it will help to contextualize a bit more about your use case.
Wait 10s and validate it on the KAI admin UI (waiting 10s is important to be sure the propagation of updated hosts on the EAI is effective).
Choose scenario
Following what you want to do on your instance, on "features" tab of your instance, choose "Search" or "Audit" or both.
search = will allow you to use search endpoints
audit = will allow you to use audit endpoints
Deploy the KAI Instance
On the tab "Instance detail", click on "launch instance"
Last updated