Understand VisibleBase
Build the one-Base-many-clients mental model before connecting real products.
The concepts section is not extra theory. It exists to keep the VisibleBase boundary clear before you wire it into real products.
If you start by treating VisibleBase as “just another backend framework” or “a generic provider gateway,” the service boundary, product split, and trust model will all drift in the wrong direction.
Start with this picture
product_id + user_token + UserClientRecommended reading order
Understand why the right shape is “one Base, many products,” and what belongs in Base versus the product layer.
Step 2ArchitectureFollow a request from the client to Base and then to the real provider.
Step 3Token modelSeparate admin_secret_key from user_token, and keep management credentials out of the frontend.
Know which capabilities should stay in trusted backend code and which ones can safely reach product clients.
Step 5Usage and billingConnect quota, metering, billing, and hooks to your own business system instead of forcing business logic into Base.
When this section is the right starting point
- you want multiple AI products to share the same AI infrastructure
- you need to understand the role of
product_id,user_token, and the model directory - you are still deciding where Base, your business backend, and an AI Gateway should stop
If you only want one working call chain first, go straight to Quickstart.