Answers come back checked
Every AI response is tested against your schema before your code ever sees it.
Wow your usersProduct intelligence
Built-in AI
Add AI features without babysitting messy model output.
You get one simple way to call AI from your server. Answers come back checked and typed, never as a messy string that crashes a feature. Swapping models is one config line, not a repo-wide hunt.
3AI providers ready to use1config file to swap models0API keys in the browser
1 of 6 in Wow your users
Interactive example
See it working
This walkthrough demonstrates the implemented backend contract without using an account, credentials, or provider calls.
Interactive example
Explore the implemented decisions
Low-latency model route
Deterministic sample data only. No account, provider, or network connection is used by the interaction.
Highlights
Why it matters
Keys stay on the server
Provider API keys live in server settings and never reach the browser.
Swap models in one line
One config file picks the provider and model, so upgrades never touch your features.
The problem this solves
Without a shared foundation, every team rebuilds these edge cases — differently.
One bad response takes down a feature
The model returns broken JSON at the worst moment. Your code trusted it. Now a customer-facing feature is crashing. Here, bad output fails loudly at the boundary, not in your UI.
A key drifts into the browser
AI calls get sprinkled wherever it's convenient. One lands client-side, and your API key ships to every visitor. Here, every call runs on the server. Full stop.
A model upgrade touches every file
Your provider retires a model. Suddenly you're doing a find-and-replace across the whole codebase. Here, it's one line in one config file.
How it works
The mechanism end to end, as implemented in the repository.
- 1
You call it with a feature key
Your server code asks for text or a structured object, tagged with a feature key. One config file decides which provider and model handle it.
- 2
The call runs on the server
The matching OpenAI, Gemini, or Anthropic adapter runs with server credentials. Nothing AI-related ships to the browser.
- 3
The answer is checked at the door
The response is parsed against your schema before your code sees it. Broken output fails loudly instead of leaking through.
- 4
You get a typed result
Your code receives a clean object plus token usage, provider, and model. Never a raw string to re-parse.
Engineering guarantees
What this does and does not promise
Safety boundary
Three providers come wired in, but the kit won't auto-switch if one goes down. Live generation still needs your own provider key.
Implementation evidence
The boundary above is code, not a claim. These files carry the contract:
src/features/ai/runtime.tssrc/features/ai/contracts.tsconfig/ai.ts
Questions
Frequently asked questions
Related capabilities
These capabilities share contracts and boundaries with this one.
Build on the built-in ai foundation
The docs describe the same contracts this page demonstrates.