Skip to main content
The AI layer in KilatKoding provides a base for chat and generate flows. That base is enough to start from, but the final AI experience still depends heavily on the product logic you build on top of it.

​
What this feature includes

  • POST /api/ai/chat
  • POST /api/ai/generate
  • provider selection
  • usage tracking
  • basic usage and plan limits
KilatKoding provides the core AI endpoints, but the screen route where users interact with AI is usually something you decide during product customization. In many implementations, these endpoints are called from a dashboard area or a custom app screen.

​
Main dependencies

​
What teams usually customize

  • prompt design,
  • chat or generation UX,
  • output formatting,
  • plan quota and limits,
  • error handling for 429 and 503,
  • business guardrails and moderation logic.

​
When this feature is disabled

It is usually disabled when:
  • AI is not yet ready as a product layer,
  • AI is only a roadmap idea and not mature enough,
  • the team is not ready to manage provider costs and provider failure states.

​
Verification checklist

  1. The selected AI provider is really configured.
  2. A test user can use chat or generate routes.
  3. Usage limits and plan messaging are consistent.
  4. The team understands 429 and 503 error behavior.
  5. Pricing and plan benefits do not overpromise compared with real limits.
If AI is core to your product, also review Billing and payments and Operational runbook.