Skip to main content

​
Architecture

AtlasFlux acts as a unified gateway between your application and AI models.
  1. You send a request to https://api.atlasflux.my
  2. AtlasFlux classifies the request type and complexity
  3. The router selects the optimal upstream model
  4. The request is forwarded to the provider
  5. The response is normalized and returned to you

​
What AtlasFlux handles

  • Model selection — Automatically picks the best model for your task
  • Fallback — If a model fails, the request is retried with another model
  • Pricing — Unified pricing across all models in MYR
  • Web search — Built-in web search with configurable depth
  • Usage tracking — Every request is logged with token counts and cost
  • Wallet billing — Prepaid balance deducted per request

​
Public model

All users access a single model identifier:
You do not need to specify model names. AtlasFlux handles routing internally.

​
Request flow

  1. Authentication — Verify your API key
  2. Validation — Validate request body and parameters
  3. Reservation — Reserve estimated cost from your wallet
  4. Routing — Classify request and select model
  5. Execution — Request is processed by the selected model
  6. Settlement — Deduct actual cost and release unused reservation
  7. Response — Return normalized response with usage data