Resources · Architecture

OpenQCore Runtime Architecture

System-level architecture for provider orchestration, routing control, contract normalization, and production reliability.

Runtime Architecture

A layered execution model from request ingress to deterministic response and cleanup.

API Gateway

Auth, scope validation, request shaping, and trace initialization.

Decision Router

Intent + modality classification with policy-aware route selection.

Execution Engine

Chat/Voice/Iris execution with streaming and non-stream handling.

Safety Controls

Rate limits, distributed locks, circuit breakers, and overload protection.

Execution flow

Persistence Layer

Session messages, dedup cache, and usage/event logging.

Provider Connectors

OpenAI, Gemini, Claude, Replicate adapters under one interface.

Contract Builder

Builds stable response envelopes for clients and downstream systems.

Unified provider abstraction with policy-aware routing and fallback management.

OpenAI

Primary for chat/voice/image generation lanes.

Gemini

Multimodal reasoning and image-capable lane.

Claude

Text-focused reasoning fallback lane.

Replicate

Specialized image/video generation lane.

Policy-aware routing

Provider Abstraction Router

Selects provider by capability, latency profile, reliability state, and environment policy.

Intent-driven dispatch with capability-aware primary/fallback selection.

Intent Type	Primary Route	Fallback Route	Output Contract
Text chat	OpenAI/Claude	Gemini	Standard text envelope
Voice conversation	OpenAI (voice lane)	Azure/OpenAI fallback	Voice + metadata envelope
Image generation	OpenAI/Replicate	Replicate/OpenAI	Images array + usage + metadata
Video generation	Replicate video lane	Policy-defined	Videos array + metadata

Normalized response envelope across all providers and modalities.

{
  "success": true,
  "content": "...",
  "provider": "...",
  "usage": {},
  "metadata": {},
  "images": [],
  "files": [],
  "videos": []
}