Skip to main content

​
ProviderRouter

ProviderRouter is the dispatch layer between the orchestrator and individual provider implementations. It is used as an async context manager. 
from mutual_dissent.config import load_config
from mutual_dissent.providers.router import ProviderRouter

config = load_config()
async with ProviderRouter(config) as router:
    response = await router.complete("claude", prompt="Hello")
On entry, the router opens connections to:
  • OpenRouterProvider if an OpenRouter key is configured
  • Direct providers for each vendor that has both an API key and a registered implementation
On exit, all provider connections are closed concurrently.

​
Routing Decision

router.route(alias) returns a RoutingDecision without making any API calls. This is used by config test to show routing before sending requests. 
decision = router.route("claude")
# RoutingDecision(vendor=Vendor.ANTHROPIC, mode='auto', via_openrouter=False)

​
complete() and complete_parallel()

complete() routes and executes a single request. complete_parallel() fans out a list of requests concurrently via asyncio.gather. Both return ModelResponse objects — errors are captured in response.error rather than raised.

​
Provider Implementations

VendorImplementationStatus
Anthropicproviders/anthropic.py✅ Direct
OpenAI—Via OpenRouter
Google—Via OpenRouter
xAI—Via OpenRouter
Nvidia—Via OpenRouter
Groq—Via OpenRouter
OpenRouterproviders/openrouter.py✅ Universal fallback
All providers implement the Provider base class in providers/base.py.

​
Vendor Resolution

The router maps model aliases and OpenRouter model IDs to vendors via their ID prefix (anthropic/, openai/, google/, x-ai/, nvidia/, groq/). Unknown prefixes fall back to Vendor.OPENROUTER.