Module 01 · Model routing
runnableSmart Router
One intended path to cloud and local models.
Runnable v0.1 slice available on GitHub — architecture and scope below.
Problem
Model access gets scattered across provider keys, endpoints, deployment rules, and separate UIs. Even a VS Code assistant and a home chatbot can end up repeating the same integration work.
Where it sits
Sits between clients and the model platforms they call: Foundry, Bedrock, Vertex/Agent Builder, or local models. It is meant to give VS Code, a home chatbot, and other tools one stable endpoint.
The router is designed to onboard services by reference and poll them for current endpoint information—including UIs that live elsewhere. Provider interfaces stay where they are; Mirevox keeps the reference, central credentials, and routing policy.
A request can target on-prem or cloud capacity according to the activity, then fan out when several models or toolsets need to answer the same input. VS Code, a side agent, and a home chatbot are meant to call the same endpoint instead of each learning every provider. Works alongside LiteLLM-class proxies; it is not locked to one proxy stack.
How it fits
This is the path through the control plane. Evals & English Governance and Data Lineage are meant to observe requests on that path using request and trace IDs, while the thin hub composes the modules without absorbing their jobs.
Capabilities
- Intended reference-based onboarding with endpoint polling
- Central handling of provider keys and credentials
- Per-activity routing between on-prem and cloud targets
- One endpoint for VS Code, home chat, agents, and external UIs by reference
- Fan-out across multiple model and toolset targets
- Works alongside LiteLLM-class proxies and alternatives
Other modules
- Evals & English GovernanceEach model platform has its own eval screens, scoring terms, and policy controls. Teams end up comparing results by hand, while governance rules live in documents that do not map cleanly to provider settings.
- Stage GateAI-assisted projects can jump from an appointment straight into generated code. The missing middle—prerequisites, requirements, decisions, and clear advance criteria—is usually spread across documents and understood by only a few people.
- Shadow VersioningOne project can have its brief in Confluence, approvals in SharePoint, code in Git, and evidence in a testing system. Humans see one piece of work, but the tools expose unrelated histories and technical version language.
- Context EngineAn agent working across 15 repositories can spend more tokens rediscovering the codebase than answering the question. Every tool builds its own partial picture, then throws much of that work away.
- Data LineageOnce a prompt crosses a gateway, it can be hard to explain which provider, region, model, or internal service handled it. Existing lineage tools cover data estates well, but AI request paths often need a smaller, trace-linked view.
- AI-DORAAI coding agents inflate commit counts and PR volume, hiding code churn, review bottlenecks, and rework. Traditional DORA dashboards track deployment frequency without telling you if half the AI-written lines were reverted two weeks later.