ailerix
Model router where Jev classifies each request's task family and code picks the cheapest capable model on a cost Pareto chain.
Overview
Ailerix is a type-safe model router and OpenRouter competitor. You never name a model; the only public slug is ailerix/auto. Each request is classified by Jev into a task family (intelligence, coding, agents, vision, factual, long_context, professional), then software walks an Artificial Analysis cost-per-task Pareto chain to pick a provider.
Jev never sees catalog model ids in routing questions; it only classifies families. It exposes a Jev-compatible /api/v1/systemone endpoint, a /route endpoint that returns the chosen model, cost-per-task and reasons, an OAuth MCP server, a Cursor plugin and a Hugging Face demo Space. A local System One engine lets the playground work without credentials. Auth is by Clerk with an anonymous tier, credits by Stripe and storage on Neon Postgres.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Choice | Which task family does this request belong to? | intelligence · coding · agents · vision · factual · long_context · professional |
How it works
- Receive a request to ailerix/auto
- Jev classifies the task family
- Walk the cost-per-task Pareto chain for that family
- Forward to the chosen provider and return reasons
Strengths
- Model choice kept in code, not in the model
- Playground works offline with a local engine
- Exposes routing reasons and costs
Caveats
- Live Artificial Analysis data needs a commercial licence
- Completions are mocked until provider keys are added
Getting started
Set TYPESAFE_API_KEY to send routing questions to TypeSafe; see docs/SETUP.md for operator setup.At a glance
- Type
- Project
- Status
- beta
- Language
- TypeScript
- Runtime
- web service + MCP server
- Has code
- yes
Use cases and integrations
How curated lists describe it
Type-safe model router. Jev (System One) banks each request to a typed catalog route.
hellogumbo/awesome-jev
Frequently asked questions
Does Jev pick the model?
No, Jev only classifies the task family; code picks the cheapest model on the Pareto chain.
Can I try it without keys?
Yes, a local System One engine powers the playground, and there is an anonymous tier of 25 routes a day.
Is it OpenAI-compatible?
It positions itself as an OpenRouter competitor using a single auto model slug.
Related projects
Local OpenAI/Anthropic-compatible proxy where one Jev call picks the model route and thinking level for each coding-agent turn.
ProxybetaGoCost-, cache- and quota-aware LLM router that classifies each turn with a Jev-class model, prices every route, and checks cheap answers.
ProjectprototypePythonJev Router for Cursor: Jev picks a COST, BALANCED or INTELLIGENCE tier per task, with soft routing in chat and hard routing via the SDK.
ProjectprototypeTypeScriptLocal-first, outcome-aware model router for coding harnesses that picks a capability tier per task with Jev and learns from history.
ProjectprototypeRustLocal-first routing engine for long-running agent tasks that uses a Jev decider to pick a model group per turn.
PluginproductionRust, TypeScriptOpen-source AI gateway for 100+ LLMs that can use Jev to classify requests for complexity routing and to compact tool results.
ProxyproductionPython