Organization repository on GitHub
Open-source LLM router that uses TypeSafe's Jev to pick a model, on top of LiteLLM
- Forks
- 1
- Last pushed
- Updated Sep 2026
Repo stats from the GitHub API, cached Oct 2026.
Jev directory
CommunityJev Tools & CLIs
Jev Router — prismhq
An OpenAI-compatible LiteLLM proxy where clients send one model id and Jev picks which configured model serves each request, with capability filtering and a cheapest-eligible fallback.
- Stars
- 14
- Language
- Python
- License
- MIT
- Reading time
- 3 min
- Published by
- Community
- Author
- prismhq
- Last verified
- Sep 2026
This is one of several Jev listings for the same project. The canonical entry is Jev Router — gargpratyush.
About this project
Open-source LLM router. Clients send one model id; Jev (TypeSafe's System One model) picks which model serves each request. Transport is LiteLLM, so the router can route across OpenRouter or any provider LiteLLM supports. You bring your own keys. Works with no third-party decision model: without a TypeSafe key, a cheapest-eligible baseline runs by default. Optional Jev: set TYPESAFEAPIKEY and Jev makes the routing call. OpenAI-compatible: point any client at the proxy and use model: "jev-router".
Quoted from the project's README. Read the full README.
Works forJev for Routing
Used inJev for Software & DevTools
Highlights
- Clients point any OpenAI-compatible client at the proxy with model jev-router; the response model field reports which model served it.
- eligible() filters candidates in router.yaml on vision, output length, and tool support before any decision is made.
- Without a TYPESAFE_API_KEY a RulesDecider picks the cheapest eligible model, and any failure falls back to router.yaml's fallback.
- Transport is LiteLLM, so it can route across OpenRouter or any provider LiteLLM supports; you bring your own keys.
- Two files: config.yaml holds the LiteLLM pool and hook registration, router.yaml the candidate descriptions, capability flags, and prices.
Quickstart
bash
pip install "litellm[proxy]" pyyaml
export OPENROUTER_API_KEY=YOUR_API_KEY
export TYPESAFE_API_KEY=YOUR_API_KEY # optional: enables Jev routing
litellm --config config.yaml --port 4000Watch out
MIT-licensed and experimental with changing APIs and config; needs LiteLLM plus an OpenRouter key, and Jev routing sends a minimized request summary to TypeSafe.
Questions about this entry
Short answers drawn from the linked project or page.
Transport yes, routing no. A pre-call hook in jev_router/hook.py intercepts requests for the jev-router model id before anything reaches an upstream provider, builds a minimized summary, and runs eligible() over router.yaml to drop candidates that cannot serve the request on vision, output length, or tools. A decider then picks one model, and any failure lands on the fallback entry in router.yaml.
It still routes. Without TYPESAFE_API_KEY a RulesDecider picks the cheapest eligible candidate and routing stays entirely local, sending nothing to TypeSafe. Set the key and JevDecider takes over and makes the routing call. That means the no-key path is genuinely free to try, though it reduces to price-based selection rather than judging the request.
A minimized summary rather than the full request: recent messages' roles and truncated text, plus image and tool signals, and that summary goes to TypeSafe when the key is set. The README's advice is blunt about the consequence: do not enable Jev routing for workloads that cannot share context with another provider.
The README labels it experimental and says APIs and config will change, so treat the schema as unsettled. You need LiteLLM plus your own provider keys, and candidate names in router.yaml must match model_list entries in config.yaml, which is the part that breaks most often when you edit the OpenRouter slugs. Self-host the proxy rather than pointing it at production before the config stops moving.
pip install "litellm[proxy]" pyyaml, export OPENROUTER_API_KEY and optionally TYPESAFE_API_KEY, then litellm --config config.yaml --port 4000. Call it like an OpenAI endpoint at /v1/chat/completions with model jev-router.
config.yaml is the LiteLLM proxy with model_list and the hook registration, and router.yaml is the routing policy with candidates, the descriptions Jev reads, capability flags, prices, the fallback, and Jev settings. Candidate names must match model_list entries.
The response model field reports the model that actually served the request. The pre-call hook in jev_router/hook.py builds a minimized summary, eligible() filters by vision, output length, and tools, the decider picks one, and any failure falls back to fallback in router.yaml.
Reactions & coverage
Posts, threads, and videos about this entry from around the web.
X: Jev 1.14 is now on OpenRouter
Facebook: Intake Router keeps diagnosis out of the answer space at the front desk.
FacebookClinicnotes
X: OpenRouter Decision Model Rankings
X: herdr + jev, an agent router
More Jev resources like this
A Python toolkit that installs nine plain SKILL.md files for Hermes, Claude Code, and Codex, letting Jev route models, filter retrieved passages, select skills, and choose bounded computer or browser actions.
Jev Tools & CLIs#community#python#agents
Community
kerpopule/hermes-jev-skillsPython
kerpopule/hermes-jev-skillsPythonA local gateway for coding agents that asks Jev which tool to call each turn and steers the LLM when Jev is confident, while every other request goes to your usual provider untouched.
Jev Tools & CLIs#community#typescript#gateway
Community
vinilana/jev-gatewayTypeScript
vinilana/jev-gatewayTypeScriptA reference router and skill that put Jev in front of Grok Bot so each request is classified before expensive research, browser, retry, or subagent work, returning actions such as reuse_cache or ask_human.
Jev Tools & CLIs#community#python#routing
Community
Bodila51/grok-bot-jevPython
Bodila51/grok-bot-jevPythonA local proxy between Claude Code and the Anthropic API that picks the model and effort level per message with Jev, routing subagents whenever cheap but gating main-chat switches on prompt-cache arithmetic.
Jev Tools & CLIs#community#typescript#routing
Community
adarshmishra07/jcm-routerTypeScript
adarshmishra07/jcm-routerTypeScriptAn experimental Codex CLI router where a local Responses proxy asks Jev one Choice per model call to select a host-available model and reasoning effort, while application-owned acceptance decides whether the task passed.
Jev Tools & CLIs#community#typescript#codex
Community
miniLV/Jev-Auto-RouterTypeScript
miniLV/Jev-Auto-RouterTypeScriptRoutes each prompt to a model tier with one Jev call: a Choice question classifies difficulty (simple, moderate, complex, frontier) and a Score rates stakes. Confidence-gated rules then fall back or escalate, and your application still makes the downstream request.
Jev Tools & CLIs#community#catalog-import#logicrw
Community
gualican/jev-model-routerPython
gualican/jev-model-routerPythonFrom the Jev guides
Original write-ups on what this entry is for.
From the Jev AI model community
Posts from builders shipping with Jev right now.