What happens if the router is unreachable?

It fails open: OmO behaves exactly as before, and skills cut from the top-K keep their names in an other_skill_names index, so no skill is ever fully hidden from the main model.

Why is this repository archived?

GitHub metadata marks it archived as of this writing; the README remains useful because it documents the fixed evaluation results and the honest limitations of one-pass skill routing.

How do I train the router on my own sessions?

Run extract/extract_dataset.py over ~/.omo/sessions and ~/.omo/skills to produce train, validation and test jsonl plus catalog.json, split by session. It is read-only, the data stays local, and training on roughly 1.4k rows takes about two minutes on CPU.

What exactly does the extension change in my prompt?

It rewrites only the available_skills block to the top-K skills, keeps any skill you invoke explicitly with $skill:name, /skill:name or skill:name, and returns the prompt unchanged on any error.

How do I point the extension at a router on another machine?

Start the container with ROUTER_BIND=0.0.0.0, or a specific IP, then set JEVLIKE_ROUTER_URL on the OmO process, for example http://192.168.1.10:8318. The default is http://127.0.0.1:8318.

What are the tuning knobs, and how do I turn it off?

JEVLIKE_ROUTER_TOPK sets how many skills stay in the prompt (default 24), JEVLIKE_ROUTER_TIMEOUT_MS is the fail-open deadline (default 2500 ms), and setting JEVLIKE_ROUTER_OFF=1 disables the router entirely.

Why do skills added after training never rank?

The router only knows the catalog it trained on, so a new skill is invisible to it and simply never scores; explicit invocation still works, and you retrain after catalog changes. Routing quality is bounded by your session data, and the most-read skill dominates short prompts.

CommunityJev Tools & CLIs

Omo Jevlike Router

A Jev-style one-pass skill router for OmO agents: a frozen Qwen2.5-0.5B with a tiny option-attention head scores the whole skill catalog per prompt, and an extension shrinks the system prompt to the top-K skills, failing open.

See highlights

Open github.com

Stars: 3
Language: Python
License: MIT
Published by: Community
Last verified: Oct 2026

islee23520/omo-jevlike-routerArchived

User repository on GitHub

View on GitHub

Jev-style one-pass skill router for OmO: shrink the skill catalog in your system prompt with one forward pass (frozen Qwen2.5-0.5B + jevlike head, fail-open extension)

Forks: 0
Last pushed: Updated Sep 2026

Repo stats from the GitHub API, cached Oct 2026.

About this project

A Jev-style one-pass skill router for OmO. On every agent start, a small local model (frozen Qwen2.5-0.5B + a tiny option-attention head, jevlike) scores your entire skill catalog against the user's prompt in a single forward pass. An OmO extension then shrinks the block of the system prompt to the top-K skills before your expensive main model sees it. Smaller prompts, fewer wasted skill reads, sub-100 ms routing on a warm cache — and fail-open: if the router is unreachable, OmO behaves exactly as before.

Quoted from the project's README. Read the full README.

Tagscommunity routing skills qwen

Works forJev for Routing Jev for Classification

Used inJev for Software & DevTools

Languages: Python
Platforms: API
Capabilities: Open sourceSelf-hostedRuns locally

Highlights

On a held-out test of 132 turns: top-1 27.3%, recall@24 84.1%, versus recall@12 74.2% and recall@32 90.2%.
Warm routing latency is roughly 50-70 ms; the first full-catalog call takes about 9 s for cache warm-up.
The README documents an earlier evaluation bug and the corrected numbers reported by the fixed evaluator.
Scores arbitrary option lists via POST /score {context, options}; the OmO extension rewrites only the available_skills block.

Watch out

The repository is archived, so it is no longer maintained; routing quality is bounded by the session data you extract. MIT licensed.

Questions about this entry

Short answers drawn from the linked project or page.

Reactions & coverage

Posts, threads, and videos about this entry from around the web.

Facebook: Intake Router keeps diagnosis out of the answer space at the front desk.

FacebookClinicnotes

YouTube: Jev explained for routing and judging, with a comparison against LLMs.

Hacker News: Vercel on using Jev for defined decisions and GPT-6 Astra when the job also has to write or use tools.

flashbrew4 pointsTypeSafe AI Jev vs. GPT-6 AstraRead the thread on Hacker News

Hacker News: ASTGL replay chose local Laya routing over hosted Jev on one Mac Studio setup.

Jmeg8r15 pointsDoes an open-weight decision model beat a hosted one? Jev vs. LayaRead the thread on Hacker News

Related terms

Glossary definitions related to this entry.

More Jev resources like this

Paca

1.9k

A Go-based team workspace with optional Jev AI for agent auto-routing, task auto-fill, auto-assign, and automation condition nodes.

Jev Tools & CLIs#community#go#agents

Community

Paca-AI/pacaGo

AutoJev

130

An open recipe for a 27B multimodal decision model trained with full-weight SFT on 73,000 examples over one H200, serving choice, noul and score on /v1/systemone with per-checkpoint provenance and released calibration plots.

Jev Models & Reimplementations#community#calibration#qwen

Community

denis-pplx/autojevPython

judgekit

CLI and Python library that runs declarative YAML judgment tasks - classify, route, score, verify - natively on Jev Choice/Score/Noul, on any OpenAI-compatible backend, or on local keyword rules, with confidence gates, fallbacks and per-decision cost records.

Jev Tools & CLIs#community#yaml#classification

Community

lexingtonhibiki/judgekitPython

Qwev

A training-free runtime that turns dense Qwen3 or Qwen3.5 checkpoints into a Jev-style Noul, Choice and Score service, sharing one state prefill across isolated questions and reading each answer from candidate-token logits.

Jev Models & Reimplementations#community#qwen#local-inference

Community

HopLee6/QwevPython

Decision 2.0 Vega

The vLLM Semantic Router project's 29.37B decision model on Qwen3.8-27B: a probability per supplied option with zero generated tokens, 74.0 on JevArena and 56.5 on Jev Decision Index 0.2.1.

Jev Models & Reimplementations#community#open-models#local

Community

AgentScope

33k

Alibaba's multi-agent platform, which implements native TypeSafe Jev classification through JevClassifierModel for binary, choice and score routing across agent pipelines. Python 3.11+, Apache-2.0, with a FastAPI agent service and web UI.

Jev Integrations & Adapters#community#multi-agent#framework

Community

agentscope-ai/agentscopePython

From the Jev guides

Original write-ups on what this entry is for.

Jev on the Gateways: Vercel, Workers AI, and OpenRouter

Oct 12, 2026

The request shape barely changes across surfaces; the surrounding contract does. When to call the API directly and when a gateway, a Worker binding, or a router entry earns its extra hop.

Jev Tools & CLIs#gateway#vercel#cloudflare

Read article

Exporting Decision Models: GGUF, ONNX, and Device Runtimes

Oct 10, 2026

Taking a decision checkpoint out of Python: what the GGUF and browser WebGPU exports preserve, what they change, and the parity evidence the model cards publish.

Jev Tools & CLIs#open-models#local#onnx

Read article

Records at Rest: Documents, Inboxes, and Risk Queues

Sep 28, 2026

Jev labels a record that already exists. Code moves, flags, or holds it. How the document, mail, and transfer tools keep the action out of the model.

Jev Tools & CLIs#classification#email#security

Read article

Back to all resources

From the Jev community

Posts from builders shipping with Jev right now.

Follow @typesafeai