Overview & result
Instead of multiple round-trips to decide model, tools, and effort level, this harness batches the entire orchestration setup into a single Jev decision turn.
From the original post
One fast call decides how an agent turn should run — then runs it.
How Jev fits in the loop
- Receive user turn in agent runtime.
- Send batched Jev request covering model tier, tool set, and effort rating.
- Let application code apply the selected model, tool, skill and effort settings, with documented fallbacks.
How to reproduce
- Inspect the source implementation at https://github.com/JoacoMarc/jev-harness-router.
- Configure local Jev API credentials and register required tool/model bindings.
- Execute test prompts and observe decision latency and routing outputs.
Why this build matters
Minimizes orchestration overhead in multi-agent workflows.
Limitations
- It does not execute tools: turning the decision into tool definitions and an execute loop is the harness's job.
- The README notes that Jev does not treat state as hostile by default, so the safety floor lives in code.
- Its figures (about 350 ms per call; one 20-question call being 2.5× cheaper and 24× faster than 20 calls) were measured on the author's own catalogue.