No Jev access yet?¶
Jev is new, and some keys may not have access yet. You can still take the whole course.
Use the adapter backend¶
Add one line to .env:
Now every ask_jev(...) call is answered by your LLM (MODEL) through TypeSafe's open-source
system-one-adapter. It exposes the same
system_one() API and returns the same response types, so no notebook code changes.
What works¶
We ran every notebook with JEV_BACKEND=adapter:
| With the adapter | |
|---|---|
| Notebooks 01 and 03-12 | run, and their built-in accuracy checks pass |
app.py and jobs/email_triage.py |
run |
| Notebook 02 (the benchmark) | needs real Jev, because it benchmarks Jev itself; it stops with a clear message |
What's different¶
- Slower. Each decision takes an LLM round-trip (~2-3 s instead of ~0.4 s).
- More expensive. You pay LLM prices, including output tokens.
- The race tab says so. In
app.py, the "Jev vs LLM race" tells you the "Jev" side is really your LLM.
When you get Jev access, switch back with JEV_BACKEND=typesafe (or delete the line). Nothing else changes.