jev-repl

# jev

A terminal REPL for shaping [TypeSafe AI](https://typesafe.ai) System One requests before you
write any code: send a `state` plus named, typed questions and read the typed answers back.

```sh
npx jev-repl          # or: npm i -g jev-repl && jev
```

With `TYPESAFE_API_KEY` set you get live answers; without it, answers are simulated locally so the
shapes can be learned offline.

| Question | Answer                                                                  |
| -------- | ----------------------------------------------------------------------- |
| `noul`   | probability of "yes" (0–1)                                              |
| `choice` | selected label, per-label probabilities, confidence                     |
| `score`  | probability-weighted level, legend, per-level probabilities, confidence |

## Inside

```text
:preset triage                                   # a ready-made session to poke at
:state The payout failed again, third time.      # bare text works too
:noul is_urgent The message conveys urgency | yes: A deadline | no: Routine
:choice department Which team | billing=Payments | technical=Bugs
:score frustration How frustrated | Calm | Annoyed | Furious
<Enter>                                          # send; answers come back with their distributions
```

- `:lesson` walks an eleven-step track from "what is a noul", past what a call costs, to taking
  the session out as code.
- `:sketch` (Ctrl-K) opens the whole request as one page of text, with the question types read
  off the punctuation, a gutter that says what each line became, and a live preview of the JSON,
  simulated answers, the same request as TypeScript, or what a call would cost:

  ```text
  The payout failed again, third time this month.
  ---
  is_urgent? The message conveys