SDKs & clients
A polished OpenAI + TypeSafe Jev terminal interface for answers with transparent decision reports
from typesafe_sdk import Choice, Noul, RetryPolicy, Score, TypeSafeClientThe link points at the commit we read, so the line number still holds.
These question sets are lifted from this project's source exactly as written.
from src/jev_system_one/jev.py:32
requires_uncertainty_noticetrue/falseShould the answer explicitly disclose uncertainty or limits?
response_depthscoreHow much detail should the answer contain?
0 — brief: one direct answer with minimal context1 — standard: a concise answer with useful explanation2 — detailed: a thorough answer with structured reasoning or stepsimport { experimental_evaluate as evaluate } from 'ai';
const { answers } = await evaluate({
model: 'typesafe-ai/jev',
state,
questions: {
requires_uncertainty_notice: { type: 'boolean', instructions: 'Should the answer explicitly disclose uncertainty or limits?' },
response_depth: { type: 'score', instructions: 'How much detail should the answer contain?',
criteria: ['brief: one direct answer with minimal context', 'standard: a concise answer with useful explanation', 'detailed: a thorough answer with structured reasoning or steps'] },
},
});from src/jev_system_one/jev.py:107
requires_revisiontrue/falseShould OpenAI revise `openai_draft.answer` before it is shown?
import { experimental_evaluate as evaluate } from 'ai';
const { answers } = await evaluate({
model: 'typesafe-ai/jev',
state,
questions: {
requires_revision: { type: 'boolean', instructions: 'Should OpenAI revise `openai_draft.answer` before it is shown?' },
},
});