IVYXSTUDIO · IVY NODE

Jev · Score

ivy.node.jev-score · v0.1.0

ivyx✓

Asks TypeSafe Jev where a text or JSON state sits on an ordered scale of 2 to 10 described levels and returns a score from 0 to the top level with the nearest level, its description and the confidence. Use it to rate or grade urgency, severity, sentiment, quality or correctness. Levels go from low to high; the score is probability-weighted, so it can land between two levels. When the state is a JSON object its fields come back beside the answer, so a for-each over rows keeps each row.

#jev#typesafe#score#rate#grade#severity#sentiment#decision

Inputs

Field	Type	Description
staterequired	Not declared	The text, or a JSON object or array, that Jev judges.
instructionsrequired	string	What to rate, in words; the model never sees anything else about it.
levelsrequired	array	Two to ten level descriptions, from the low end of the scale to the high end.
jev_api_keyrequired	string	The TypeSafe API key; reference the stored secret jev-api-key, never the key itself.
base_url	string	The System One server.
model	string	The Jev model; pin a version such as jev-1.13.0 before tuning a threshold.

Outputs

Field	Type	Description
scorerequired	number	0 to the top level number; can fall between levels.
levelrequired	integer	The nearest level number, 0 for the first.
level_textrequired	string	The nearest level's description.
is	object	For each level, as lowercase words joined by _, whether it is the nearest one; a branch reads is.<level>.
normalizedrequired	number	The score scaled to 0–1.
confidencerequired	number	0–1, how peaked the probabilities are.
probabilities	object	Each level's probability, keyed by level number.
model	string	The model version that answered.

Source

python

inp = __ivy_ctx__["nodes"][__ivy_node_id__]["input"]

import json
import re
import requests

def slug(text):
    """The key a branch reads: "Needs action today" → "needs_action_today"."""
    return re.sub(r"[^a-z0-9]+", "_", str(text).lower()).strip("_")

def ask(question):
    state = inp.get("state")
    if state is None or (isinstance(state, str) and not state.strip()) or state == {} or state == []:
        raise ValueError("state is empty; give Jev the text or JSON to judge.")
    instructions = str(inp.get("instructions") or "").strip()
    if not instructions:
        raise ValueError("instructions is empty; say what Jev should decide.")
    key = inp.get("jev_api_key")
    if not key:
        raise ValueError("jev_api_key is missing; reference the stored secret jev-api-key.")
    question["instructions"] = instructions
    base = str(inp.get("base_url") or "https://api.typesafe.ai").rstrip("/")
    response = requests.post(
        base + "/v1/systemone",
        headers={"Authorization": "Bearer " + str(key), "Content-Type": "application/json"},
        json={"state": state, "model": inp.get("model") or "jev-latest", "questions": {"q": question}},
        timeout=60,
    )
    if response.status_code in (401, 403):
        raise PermissionError(f"Jev refused the API key (HTTP {response.status_code}).")
    if response.status_code != 200:
        raise RuntimeError(f"Jev answered HTTP {response.status_code}: {response.text[:300]}")
    body = response.json()
    answer = (body.get("answers") or {}).get("q")
    if not isinstance(answer, dict):
        raise RuntimeError("Jev's answer has no entry for the question.")
    return answer, body.get("model")

levels = [str(l) for l in (inp.get("levels") or []) if str(l).strip()]
if not 2 <= len(levels) <= 10:
    raise ValueError("levels needs between 2 and 10 descriptions, lowest first.")
answer, model = ask({"type": "score", "criteria": levels})
score = float(answer.get("score"))
level = max(0, min(len(levels) - 1, int(round(score))))

out = __ivy_ctx__["nodes"][__ivy_node_id__]["output"]
out["score"] = score
out["level"] = level
out["level_text"] = levels[level]
out["is"] = {slug(t): i == level for i, t in enumerate(levels)}
out["normalized"] = round(score / (len(levels) - 1), 4)
out["confidence"] = answer.get("confidence")
out["probabilities"] = answer.get("probabilities") or {}
out["model"] = model
# An object state's fields come back beside the answer, so a for-each over
# rows keeps each row (its date, its question) next to its score.
if isinstance(inp.get("state"), dict):
    for _k, _v in inp["state"].items():
        out.setdefault(_k, _v)

Tests

Requires: python:3.9, requests

severe
Two exclamation marks land on the top of a three-level scale.
calm
A calm message is level 0 on a five-level scale.
row
A row as the state comes back beside its score, and the node's own fields win.
one-level
A single level is refused before any call.
empty-state
An empty state is refused before any call.