feat(intern-decision-serve): 0.1.1 adds POST /v1/systemone (Jev wire shape)

Straight passthrough to the checkpoint's own DecisionEngine.predict — never the
semif mapping, whose different prompt would change the answers. Reuses the one
inference thread, bearer auth, MAX_QUEUE, VRAM cap and error envelope; no new
concurrency. 1..16 questions in ONE call (never chunked: Jev questions share a
prompt); images 422; over MAX_TOKENS 422 before the forward. /health advertises
the surface. Response 'model' is a string name@revision (JevBench's runner
hashes it; a dict broke its manifest step).

Acceptance on the live service (see acceptance/systemone-2026-09-30/): JevBench
v1.2.16 typesafe adapter over the 231 public items scores all 202/231, hard
83/111, with 0 changed answers across all 924 rows of the bench's own r1..r4;
controls 401/422x3 (token boundary proven at 7168 pass / 7169 refuse); GPU 1
per-process peak 9,866 MiB under the largest accepted request (budget 9,876);
/decide/shared positive control unchanged. 120 tests green.

services/intern-decision-serve/tests/fake_engine.py

@@ -22,6 +22,32 @@ def score_by_description(field: str, question: dict) -> list[float]:
     return [len(d) + (0.5 if i == 0 else 0.0) for i, d in enumerate(descs)]
 
 
+def answer_for(field: str, question: dict, scorer) -> dict:
+    """One Jev answer in the shape DecisionEngine.predict returns, for any of the three types."""
+    kind = question["type"]
+    if kind == "noul":
+        probs = {"yes": 0.75, "no": 0.25}
+        return {"type": "noul", "probabilities": probs, "confidence": 0.75, "noul": 0.75,
+                "source": "local", "decision": "yes"}
+    criteria = question["criteria"]
+    if isinstance(criteria, dict):
+        keys = list(criteria)
+        descriptions = list(criteria.values())
+    else:                                                   # score: a list of levels
+        keys = [str(i) for i in range(len(criteria))]
+        descriptions = [str(level) for level in criteria]
+    probs = dict(zip(keys, softmax(scorer(field, {**question, "criteria": dict(zip(keys, descriptions))}))))
+    best = min(keys, key=lambda k: (-probs[k], k))
+    answer = {"type": kind, "probabilities": probs, "confidence": probs[best], "source": "local", "decision": best}
+    if kind == "choice":
+        answer["choice"] = best
+    else:
+        answer["score"] = sum(float(k) * p for k, p in probs.items())
+        answer["legend"] = {k: (criteria[i] if isinstance(criteria, list) else v)
+                            for i, (k, v) in enumerate(zip(keys, criteria.values() if isinstance(criteria, dict) else criteria))}
+    return answer
+
+
 class FakeEngine:
     def __init__(self, scorer=score_by_description, tokens_per_question: int = 100):
         self.scorer = scorer
@@ -35,13 +61,8 @@ class FakeEngine:
 
     def predict(self, request: dict) -> tuple[dict, str]:
         self.calls.append(copy.deepcopy(request))
-        answers = {}
-        for field, question in request["questions"].items():
-            ids = list(question["criteria"])
-            probs = dict(zip(ids, softmax(self.scorer(field, question))))
-            best = min(ids, key=lambda i: (-probs[i], i))
-            answers[field] = {"type": "choice", "probabilities": probs, "confidence": probs[best],
-                              "choice": best, "source": "local", "decision": best}
+        answers = {field: answer_for(field, question, self.scorer)
+                   for field, question in request["questions"].items()}
         response = {"answers": answers,
                     "usage": {"input_tokens": self.tokens_per_question * len(answers),
                               "output_tokens": len(answers), "decision_count": len(answers)},