Zum Inhalt

tests/test_gui_explain.py

  1
  2
  3
  4
  5
  6
  7
  8
  9
 10
 11
 12
 13
 14
 15
 16
 17
 18
 19
 20
 21
 22
 23
 24
 25
 26
 27
 28
 29
 30
 31
 32
 33
 34
 35
 36
 37
 38
 39
 40
 41
 42
 43
 44
 45
 46
 47
 48
 49
 50
 51
 52
 53
 54
 55
 56
 57
 58
 59
 60
 61
 62
 63
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
import json

from fastapi.testclient import TestClient

from touchstone.gui import app as gui_app
from touchstone.gui.control import RunRegistry

PACK = "packs/ndassist.yaml"


class _FakeLauncher:
    def spawn(self, argv):
        return 1

    def alive(self, pid):
        return False

    def terminate(self, pid):
        return None


def _client(tmp_path):
    reg = RunRegistry(runs_dir=tmp_path, launcher=_FakeLauncher())
    return TestClient(gui_app.create_app(runs_dir=tmp_path, registry=reg))


def _resp_dict(model, variant):
    from touchstone.results import EvalResponse

    return EvalResponse(
        pack_id="ndassist",
        pack_version=1,
        machine="t",
        model=model,
        quant="q",
        engine="e",
        engine_version="x",
        variant=variant,
        category="A",
        prompt_id="A1",
        repeat=0,
        response_text="ok",
        content_empty=False,
        ttft_s=0.1,
        decode_tps=1.0,
        prefill_tps=1.0,
        e2e_s=1.0,
        prompt_tokens=1,
        completion_tokens=1,
        is_cold_start=False,
        power_source="ac",
        peak_rss_mb=0.0,
        sys_used_mb=0.0,
        mem_pressure_max="normal",
        throttled=False,
        ok=True,
        error="",
        seed=42,
        t_start=0.0,
        t_end=1.0,
        reasoning_chars=0,
    ).as_dict()


def _write_bundle(d, *, groups, scores_by_group, blank_dims=()):
    """Build a self-contained judged bundle under d using the real in-repo pack."""
    from touchstone.pack import load_pack

    d.mkdir(parents=True, exist_ok=True)
    pk = load_pack(PACK)
    (d / "bundle.json").write_text(
        json.dumps(
            {
                "pack_id": pk.id,
                "pack_path": PACK,
                "models": [{"id": g[0], "quant": "q"} for g in groups],
                "date": "2026-06-20",
            }
        ),
        encoding="utf-8",
    )
    (d / "responses.jsonl").write_text(
        "\n".join(json.dumps(_resp_dict(m, v)) for (m, v) in groups) + "\n",
        encoding="utf-8",
    )
    header = ["model", "variant", "metric_type", "metric", "weight", "score"]
    lines = [",".join(header)]
    for m, v in groups:
        scores = scores_by_group[(m, v)]
        for dim in pk.dimensions:
            raw = "" if dim.id in blank_dims else str(scores.get(dim.id, ""))
            lines.append(f"{m},{v},dimension,{dim.id},{dim.weight},{raw}")
    (d / "scores.csv").write_text("\n".join(lines) + "\n", encoding="utf-8")
    return pk


def test_compare_headers_have_glossary_tooltips(tmp_path):
    _write_bundle(
        tmp_path / "2026_eval_nd",
        groups=[("m", "baseline")],
        scores_by_group={("m", "baseline"): {"Q1": 4}},
    )
    r = _client(tmp_path).get("/compare")
    assert r.status_code == 200
    assert "Time-To-First-Token" in r.text  # tooltip text from glossary


def test_base_no_longer_references_htmx(tmp_path):
    r = _client(tmp_path).get("/")
    assert "htmx.min.js" not in r.text