fare-policy-assistant/docs/pages/index.html at main · ghzhost/fare-policy-assistant · GitHub

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
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
<!doctype html>
<html lang="en">
<head>
<meta charset="utf-8">
<meta name="viewport" content="width=device-width, initial-scale=1">
<meta http-equiv="Content-Security-Policy"
  content="default-src 'none'; style-src 'unsafe-inline'; img-src 'self'; base-uri 'none'; form-action 'none'">
<title>Verified evaluation evidence — Transit Fare Policy Assistant</title>
<style>
  :root { color-scheme: light dark; }
  * { box-sizing: border-box; }
  body {
    margin: 0;
    color: #17201b;
    background: #f7faf8;
    font: 1rem/1.55 system-ui, -apple-system, BlinkMacSystemFont, "Segoe UI", sans-serif;
  }
  main { max-width: 64rem; margin: 0 auto; padding: 1.5rem 1rem 4rem; }
  h1 { max-width: 42rem; font-size: clamp(1.8rem, 5vw, 2.8rem); line-height: 1.15; }
  h2 { font-size: 1.2rem; }
  .lede { max-width: 48rem; color: #425048; }
  .notice, .card {
    margin: 1rem 0;
    padding: 1rem;
    border: 1px solid #a8b4ad;
    border-radius: .6rem;
    background: #fff;
  }
  .notice { border-left-width: .5rem; }
  .notice.verified { border-left-color: #14733b; }
  .notice.warning { border-left-color: #9a6500; }
  .red { border-left: .35rem solid #991b1b; padding-left: .7rem; }
  .status-label { display: block; font-size: 1.15rem; }
  dl { display: grid; grid-template-columns: minmax(9rem, 1fr) 2fr; gap: .4rem 1rem; }
  dt { font-weight: 700; }
  dd { margin: 0; overflow-wrap: anywhere; }
  table { width: 100%; border-collapse: collapse; }
  caption { padding: .5rem 0; text-align: left; font-weight: 700; }
  th, td { border: 1px solid #a8b4ad; padding: .55rem; text-align: left; }
  code { overflow-wrap: anywhere; }
  img { display: block; width: 100%; height: auto; }
  a { color: #075ea8; }
  a:focus-visible { outline: 3px solid #075ea8; outline-offset: 3px; }
  @media (max-width: 35rem) {
    dl { grid-template-columns: 1fr; }
    dd { margin-bottom: .6rem; }
  }
  @media (prefers-color-scheme: dark) {
    body { color: #edf5ef; background: #131a16; }
    .lede { color: #bdc8c0; }
    .notice, .card { background: #1c2620; border-color: #64736a; }
    a { color: #79bfff; }
  }
</style>
</head>
<body>
<main>
  <p><strong>Reference implementation.</strong> This portfolio project is not an
    official service of any transit agency.</p>
  <h1>Verified evaluation evidence</h1>
  <p class="lede">This page is generated from a canonical promotion receipt. It
    publishes aggregate scores and safe identifiers only—never rider inputs,
    model responses, prompts, rationales, or retrieved passages.</p>
  <p><a href="https://evals.chelseakr.com/" aria-current="page">Canonical
      evaluation evidence hub</a> ·
    <a href="https://yahp6ddfo1.execute-api.us-west-2.amazonaws.com/">Live rider
      assistant</a></p>

  <section class="notice {{STATUS_CLASS}}" aria-labelledby="status-heading">
    <h2 id="status-heading">Evidence status</h2>
    <strong class="status-label">{{STATUS_LABEL}}</strong>
    <p>{{STATUS_DETAIL}}</p>
    <p>Evaluated <time datetime="{{RUN_AT}}">{{RUN_AT}}</time>; promoted
      <time datetime="{{PROMOTED_AT}}">{{PROMOTED_AT}}</time>.</p>
  </section>

  <section class="card" aria-labelledby="benchmark-heading">
    <h2 id="benchmark-heading">Published evaluation context</h2>
    <p><strong>Committed promoted baseline: 192/201 (95.5%).</strong></p>
    <p class="red"><strong>Latest observed: 190/201 overall and 175/186
      production-core.</strong> The cross-agency promotion gate is red, so
      the observed run has not replaced the committed baseline.</p>
    <p><strong>What these scores do not cover.</strong> They were produced over
      201 cases against five agencies. The suite now holds 385 cases across
      eighteen agencies: Elk Grove e-tran, Santa Cruz METRO, SolTrans and
      Fresno Area Express were added on 12&ndash;13 August 2026, and County
      Connection, San Joaquin RTD, AC Transit, WestCAT, SLO RTA, VTA, Napa
      VINE, SamTrans, and Marin Transit on 13&ndash;14 August 2026, and
      <em>none of their cases
      has been scored in a promoted run</em>. The scores also predate system
      prompt v11, which widened the assistant's scope rule from five named
      agencies to all in the corpus (and v12 through v20, which name the
      newest nine);
      every number above was produced under v10, which
      instructed the model to decline the newer agencies. Treat this page as
      evidence about the five agencies it measured, and about nothing else.</p>
  </section>

  <section class="card" aria-labelledby="score-heading">
    <h2 id="score-heading">Promotion score</h2>
    <p><strong>{{TOTAL_SCORE}}</strong> across {{CASE_COUNT}} complete cases.</p>
    <table>
      <caption>Scores by evaluation suite</caption>
      <thead>
        <tr><th scope="col">Suite</th><th scope="col">Passed</th>
          <th scope="col">Pass rate</th></tr>
      </thead>
      <tbody>{{SUITE_ROWS}}</tbody>
    </table>
    <p><a href="report.html">Review the sanitized case-level report</a></p>
  </section>

  <section class="card" aria-labelledby="release-heading">
    <h2 id="release-heading">Immutable runtime release</h2>
    <dl>
      <dt>Lambda version</dt><dd>{{FUNCTION_VERSION}}</dd>
      <dt>Release digest</dt><dd><code>{{RELEASE_VERSION}}</code></dd>
      <dt>Source revision</dt><dd><code>{{SOURCE_REVISION}}</code></dd>
    </dl>
    <p><a href="release.json">Machine-readable release receipt</a> ·
      <a href="public-evidence.json">Canonical public evidence</a></p>
  </section>

  {{TREND_SECTION}}
</main>
</body>
</html>