File size: 14,324 Bytes
064fb44
f3d0a56
 
064fb44
f3d0a56
 
 
 
 
a264e34
c31f724
52716e4
 
a264e34
064fb44
f3d0a56
 
 
 
 
 
 
 
 
 
 
 
064fb44
 
 
6ae309d
f3d0a56
 
 
6ae309d
f3d0a56
 
064fb44
 
 
f3d0a56
 
 
 
 
 
 
 
064fb44
 
 
 
f3d0a56
064fb44
f3d0a56
 
064fb44
f3d0a56
a56caea
52716e4
 
 
 
 
 
 
 
 
 
 
 
 
 
a56caea
 
 
 
 
 
 
 
 
 
 
52716e4
 
 
 
 
 
 
 
 
 
064fb44
f3d0a56
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
4986b4a
 
 
 
 
 
 
 
 
 
 
 
f3d0a56
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
064fb44
f3d0a56
 
 
 
 
 
 
 
 
 
 
 
 
49aa5e3
52716e4
 
 
 
 
 
 
 
 
a56caea
 
 
 
 
52716e4
49aa5e3
 
 
 
 
 
 
 
 
 
 
 
f3d0a56
 
4986b4a
 
 
 
 
 
 
 
 
 
 
f3d0a56
a264e34
4986b4a
 
f3d0a56
a264e34
 
 
064fb44
f3d0a56
 
52716e4
 
 
 
 
 
 
 
 
 
 
 
a264e34
 
064fb44
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
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
# JustDial-themed Category -> Vertical & Sub-vertical Classifier (Gradio 6.x safe)
try:
    import spaces  # ZeroGPU runtime on Hugging Face
except ImportError:                       # local / CPU fallback
    class spaces:
        @staticmethod
        def GPU(fn=None, **kw):
            return (lambda f: f)(fn) if fn else (lambda f: f)

import gradio as gr
from inference import classify, VERT_THR, SUB_THR
import os, tempfile
import pandas as pd

JD_BLUE, JD_ORANGE, GREEN, RED = "#00457C", "#F58220", "#16a34a", "#dc2626"

def conf_color(c):
    return GREEN if c >= 0.80 else (JD_ORANGE if c >= 0.60 else RED)

def bar(pct, color):
    return (f'<div class="jd-track"><div class="jd-fill" '
            f'style="width:{pct:.1f}%;background:{color}"></div></div>')

def render(r):
    vc, sc = r["vertical_confidence"], r["sub_vertical_confidence"]
    v_flag = ('<span class="jd-badge jd-badge-warn">&#9888; Review &middot; low confidence</span>'
              if r["vertical_review"] else '<span class="jd-badge jd-badge-ok">&#10003; Confident</span>')
    v_card = (f'<div class="jd-card"><div class="jd-card-top"><span class="jd-eyebrow">VERTICAL</span>'
              f'{v_flag}</div><div class="jd-value jd-blue">{r["vertical"]}</div>'
              f'{bar(vc*100, conf_color(vc))}<div class="jd-pct">{vc*100:.1f}% confidence</div></div>')
    if r["propose_new_sub_vertical"]:
        s_flag = '<span class="jd-badge jd-badge-new">&#10024; New sub-vertical suggested</span>'
        s_extra = (f'<div class="jd-suggest">Suggested name: '
                   f'<span class="jd-chip">{r["suggested_sub_vertical"]}</span></div>')
    else:
        s_flag = '<span class="jd-badge jd-badge-ok">&#10003; Fits existing</span>'
        s_extra = ''
    s_card = (f'<div class="jd-card"><div class="jd-card-top"><span class="jd-eyebrow">SUB-VERTICAL</span>'
              f'{s_flag}</div><div class="jd-value jd-orange">{r["sub_vertical"]}</div>'
              f'{bar(sc*100, conf_color(sc))}<div class="jd-pct">{sc*100:.1f}% confidence</div>{s_extra}</div>')
    rows = ""
    for v, c in r["vertical_top3"]:
        rows += (f'<div class="jd-top-row"><div class="jd-top-name">{v}</div>'
                 f'<div class="jd-top-bar">{bar(c*100, JD_BLUE)}</div>'
                 f'<div class="jd-top-pct">{c*100:.1f}%</div></div>')
    top = f'<div class="jd-card"><div class="jd-eyebrow">TOP&nbsp;3 VERTICALS</div>{rows}</div>'
    return f'<div class="jd-results">{v_card}{s_card}{top}</div>'

EMPTY_HTML = ('<div class="jd-empty">Enter a category name above and click '
              '<b>Classify</b> to see its vertical, sub-vertical and confidence.</div>')

def _run(name, vthr, sthr):
    if not name or not name.strip():
        return EMPTY_HTML
    return render(classify(name, vert_thr=vthr, sub_thr=sthr))

run = spaces.GPU(_run)   # ZeroGPU-detectable; also works on CPU

def process_file(file, vthr, sthr, progress=gr.Progress()):
    if file is None:
        return None, ('<div class="jd-status jd-status-info">Upload an Excel/CSV file with a '
                      '<b>category_name</b> column, then click <b>Process file</b>.</div>')
    path = file if isinstance(file, str) else getattr(file, "name", None)
    try:
        df = pd.read_csv(path, dtype=str) if str(path).lower().endswith(".csv") \
             else pd.read_excel(path, dtype=str)
    except Exception as e:
        return None, f'<div class="jd-status jd-status-err">Could not read the file: {e}</div>'
    cols = {c.strip().lower(): c for c in df.columns}
    if "category_name" not in cols:
        return None, ('<div class="jd-status jd-status-err">No <b>category_name</b> column found. '
                      f'Columns present: {", ".join(map(str, df.columns))}</div>')
    names = df[cols["category_name"]].fillna("").astype(str)
    R = [classify(n, vert_thr=vthr, sub_thr=sthr)
         for n in progress.tqdm(names, desc="Classifying")]
    # keep every original column, then append the classification columns
    out = df.copy()
    out["vertical"] = [r["vertical"] for r in R]
    out["vertical_confidence_%"] = [round(r["vertical_confidence"]*100, 1) for r in R]
    out["vertical_flag"] = ["review" if r["vertical_review"] else "ok" for r in R]
    out["sub_vertical"] = [r["sub_vertical"] for r in R]
    out["sub_vertical_confidence_%"] = [round(r["sub_vertical_confidence"]*100, 1) for r in R]
    out["sub_vertical_status"] = ["new-suggested" if r["propose_new_sub_vertical"] else "fits-existing" for r in R]
    out["suggested_new_sub_vertical"] = [r["suggested_sub_vertical"] or "" for r in R]
    out_path = os.path.join(tempfile.gettempdir(), "classified_results.xlsx")
    out.to_excel(out_path, index=False)
    n = len(out); rev = int((out["vertical_flag"] == "review").sum())
    new = int((out["sub_vertical_status"] == "new-suggested").sum())
    status = (f'<div class="jd-status jd-status-ok">Processed <b>{n}</b> rows &nbsp;&middot;&nbsp; '
              f'{rev} flagged for vertical review &nbsp;&middot;&nbsp; {new} with a suggested new '
              f'sub-vertical. <br>Download the results file below.</div>')
    return out_path, status


HEADER = '''
<div class="jd-header">
  <div class="jd-brand"><span class="jd-just">Just</span><span class="jd-dial">dial</span></div>
  <div class="jd-divider"></div>
  <div class="jd-apptitle">AI Category Classifier</div>
  <span class="jd-live">&#9679; LIVE</span>
</div>
<div class="jd-stats">
  <div class="jd-stat"><div class="jd-stat-label">Verticals</div><div class="jd-stat-num">21</div></div>
  <div class="jd-stat"><div class="jd-stat-label">Sub-verticals</div><div class="jd-stat-num">122</div></div>
  <div class="jd-stat"><div class="jd-stat-label">Vertical accuracy</div><div class="jd-stat-num">97.1%</div></div>
  <div class="jd-stat"><div class="jd-stat-label">Exact match</div><div class="jd-stat-num">97.6%</div></div>
</div>
<div class="jd-sub">Type a category name to get its <b>vertical</b> (1 of 21) and <b>sub-vertical</b>,
each with a confidence score. Low-confidence verticals are flagged for review; when no existing
sub-vertical fits, a new one is proposed.</div>
'''

CSS = """

gradio-app, body {background:#ffffff !important;}
.gradio-container, .gradio-container.dark, body {
  --body-background-fill:#ffffff; --background-fill-primary:#ffffff;
  --background-fill-secondary:#f7f8fa; --block-background-fill:#ffffff;
  --block-border-color:#eef0f2; --block-label-text-color:#6b7280;
  --body-text-color:#1f2937; --body-text-color-subdued:#6b7280;
  --input-background-fill:#ffffff; --input-border-color:#dfe3e8;
  --border-color-primary:#eef0f2; --color-accent:#F58220;
  --link-text-color:#00457C; --neutral-50:#f7f8fa; color-scheme:light !important;}
.jd-blue{color:#00457C !important;} .jd-dial{color:#F58220 !important;}
.jd-just{color:#00457C !important;} .jd-orange{color:#F58220 !important;}
.gradio-container {background:#ffffff !important; max-width:960px !important; margin:0 auto !important;
  font-family:-apple-system,'Segoe UI',Roboto,Helvetica,Arial,sans-serif !important;}
.gradio-container::before {content:""; position:fixed; top:0; left:0; right:0; height:5px; z-index:50;
  background:linear-gradient(90deg,#ff5f6d,#ff8a5b,#ffc371);}
footer {display:none !important;}
.jd-header {display:flex; align-items:center; gap:14px; padding:26px 4px 6px;}
.jd-brand {font-size:30px; font-weight:800; letter-spacing:-.5px;}
.jd-just {color:#00457C;} .jd-dial {color:#F58220;}
.jd-divider {width:1px; height:26px; background:#e5e7eb;}
.jd-apptitle {font-size:20px; font-weight:600; color:#1f2937;}
.jd-live {margin-left:auto; color:#16a34a; font-weight:700; font-size:12px; letter-spacing:.5px;
  background:#eafaf0; border:1px solid #bbe9ca; padding:4px 10px; border-radius:20px;}
.jd-stats {display:grid; grid-template-columns:repeat(4,1fr); gap:14px; margin:18px 4px 6px;}
.jd-stat {background:#fff; border:1px solid #eef0f2; border-radius:12px; padding:14px 16px;
  box-shadow:0 1px 2px rgba(0,0,0,.03);}
.jd-stat-label {font-size:12px; color:#6b7280; margin-bottom:6px;}
.jd-stat-num {font-size:26px; font-weight:800; color:#1f2937;}
.jd-sub {color:#6b7280; font-size:14px; line-height:1.5; margin:12px 4px 8px;}
#jd-input textarea {border:1px solid #dfe3e8 !important; border-radius:10px !important; font-size:16px !important;}
#jd-input textarea:focus {border-color:#F58220 !important; box-shadow:0 0 0 3px rgba(245,130,32,.15) !important;}
#jd-btn {background:#F58220 !important; color:#fff !important; border:none !important; font-weight:700 !important;
  border-radius:10px !important; box-shadow:0 2px 6px rgba(245,130,32,.25) !important;}
#jd-btn:hover {background:#e2710f !important;}
.jd-results {display:flex; flex-direction:column; gap:14px; margin-top:6px;}
.jd-card {background:#fff; border:1px solid #eef0f2; border-radius:14px; padding:18px 20px;
  box-shadow:0 1px 3px rgba(0,0,0,.04);}
.jd-card-top {display:flex; align-items:center; justify-content:space-between; margin-bottom:8px;}
.jd-eyebrow {font-size:11px; font-weight:700; letter-spacing:1px; color:#9aa2ad;}
.jd-value {font-size:24px; font-weight:800; margin:2px 0 12px; text-transform:capitalize;}
.jd-blue {color:#00457C;} .jd-orange {color:#F58220;}
.jd-track {height:9px; background:#f0f2f5; border-radius:6px; overflow:hidden;}
.jd-fill {height:100%; border-radius:6px;}
.jd-pct {font-size:12px; color:#6b7280; margin-top:6px;}
.jd-badge {font-size:12px; font-weight:700; padding:4px 10px; border-radius:20px;}
.jd-badge-ok {background:#eafaf0; color:#16a34a; border:1px solid #bbe9ca;}
.jd-badge-warn {background:#fff5e6; color:#b45309; border:1px solid #fcd9a3;}
.jd-badge-new {background:#eef4ff; color:#00457C; border:1px solid #c3d7f7;}
.jd-suggest {margin-top:12px; font-size:13px; color:#374151;}
.jd-chip {background:#fff5e6; color:#b45309; border:1px solid #fcd9a3; padding:3px 10px;
  border-radius:8px; font-weight:700;}
.jd-top-row {display:flex; align-items:center; gap:12px; margin:9px 0;}
.jd-top-name {width:150px; font-size:13px; color:#374151; text-transform:capitalize;}
.jd-top-bar {flex:1;} .jd-top-pct {width:52px; text-align:right; font-size:12px; color:#6b7280; font-weight:600;}
.jd-empty {background:#fafbfc; border:1px dashed #dfe3e8; border-radius:14px; padding:26px;
  text-align:center; color:#6b7280; font-size:14px;}

.jd-section {font-size:16px; font-weight:800; color:#1f2937; margin:26px 4px 2px; padding-top:10px;
  border-top:1px solid #eef0f2;}
#jd-btn2 {background:#00457C !important; color:#fff !important; border:none !important; font-weight:700 !important;
  border-radius:10px !important; box-shadow:0 2px 6px rgba(0,69,124,.2) !important;}
#jd-btn2:hover {background:#013a68 !important;}
.jd-status {font-size:13px; border-radius:10px; padding:12px 14px; margin:8px 4px;}
.jd-status-info {background:#f7f8fa; border:1px solid #eef0f2; color:#374151;}
.jd-status-ok {background:#eff8f1; border:1px solid #bbe9ca; color:#166534;}
.jd-status-err {background:#fef2f2; border:1px solid #f3c0c0; color:#b91c1c;}
#jd-file, #jd-file *, #jd-fileout, #jd-fileout * {background:#ffffff !important;
  color:#374151 !important; -webkit-text-fill-color:#374151 !important;}
#jd-file, #jd-fileout {border:1px solid #dfe3e8 !important; border-radius:10px !important;}
#jd-file a, #jd-fileout a {color:#00457C !important; -webkit-text-fill-color:#00457C !important;}
#jd-file svg, #jd-fileout svg {color:#6b7280 !important; opacity:1 !important;}

/* --- force native components to light (input text, labels, examples) --- */
#jd-input textarea, #jd-input input {background:#ffffff !important; color:#1f2937 !important;
  -webkit-text-fill-color:#1f2937 !important; caret-color:#F58220 !important;}
.gradio-container label, .gradio-container .label-wrap, .gradio-container .label-wrap > span,
.gradio-container span[data-testid="block-info"], .gradio-container .head {
  color:#374151 !important; opacity:1 !important;}
.gradio-container .label-wrap svg, .gradio-container .icon {color:#374151 !important; opacity:1 !important;}
.gradio-container [class*="sample"] {background:#ffffff !important; color:#1f2937 !important;
  -webkit-text-fill-color:#1f2937 !important; border:1px solid #dfe3e8 !important;}
.gradio-container [class*="sample"]:hover {background:#fff5e6 !important; color:#b45309 !important;
  -webkit-text-fill-color:#b45309 !important; border-color:#fcd9a3 !important;}
.gradio-container .accordion, .gradio-container .accordion * {opacity:1 !important;}
"""

FORCE_LIGHT = """
() => {
  const u = new URL(window.location.href);
  if (u.searchParams.get('__theme') !== 'light') {
    u.searchParams.set('__theme', 'light');
    window.location.replace(u.href);
  }
}
"""

with gr.Blocks(title="JustDial AI Category Classifier", js=FORCE_LIGHT) as demo:
    gr.HTML(HEADER)
    with gr.Row():
        inp = gr.Textbox(show_label=False, placeholder="e.g. Bridal Makeup Artists",
                         elem_id="jd-input", scale=5, lines=1)
        btn = gr.Button("Classify", elem_id="jd-btn", scale=1)
    with gr.Accordion("Thresholds (advanced)", open=False):
        vthr = gr.Slider(0.0, 1.0, value=VERT_THR, step=0.05, label="Vertical review threshold")
        sthr = gr.Slider(0.0, 1.0, value=SUB_THR, step=0.05, label="New sub-vertical threshold")
    out = gr.HTML(EMPTY_HTML)
    btn.click(run, [inp, vthr, sthr], out)
    inp.submit(run, [inp, vthr, sthr], out)

    gr.HTML('<div class="jd-section">Bulk classify &mdash; upload an Excel/CSV</div>'
            '<div class="jd-sub" style="margin-top:-2px">File must have a <b>category_name</b> column. '
            'You get back vertical, sub-vertical and both confidence scores as a downloadable Excel.</div>')
    with gr.Row():
        fin = gr.File(label="Upload .xlsx / .csv", file_types=[".xlsx", ".xls", ".csv"],
                      type="filepath", elem_id="jd-file", scale=4)
        pbtn = gr.Button("Process file", elem_id="jd-btn2", scale=1)
    fstatus = gr.HTML('<div class="jd-status jd-status-info">Upload a file with a '
                      '<b>category_name</b> column, then click <b>Process file</b>.</div>')
    fout = gr.File(label="Download results (.xlsx)", elem_id="jd-fileout")
    pbtn.click(process_file, [fin, vthr, sthr], [fout, fstatus])

if __name__ == "__main__":
    demo.launch(css=CSS)