File size: 11,081 Bytes
d7a40b3
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
from __future__ import annotations

import sys
from pathlib import Path

import gradio as gr

sys.path.insert(0, str(Path(__file__).parent))

from ui import components as ui
from ui.controller import (
    OUTPUT_FIELDS,
    PRECISION_CHOICES,
    PRECISION_EXACT,
    DashboardController,
    empty_view,
)
from ui.theme import HEAD, build_css, build_theme

EXAMPLES_PATH = Path(__file__).parent / "examples.csv"
FOOTER = (
    "FLAIRD is a decision-support tool. Scores are prior-adjusted model outputs, not calibrated "
    "probabilities, and explanations describe the model's behavior, not ground truth."
)


def _card(title: str, subtitle: str = "") -> gr.Column:
    """Open a white card with a title; use as a context manager."""
    column = gr.Column(elem_classes="fx-card")
    with column:
        gr.HTML(
            f'<div class="fx-title">{title}</div>'
            + (f'<div class="fx-sub">{subtitle}</div>' if subtitle else "")
        )
    return column


def build_app(controller: DashboardController | None = None) -> gr.Blocks:
    controller = controller or DashboardController(examples_path=EXAMPLES_PATH)
    initial = empty_view()
    components: dict[str, gr.components.Component] = {}

    with gr.Blocks(title="FLAIRD - AI-text forensic analysis", fill_width=True) as app:
        gr.HTML(ui.header_html(), elem_classes="fx-bleed")
        components["hero"] = gr.HTML(initial.hero, elem_classes="fx-bleed")

        with gr.Column(elem_classes="fx-page"):
            components["notice"] = gr.HTML(initial.notice)
            with gr.Tabs(elem_classes="fx-tabs"):
                # ------------------------------------------------------------ Dashboard
                with gr.Tab("Dashboard"):
                    with _card(
                        "Analyze a text",
                        "Paste a text of at least a few sentences, or pick an example.",
                    ):
                        with gr.Row(elem_classes="fx-row"):
                            with gr.Column(scale=3, min_width=340):
                                text = gr.Textbox(
                                    label="Text to analyze",
                                    lines=9,
                                    max_lines=14,
                                    placeholder="Paste the text you want to check ...",
                                )
                            with gr.Column(scale=2, min_width=300):
                                model = gr.Dropdown(
                                    choices=list(controller.models),
                                    value=next(iter(controller.models)),
                                    label="Model",
                                    info="Both are the released multitask checkpoints.",
                                )
                                precision = gr.Radio(
                                    PRECISION_CHOICES,
                                    value=PRECISION_EXACT,
                                    label="Feature attribution",
                                    info="Exact enumerates ~78,000 combinations; the sampled estimate is faster on CPU.",
                                )
                                example = gr.Dropdown(
                                    choices=controller.example_choices(),
                                    value=None,
                                    label="Example",
                                    info="Loads a text from the project's example set.",
                                )
                        example_chip = gr.HTML("")
                        with gr.Row():
                            analyze = gr.Button("Analyze", variant="primary", scale=3)
                            clear = gr.Button("Clear", scale=1)

                    with gr.Row(elem_classes="fx-row"):
                        with _card(
                            "Path balance",
                            "How meaning and writing style contributed to the decision",
                        ):
                            components["balance_panel"] = gr.HTML(initial.balance_panel)
                            components["balance_plot"] = gr.Plot(
                                initial.balance_plot, show_label=False, container=False
                            )
                        with _card(
                            "Generator distribution",
                            "Estimated model family of the text",
                        ):
                            components["generator_panel"] = gr.HTML(
                                initial.generator_panel
                            )
                            components["generator_plot"] = gr.Plot(
                                initial.generator_plot,
                                show_label=False,
                                container=False,
                            )

                    with gr.Row(elem_classes="fx-row"):
                        with _card("Detection score"):
                            components["score_plot"] = gr.Plot(
                                initial.score_plot, show_label=False, container=False
                            )
                            components["score_meta"] = gr.HTML(initial.score_meta)
                        with _card("Input reliability"):
                            components["reliability_plot"] = gr.Plot(
                                initial.reliability_plot,
                                show_label=False,
                                container=False,
                            )
                            components["reliability_meta"] = gr.HTML(
                                initial.reliability_meta
                            )
                        with _card("Fusion gate"):
                            components["gate_plot"] = gr.Plot(
                                initial.gate_plot, show_label=False, container=False
                            )
                            components["gate_meta"] = gr.HTML(initial.gate_meta)

                # ------------------------------------------------------------ Reasoning
                with gr.Tab("Reasoning"):
                    with _card(
                        "From semantic evidence to the final decision",
                        "Each step shows how a feature group moved the machine logit",
                    ):
                        components["waterfall_plot"] = gr.Plot(
                            initial.waterfall_plot, show_label=False, container=False
                        )
                        components["waterfall_caption"] = gr.HTML(
                            initial.waterfall_caption
                        )
                    with gr.Row(elem_classes="fx-row"):
                        with _card(
                            "Groups and features",
                            "Inner ring: groups. Outer ring: features, sized by absolute contribution",
                        ):
                            components["sunburst_plot"] = gr.Plot(
                                initial.sunburst_plot, show_label=False, container=False
                            )
                            components["sunburst_caption"] = gr.HTML(
                                initial.sunburst_caption
                            )
                        with _card(
                            "Attention versus contribution",
                            "Where the model looked compared with what changed the decision",
                        ):
                            components["attention_plot"] = gr.Plot(
                                initial.attention_plot,
                                show_label=False,
                                container=False,
                            )
                            components["attention_caption"] = gr.HTML(
                                initial.attention_caption
                            )

                # ------------------------------------------------------------ Forensic evidence
                with gr.Tab("Forensic Evidence"):
                    with (
                        gr.Row(elem_classes="fx-row"),
                        _card(
                            "Feature groups",
                            "Positive pushes toward machine, negative toward human",
                        ),
                    ):
                        components["groups_plot"] = gr.Plot(
                            initial.groups_plot, show_label=False, container=False
                        )
                        components["groups_caption"] = gr.HTML(initial.groups_caption)
                    with _card(
                        "Top forensic indicators",
                        "Ranked by absolute contribution to the decision",
                    ):
                        components["indicators"] = gr.HTML(initial.indicators)
                    with _card("All 35 features", "Click a column header to sort"):
                        components["features_table"] = gr.Dataframe(
                            value=initial.features_table,
                            interactive=False,
                            wrap=True,
                            show_label=False,
                            max_height=520,
                            column_widths=[
                                "70px",
                                "210px",
                                "190px",
                                "90px",
                                "120px",
                                "100px",
                                "380px",
                            ],
                        )

                # ------------------------------------------------------------ Method
                with gr.Tab("Method"), gr.Column(elem_classes="fx-card"):
                    gr.Markdown(ui.METHOD_MARKDOWN, elem_classes="fx-prose")

        gr.HTML(f'<div class="fx-footer">{FOOTER}</div>')

        outputs = [components[name] for name in OUTPUT_FIELDS]

        def run(text_value, model_label, precision_value):
            return controller.analyze(text_value, model_label, precision_value)

        analyze.click(run, [text, model, precision], outputs, show_progress="minimal")
        example.change(
            controller.load_example,
            example,
            [text, example_chip],
            show_progress="minimal",
        )
        clear.click(
            lambda: ("", None, "", *empty_view().as_tuple()),
            None,
            [text, example, example_chip, *outputs],
            show_progress="minimal",
        )
    return app


def main() -> None:
    app = build_app()
    app.queue(max_size=8).launch(theme=build_theme(), css=build_css(), head=HEAD)


if __name__ == "__main__":
    main()