File size: 10,494 Bytes
29cc4c5
90cb4f4
 
 
 
2b2f0d1
48fa8cf
6cdfa00
90cb4f4
 
 
2b2f0d1
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
b7a713a
 
46dae9a
b7a713a
 
a2b3c97
b7a713a
46dae9a
b7a713a
 
 
 
 
 
46dae9a
 
dd3763f
46dae9a
b7a713a
 
2b2f0d1
b7a713a
46dae9a
4f7c053
 
 
46dae9a
 
2b2f0d1
46dae9a
4f7c053
46dae9a
 
2b2f0d1
46dae9a
4f7c053
46dae9a
 
2b2f0d1
46dae9a
4f7c053
b7a713a
 
 
 
a2b3c97
 
 
 
 
 
b7a713a
 
 
 
 
 
 
 
 
 
 
4f7c053
a2b3c97
 
 
 
 
 
 
 
 
b7a713a
a2b3c97
 
 
 
 
b7a713a
 
 
48fa8cf
b7a713a
6cdfa00
b7a713a
 
 
0960577
 
 
 
a2b3c97
0960577
b7a713a
 
a2b3c97
4e9003c
a2b3c97
 
4e9003c
a2b3c97
0960577
b7a713a
 
 
934c580
 
0960577
6cdfa00
 
 
 
 
 
 
 
 
 
0960577
 
b7a713a
 
4f7c053
b7a713a
 
 
 
 
6cdfa00
4f7c053
b7a713a
 
 
 
 
 
 
6cdfa00
 
 
c2ca81b
1a117cc
 
6cdfa00
 
 
b7a713a
 
876e6bb
 
48fa8cf
b7a713a
6cdfa00
b7a713a
 
 
 
a2b3c97
b7a713a
 
a2b3c97
 
b7a713a
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
876e6bb
0759822
 
 
b7a713a
90cb4f4
b7a713a
0759822
 
c96708d
0759822
 
 
 
a2b3c97
dd7da04
48fa8cf
dd7da04
48fa8cf
 
0759822
 
 
b7a713a
48fa8cf
 
 
 
 
c96708d
0960577
 
0759822
 
 
0960577
0759822
 
 
29cc4c5
 
0960577
 
 
29cc4c5
 
0960577
 
 
b7a713a
29cc4c5
0759822
b7a713a
90cb4f4
46dae9a
 
 
 
0759822
 
 
 
 
 
29cc4c5
0759822
 
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
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
from db.schema import Response, ModelRatings
import streamlit as st
from datetime import datetime
from dotenv import load_dotenv
from views.nav_buttons import navigation_buttons
import random
from utils.loaders import load_html

load_dotenv()


def display_completion_message():
    """Display a standardized survey completion message."""
    st.markdown(
        """
        <div class='exit-container'>
            <h1>You have already completed the survey! Thank you for participating!</h1>
            <p>Your responses have been saved successfully.</p>
            <p>You can safely close this window or start a new survey.</p>
        </div>
        """,
        unsafe_allow_html=True,
    )
    st.session_state.show_questions = False
    st.session_state.completed = True
    st.session_state.start_new_survey = True


def get_previous_ratings(model_name, query_key, current_index):
    """Retrieve previous ratings from session state."""
    previous_ratings = {}

    if current_index < st.session_state.current_index and len(
            st.session_state.responses
    ) > current_index:
        if st.session_state.previous_ratings:
            previous_ratings = st.session_state.previous_ratings.get(
                st.session_state.data.iloc[current_index]["config_id"], {}
            )
            previous_ratings = previous_ratings.get(
                model_name, None
            )  # Fix: Model key from session state
    elif len(st.session_state.responses) <= current_index:
        previous_ratings = {}
    else:
        response_from_session = st.session_state.responses[current_index]
        try:
            previous_ratings = response_from_session.model_ratings.get(model_name, {})
        except AttributeError:
            previous_ratings = response_from_session["model_ratings"].get(model_name, {})

    stored_query_ratings = {}
    if previous_ratings:
        if "query_v" in query_key:
            try:
                stored_query_ratings = previous_ratings.query_v_ratings
            except AttributeError:
                stored_query_ratings = previous_ratings["query_v_ratings"]
        elif "query_p0" in query_key:
            try:
                stored_query_ratings = previous_ratings.query_p0_ratings
            except AttributeError:
                stored_query_ratings = previous_ratings["query_p0_ratings"]
        elif "query_p1" in query_key:
            try:
                stored_query_ratings = previous_ratings.query_p1_ratings
            except AttributeError:
                stored_query_ratings = previous_ratings["query_p1_ratings"]

    return stored_query_ratings if stored_query_ratings else {}


def render_single_rating(
        label,
        options,
        format_func,
        key_prefix,
        stored_rating,
        col,
):
    """Renders a single rating widget (radio)."""
    with col:
        return st.radio(
            label,
            options=options,
            format_func=format_func,
            key=f"{key_prefix}",
            index=stored_rating if stored_rating is not None else 0,
        )


def clean_query_text(query_text):
    """Clean the query text for display."""
    if query_text.startswith('"') or query_text.startswith("'") or query_text.endswith('"') or query_text.endswith("'"):
        query_text = query_text.replace('"', '').replace("'", "")
    if query_text[-1] not in [".", "?", "!", "\n"]:
        query_text += "."
    return query_text.capitalize()


def render_query_ratings(
        model_name,
        config,
        query_key,
        current_index,
        has_persona_alignment=False,
):
    """Helper function to render ratings for a given query."""
    stored_query_ratings = get_previous_ratings(model_name, query_key, current_index)
    stored_groundedness = stored_query_ratings.get("groundedness", 0)
    stored_clarity = stored_query_ratings.get("clarity", 0)
    stored_overall_rating = stored_query_ratings.get("overall", 0)
    stored_persona_alignment = (
        stored_query_ratings.get("persona_alignment", 0) if has_persona_alignment else 0
    )
    if model_name == "gemini":
        bg_color = "#e0f7fa"
    else:
        bg_color = "#f0f4c3"
    query_text = clean_query_text(config[model_name + "_" + query_key])
    with st.container():
        st.markdown(
            f"""
               <div style="background-color:{bg_color}; padding:1rem;">
                    <h3 style="text-align:left;color:black;">
                    {config.index.get_loc(model_name + "_" + query_key) - 5}
                    </h3> 
                   <p style="text-align:left;color:black;">
                   {query_text}</p>
                </div>
           """,
            unsafe_allow_html=True,
        )
        col_no = 4 if has_persona_alignment else 3
        cols = st.columns(col_no)
        options = [0, 1, 2, 3, 4]
        groundedness_rating = render_single_rating(
            "Groundedness:",
            options,
            lambda x: ["N/A", "Not Grounded", "Partially Grounded", "Grounded", "Unclear"][
                x
            ],
            f"rating_{model_name}{query_key}_groundedness_",
            stored_groundedness,
            cols[0],
        )
        persona_alignment_rating = None
        if has_persona_alignment:
            persona_alignment_rating = render_single_rating(
                "Persona Alignment:",
                options,
                lambda x: ["N/A", "Not Aligned", "Partially Aligned", "Aligned", "Unclear"][
                    x
                ],
                f"rating_{model_name}{query_key}_persona_alignment_",
                stored_persona_alignment,
                cols[1],
            )

        clarity_rating = render_single_rating(
            "Clarity:",
            [0, 1, 2, 3],
            lambda x: ["N/A", "Not Clear", "Somewhat Clear", "Very Clear"][x],
            f"rating_{model_name}{query_key}_clarity_",
            stored_clarity,
            cols[2] if has_persona_alignment else cols[1],
        )
        overall_rating = render_single_rating(
            "Overall Fit:",
            [0, 1, 2, 3],
            lambda x: ["N/A", "Poor", "Moderate", "Strong Fit"][x],
            f"rating_{model_name}{query_key}_overall_",
            stored_overall_rating,
            cols[3] if has_persona_alignment else cols[2],
        )

    return {
        "clarity": clarity_rating,
        "groundedness": groundedness_rating,
        "persona_alignment": persona_alignment_rating if has_persona_alignment else None,
        "overall": overall_rating,
    }


def display_ratings_row(model_name, config, current_index):
    # st.markdown(f"## {model_name.capitalize()} Ratings")

    cols = st.columns(3)
    # combinations = ["query_v", "query_p0", "query_p1"]
    # random.shuffle(combinations)
    with cols[0]:
        query_v_ratings = render_query_ratings(
            model_name,
            config,
            "query_v",
            current_index,
            has_persona_alignment=False,
        )
    with cols[1]:
        query_p0_ratings = render_query_ratings(
            model_name,
            config,
            "query_p0",
            current_index,
            has_persona_alignment=True,
        )
    with cols[2]:
        query_p1_ratings = render_query_ratings(
            model_name,
            config,
            "query_p1",
            current_index,
            has_persona_alignment=True,
        )
    if "persona_alignment" in query_v_ratings:
        query_v_ratings.pop("persona_alignment")
    return {
        "query_v_ratings": query_v_ratings,
        "query_p0_ratings": query_p0_ratings,
        "query_p1_ratings": query_p1_ratings,
    }


def questions_screen(data):
    """Display the questions screen with split layout."""
    current_index = st.session_state.current_index

    try:
        config = data.iloc[current_index]
        st.markdown(f"## Hello {st.session_state.username.capitalize()} πŸ‘‹")
        # Progress bar
        progress = (current_index + 1) / len(data)
        st.progress(progress)
        st.write(f"Question {current_index + 1} of {len(data)}")
        # st.subheader(f"Config ID: {config['config_id']}")
        st.markdown("### Instructions")
        instructions_html = load_html("static/instructions.html")
        with st.expander("Instructions", expanded=False):
            st.html(instructions_html)

        # Context information
        st.markdown("### Context Information")
        with st.expander("Persona", expanded=True):
            st.write(config["persona"])
        with st.expander("Filters", expanded=True):
            st.code(config["filters"], language="json")
            # st.write("**Cities:**", config["city"])
        # with st.expander("Full Context", expanded=False):
        #     st.text_area("", config["context"], height=300, disabled=False)
        st.markdown("### Rate the following queries based on the above context.")
        g_ratings = display_ratings_row("gemini", config, current_index)
        l_ratings = display_ratings_row("llama", config, current_index)

        # Additional comments
        comment = st.text_area("Additional Comments (Optional):")

        # Collecting the response data
        response = Response(
            config_id=config["config_id"],
            model_ratings={
                "gemini": ModelRatings(
                    query_v_ratings=g_ratings["query_v_ratings"],
                    query_p0_ratings=g_ratings["query_p0_ratings"],
                    query_p1_ratings=g_ratings["query_p1_ratings"],
                ),
                "llama": ModelRatings(
                    query_v_ratings=l_ratings["query_v_ratings"],
                    query_p0_ratings=l_ratings["query_p0_ratings"],
                    query_p1_ratings=l_ratings["query_p1_ratings"],
                ),
            },
            comment=comment,
            timestamp=datetime.now().isoformat(),
        )
        try:
            st.session_state.ratings[current_index] = response["model_ratings"]
        except TypeError:
            st.session_state.ratings[current_index] = response.model_ratings
        if len(st.session_state.responses) > current_index:
            st.session_state.responses[current_index] = response
        else:
            st.session_state.responses.append(response)

        # Navigation buttons
        navigation_buttons(data, response)
    except IndexError:
        print("Survey completed!")