File size: 16,802 Bytes
570952a
 
 
 
 
 
 
 
 
 
 
 
92707e3
570952a
 
92707e3
570952a
 
6329f74
570952a
 
 
 
 
 
 
 
 
 
 
 
 
 
 
996bf1f
570952a
 
 
 
996bf1f
570952a
a1793e4
570952a
7262aba
996bf1f
 
 
 
570952a
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
996bf1f
570952a
 
 
 
 
 
 
 
 
 
996bf1f
570952a
 
996bf1f
570952a
 
 
 
996bf1f
570952a
 
 
996bf1f
570952a
 
 
92707e3
570952a
92707e3
 
570952a
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
996bf1f
570952a
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
6329f74
996bf1f
 
6329f74
 
996bf1f
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
6329f74
996bf1f
570952a
6329f74
 
 
570952a
996bf1f
 
570952a
8d05292
6329f74
 
 
 
 
 
8d05292
 
 
 
 
 
6329f74
 
 
87a8721
6329f74
 
 
 
 
 
87a8721
 
 
6329f74
 
 
570952a
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
92707e3
570952a
 
 
 
 
 
 
 
 
92707e3
570952a
92707e3
570952a
 
 
 
 
 
996bf1f
 
 
570952a
 
92707e3
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
f361f6e
92707e3
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
570952a
 
 
 
92707e3
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
570952a
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "3bae1d7d-a2be-444d-97cc-d1cbf8843bf1",
   "metadata": {},
   "source": [
    "# Invisible RAG Pilot Demo App"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "2a8e18f7-cc88-4bbf-a6e1-095237ed7714",
   "metadata": {},
   "outputs": [],
   "source": [
    "import json\n",
    "import gspread\n",
    "import gradio as gr\n",
    "\n",
    "\n",
    "class RAGInterface:\n",
    "    \"\"\"\n",
    "    Setup the gradio app for loading/saving/syncronizing the mockup A/B evaluation RAG tasks.\n",
    "    The app is deployed on Hugging Face spaces at https://huggingface.co/spaces/sukiboo/invisible-rag-demo\n",
    "    \"\"\"\n",
    "\n",
    "    def __init__(self):\n",
    "        self.setup_interface()\n",
    "        self.launch_interface()\n",
    "\n",
    "    def setup_interface(self):\n",
    "        \"\"\"Configure the A/B Evaluation RAG task interface.\"\"\"\n",
    "        with gr.Blocks(title='Demo AB Evaluate RAG') as self.interface:\n",
    "\n",
    "            # protected fields\n",
    "            _task_id = gr.Textbox(label='Task ID', interactive=False, visible=False)\n",
    "\n",
    "            # task id and load/save/reset buttons\n",
    "            with gr.Row():\n",
    "                task_id = gr.Textbox(container=False, placeholder='Enter a task ID: 1--11', scale=9)\n",
    "                load_button = gr.Button('Load Task', scale=1)\n",
    "                save_button = gr.Button('Save Task', scale=1, variant='primary')\n",
    "                reset_button = gr.Button('Reset Task', scale=1, variant='stop')\n",
    "\n",
    "            # chat history and search results\n",
    "            chat = gr.Chatbot(height=700, layout='bubble', bubble_full_width=False, label='Chat History')\n",
    "            sources = gr.Markdown()\n",
    "\n",
    "            # model completions for answers 1 and 2\n",
    "            with gr.Row():\n",
    "                with gr.Column():\n",
    "                    answer1 = gr.Textbox(label='Answer 1', max_lines=50)\n",
    "                with gr.Column():\n",
    "                    answer2 = gr.Textbox(label='Answer 2', max_lines=50)\n",
    "\n",
    "            # individual ratings for answers 1 and 2\n",
    "            with gr.Row():\n",
    "                with gr.Column():\n",
    "                    groundedness1 = gr.Radio(label='Groundedness', choices=['Bad', 'Good', 'Perfect'])\n",
    "                    fluency1 = gr.Radio(label='Fluency', choices=['Bad', 'Good', 'Perfect'])\n",
    "                    utility1 = gr.Radio(label='Utility', choices=['Catastrophic', 'Bad', 'Good', 'Perfect'])\n",
    "                    notes1 = gr.Textbox(label='Notes', placeholder='N/A')\n",
    "                with gr.Column():\n",
    "                    groundedness2 = gr.Radio(label='Groundedness', choices=['Bad', 'Good', 'Perfect'])\n",
    "                    fluency2 = gr.Radio(label='Fluency', choices=['Bad', 'Good', 'Perfect'])\n",
    "                    utility2 = gr.Radio(label='Utility', choices=['Catastrophic', 'Bad', 'Good', 'Perfect'])\n",
    "                    notes2 = gr.Textbox(label='Notes', placeholder='N/A')\n",
    "\n",
    "            # overall rating\n",
    "            overall = gr.Radio(label='Overall Rating', choices=['#1 Better', 'Equally Bad', 'Equally Good', '#2 Better'])\n",
    "            notes = gr.Textbox(label='Notes', placeholder='A brief justification for the overall rating')\n",
    "\n",
    "            # input/output fields\n",
    "            answers = (answer1, answer2)\n",
    "            ratings1 = (groundedness1, fluency1, utility1, notes1)\n",
    "            ratings2 = (groundedness2, fluency2, utility2, notes2)\n",
    "            ratings = (*ratings1, *ratings2, overall, notes)\n",
    "\n",
    "            # button clicks\n",
    "            load_button.click(self.load_task, inputs=[task_id], outputs=[_task_id, chat, sources, *answers, *ratings])\n",
    "            save_button.click(self.save_task, inputs=[_task_id, *ratings], outputs=None)\n",
    "            reset_button.click(self.reset_task, inputs=[_task_id], outputs=[*ratings])\n",
    "\n",
    "    def load_task(self, task_id):\n",
    "        \"\"\"Load the task and parse the info.\"\"\"\n",
    "        task = self.read_task(task_id)\n",
    "        try:\n",
    "            id = task['id']\n",
    "            chat = task['chat_history'] + [[task['question'], task['search_query']]]\n",
    "            answers = [task['answer_1'], task['answer_2']]\n",
    "            sources = self.load_sources(task)\n",
    "            ratings = self.load_ratings(task)\n",
    "            gr.Info(f'Task demo_task_{task_id} is loaded!')\n",
    "            return id, chat, sources, *answers, *ratings\n",
    "        except:\n",
    "            raise gr.Error(f'Could not load the task demo_task_{task_id} :(')\n",
    "\n",
    "    def read_task(self, task_id):\n",
    "        \"\"\"Read the json task file.\"\"\"\n",
    "        try:\n",
    "            with open(f'./data/demo_task_{task_id}.json') as task_file:\n",
    "                task = json.load(task_file)\n",
    "            return task\n",
    "        except FileNotFoundError:\n",
    "            raise gr.Error(f'Task demo_task_{task_id} is not found :(')\n",
    "\n",
    "    def load_sources(self, task):\n",
    "        \"\"\"Parse the search results.\"\"\"\n",
    "        sources = []\n",
    "        for idx, source in enumerate(task['search_results']):\n",
    "            sources.append(f'##### {idx+1}. {source.replace(\"<\", f\"{chr(92)}<\")}\\n')\n",
    "        return '\\n---\\n'.join(['## Search Results'] + sources + ['']) if sources else ''\n",
    "\n",
    "    def load_ratings(self, task):\n",
    "        \"\"\"Parse the ratings for each answer.\"\"\"\n",
    "        # load ratings for answer 1\n",
    "        ratings1 = (task['ratings_1']['groundedness'],\n",
    "                    task['ratings_1']['fluency'],\n",
    "                    task['ratings_1']['utility'],\n",
    "                    task['ratings_1']['notes'])\n",
    "        # load ratings for answer 2\n",
    "        ratings2 = (task['ratings_2']['groundedness'],\n",
    "                    task['ratings_2']['fluency'],\n",
    "                    task['ratings_2']['utility'],\n",
    "                    task['ratings_2']['notes'])\n",
    "        # load overall ratings\n",
    "        overall = task['overall']\n",
    "        notes = task['notes']\n",
    "        return (*ratings1, *ratings2, overall, notes)\n",
    "\n",
    "    def save_task(self, task_id, *ratings):\n",
    "        \"\"\"Save the task into a new json file.\"\"\"\n",
    "        # load the original task\n",
    "        with open(f'./data/demo_task_{task_id}.json') as task_file:\n",
    "            task = json.load(task_file)\n",
    "        # parse the ratings\n",
    "        groundedness1, fluency1, utility1, notes1, \\\n",
    "        groundedness2, fluency2, utility2, notes2, \\\n",
    "        overall, notes = ratings\n",
    "        # update the ratings for answer 1\n",
    "        task['ratings_1']['groundedness'] = groundedness1\n",
    "        task['ratings_1']['fluency'] = fluency1\n",
    "        task['ratings_1']['utility'] = utility1\n",
    "        task['ratings_1']['notes'] = notes1\n",
    "        # update the ratings for answer 2\n",
    "        task['ratings_2']['groundedness'] = groundedness2\n",
    "        task['ratings_2']['fluency'] = fluency2\n",
    "        task['ratings_2']['utility'] = utility2\n",
    "        task['ratings_2']['notes'] = notes2\n",
    "        # update overall ratings\n",
    "        task['overall'] = overall\n",
    "        task['notes'] = notes\n",
    "        try:\n",
    "            # save the task to json file\n",
    "            with open(f'./data/demo_task_{task_id}.json', 'w', encoding='utf-8') as task_file:\n",
    "                json.dump(task, task_file, ensure_ascii=False, indent=4)\n",
    "            # save the task to google spreadsheet\n",
    "            self.save_gsheet(task_id, ratings)\n",
    "            gr.Info(f'Task demo_task_{task_id} is saved!')\n",
    "        except:\n",
    "            raise gr.Error(f'Could not save the task demo_task_{task_id} :(')\n",
    "\n",
    "    def reset_task(self, task_id):\n",
    "        \"\"\"Reset the task by erasing the ratings and operator notes.\"\"\"\n",
    "        # load the original task\n",
    "        with open(f'./data/demo_task_{task_id}.json') as task_file:\n",
    "            task = json.load(task_file)\n",
    "        # erase the ratings for answer 1\n",
    "        task['ratings_1']['groundedness'] = ''\n",
    "        task['ratings_1']['fluency'] = ''\n",
    "        task['ratings_1']['utility'] = ''\n",
    "        task['ratings_1']['notes'] = ''\n",
    "        # erase the ratings for answer 2\n",
    "        task['ratings_2']['groundedness'] = ''\n",
    "        task['ratings_2']['fluency'] = ''\n",
    "        task['ratings_2']['utility'] = ''\n",
    "        task['ratings_2']['notes'] = ''\n",
    "        # erase overall ratings\n",
    "        task['overall'] = ''\n",
    "        task['notes'] = ''\n",
    "        try:\n",
    "            # save the reset task to json file\n",
    "            with open(f'./data/demo_task_{task_id}.json', 'w', encoding='utf-8') as task_file:\n",
    "                json.dump(task, task_file, ensure_ascii=False, indent=4)\n",
    "            # save the reset task to google spreadsheet\n",
    "            self.reset_gsheet(task_id)\n",
    "            gr.Info(f'Task demo_task_{task_id} is reset!')\n",
    "        except:\n",
    "            raise gr.Error(f'Could not reset the task demo_task_{task_id} :(')\n",
    "        return '', '', '', '', '', '', '', '', '', ''\n",
    "\n",
    "    def save_gsheet(self, task_id, ratings):\n",
    "        \"\"\"Save the task to google spreadsheet.\"\"\"\n",
    "        try:\n",
    "            # configure gsheet credentials\n",
    "            gc = gspread.service_account('./gsheet_service_account.json')\n",
    "            sheet_id = '1D2sfE9YXKtd7cKlgalo5UnuNKC-GhxlGqHVYUlkQlCY'\n",
    "            sh = gc.open_by_key(sheet_id).worksheet('demo-app')\n",
    "            # update non-empty task ratings in the worksheet\n",
    "            cell_val = []\n",
    "            for idx, rating in enumerate([ratings[idx] for idx in [0,1,2,4,5,6,8,9]]):\n",
    "                if rating:\n",
    "                    cell_val.append({'range': f'{chr(67+idx)}{3+int(task_id)}', 'values': [[f'{rating}']]})\n",
    "            sh.batch_update(cell_val)\n",
    "        except:\n",
    "            gr.Warning(f'Could not save the task demo_task_{task_id} to the spreadsheet :(')\n",
    "\n",
    "    def reset_gsheet(self, task_id):\n",
    "        \"\"\"Reset the task ratings in google spreadsheet.\"\"\"\n",
    "        try:\n",
    "            # configure gsheet credentials\n",
    "            gc = gspread.service_account('./gsheet_service_account.json')\n",
    "            sheet_id = '1D2sfE9YXKtd7cKlgalo5UnuNKC-GhxlGqHVYUlkQlCY'\n",
    "            sh = gc.open_by_key(sheet_id).worksheet('demo-app')\n",
    "            # update task ratings in the worksheet if there are any (clearing empty values removes formatting)\n",
    "            if sh.get(f'C{3+int(task_id)}:J{3+int(task_id)}')[0]:\n",
    "                sh.batch_clear([f'C{3+int(task_id)}:J{3+int(task_id)}'])\n",
    "        except:\n",
    "            gr.Warning(f'Could not reset the task demo_task_{task_id} in the spreadsheet :(')\n",
    "\n",
    "    def launch_interface(self):\n",
    "        \"\"\"Launch the A/B Evaluation RAG task interface.\"\"\"\n",
    "        gr.close_all()\n",
    "        self.interface.queue(default_concurrency_limit=None)\n",
    "        self.interface.launch()\n",
    "\n",
    "\n",
    "rag = RAGInterface()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "ade1097d-35ce-4f7a-a689-1b51973cbc70",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "6707866e-8f1b-4bda-9b12-0008e289ab77",
   "metadata": {},
   "outputs": [],
   "source": [
    "# create placeholder tasks\n",
    "import os\n",
    "import json\n",
    "\n",
    "os.makedirs('./data/', exist_ok=True)\n",
    "for idx in range(1):\n",
    "    task = {\n",
    "        'id': f'{idx}',\n",
    "        'chat_history': [['user message 1', 'bot message 1'], ['user message 2', 'bot message 2']],\n",
    "        'question': 'question',\n",
    "        'search_query': 'search query',\n",
    "        'search_results': ['source 1', 'source 2', 'source 3'],\n",
    "        'answer_1': 'answer 1',\n",
    "        'answer_2': 'answer 2',\n",
    "        'ratings_1': {'groundedness': '', 'utility': '', 'fluency': '', 'notes': ''},\n",
    "        'ratings_2': {'groundedness': '', 'utility': '', 'fluency': '', 'notes': ''},\n",
    "        'overall': '',\n",
    "        'notes': ''\n",
    "    }\n",
    "    with open(f'./data/demo_task_{idx}.json', 'w', encoding='utf-8') as task_file:\n",
    "        json.dump(task, task_file, ensure_ascii=False, indent=4)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "e99c2d79-d544-4d30-ab22-6452385d3593",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "98f44682-2088-4925-8e2a-563197a50b66",
   "metadata": {},
   "outputs": [],
   "source": [
    "# make demo tasks from the csv of the spreadsheet\n",
    "# https://docs.google.com/spreadsheets/d/1kYW0cABv2C-mMmmw2Uc50mQC0MmOuoqKJQaBp7IyCho/edit#gid=1934745276\n",
    "import os\n",
    "import json\n",
    "import pandas as pd\n",
    "\n",
    "df = pd.read_csv('./dev.csv')\n",
    "df\n",
    "\n",
    "os.makedirs('./data/', exist_ok=True)\n",
    "for idx in range(len(df)):\n",
    "    row = df.iloc[idx]\n",
    "    task = {\n",
    "        'id': f'{idx+1}',\n",
    "        'chat_history': [],\n",
    "        'question': f'{row[\"question\"]}',\n",
    "        'search_query': '',\n",
    "        'search_results': [],\n",
    "        'answer_1': f'{row[\"answer_1\"]}',\n",
    "        'answer_2': f'{row[\"answer_2\"]}',\n",
    "        'ratings_1': {'groundedness': '', 'utility': '', 'fluency': '', 'notes': ''},\n",
    "        'ratings_2': {'groundedness': '', 'utility': '', 'fluency': '', 'notes': ''},\n",
    "        'overall': '',\n",
    "        'notes': ''\n",
    "    }\n",
    "\n",
    "    # chat history\n",
    "    try:\n",
    "        i = 1\n",
    "        while not pd.isna(row[f'user message {i}']):\n",
    "            task['chat_history'].append([row[f'user message {i}'], row[f'bot message {i}']])\n",
    "            i += 1\n",
    "    except:\n",
    "        pass\n",
    "\n",
    "    # search query\n",
    "    if not pd.isna(row['search_2']):\n",
    "        task['search_query'] = f'{row[\"search_1\"]}\\n{row[\"search_2\"]}'\n",
    "    else:\n",
    "        task['search_query'] = f'{row[\"search_1\"]}'\n",
    "\n",
    "    # search results\n",
    "    try:\n",
    "        i = 1\n",
    "        while not pd.isna(row[f'source {i}']):\n",
    "            task['search_results'].append(row[f'source {i}'])\n",
    "            i += 1\n",
    "    except:\n",
    "        pass\n",
    "\n",
    "    # save the task\n",
    "    with open(f'./data/demo_task_{idx+1}.json', 'w', encoding='utf-8') as task_file:\n",
    "        json.dump(task, task_file, ensure_ascii=False, indent=4)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "d293fcca-659d-41c5-b043-15fd2e57b216",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "9ab3763f-fa7b-406b-9bc4-22bc4f7a4ea3",
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "d5023979-626b-4135-8805-3de1a846586e",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.12.1"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}