-
Notifications
You must be signed in to change notification settings - Fork 0
/
app.py
635 lines (575 loc) · 29.3 KB
/
app.py
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
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
import datetime
import uuid
from datetime import timezone, datetime
import streamlit as st
from elasticsearch import Elasticsearch
import os
import math
import tiktoken
from langchain_community.chat_models import BedrockChat, AzureChatOpenAI, ChatOllama
from langchain.schema import (
SystemMessage,
HumanMessage
)
import nltk
from nltk.tokenize import word_tokenize
import time
import boto3
import pandas as pd
# Initialise ES connection and configuration details
es = Elasticsearch(os.environ['elastic_url'], api_key=os.environ['elastic_api_key'])
source_index = os.environ['default_index']
logging_index = os.environ['logging_index']
source_pipeline = os.environ['default_pipeline']
logging_pipeline = os.environ['logging_pipeline']
benchmarking_index = os.environ['benchmarking_index']
benchmarking_qa_index = os.environ['benchmarking_qa_index']
# Handle some Streamlit setup
st.set_page_config(
page_title="RAG workbench",
page_icon="🧊",
layout="wide",
initial_sidebar_state="expanded",
)
# Initialise the model temperature session variable
if 'model_temp' not in st.session_state:
st.session_state['model_temp'] = 0
# Check the configured providers in the secrets file and create the mapping needed later on
model_provider_map = []
if 'openai_api_model' in os.environ:
azure_openai_list_entry = {
'model_name': os.environ['openai_api_model'],
'provider_name': 'Azure OpenAI',
'prompt': 0.06,
'response': 0.12
}
model_provider_map.append(azure_openai_list_entry)
if 'aws_model_id' in os.environ:
aws_bedrock_list_entry = {
'model_name': os.environ['aws_model_id'],
'provider_name': 'AWS Bedrock',
'prompt': 0.008,
'response': 0.024
}
model_provider_map.append(aws_bedrock_list_entry)
if 'ollama_chat_model' in os.environ:
ollama_list_entry = {
'model_name': os.environ['ollama_chat_model'],
'provider_name': 'Ollama',
'prompt': 0,
'response': 0
}
model_provider_map.append(ollama_list_entry)
# Initialise the explanation for the RAG pipelines and prompt techniques which get rendered in the main page
pattern_explainer_map = [
{
'pattern_name': 'zero-shot-rag',
'description': 'question --> hybrid search --> prompt with context --> response --> output'
},
{
'pattern_name': 'few-shot-rag',
'description': 'question --> hybrid search --> prompt with examples + context --> response --> output'
},
{
'pattern_name': 'reflection-rag',
'description': 'question --> hybrid search --> prompt with context --> response --> reflect --> '
're-prompt with reflection notes --> response --> output'
},
{
'pattern_name': 'auto-prompt-engineer',
'description': 'question --> hybrid search --> design prompt + context --> prompt + context --> response --> output'
},
]
# --------------- UTILITY FUNCTIONS ----------------
# Calculate the number of tokens from a string of characters
def num_tokens_from_string(string: str, encoding_name: str) -> int:
"""Returns the number of tokens in a text string."""
encoding = tiktoken.get_encoding(encoding_name)
num_tokens = len(encoding.encode(string))
return num_tokens
# Find the provider from the model name
def get_provider_from_model(model_name):
for p in model_provider_map:
if p['model_name'] == model_name:
provider = p['provider_name']
return provider
# Find the description from the pattern name
def get_pattern_from_name(pattern_name):
for p in pattern_explainer_map:
if p['pattern_name'] == pattern_name:
description = p['description']
return description
# With this function we trim the context to fit into the LLMs context window
def truncate_text(text, max_tokens):
nltk.download('punkt')
tokens = word_tokenize(text)
trimmed_text = ' '.join(tokens[:max_tokens])
return trimmed_text
# Calculate the cost of the prompt or response for use in the logging of the LLM interaction
def calculate_cost(message, type):
for p in model_provider_map:
if p['provider_name'] == st.session_state.provider_name:
cost_per_1k = p[type]
message_token_count = num_tokens_from_string(message, "cl100k_base")
billable_message_tokens = message_token_count / 1000
rounded_up_message_tokens = math.ceil(billable_message_tokens)
message_cost = rounded_up_message_tokens * cost_per_1k
return message_cost
# --------------- RAG FLOW FUNCTIONS ----------------
# Invoke the LLM object so that we can pass it a prompt
def llm_response(provider_name):
if provider_name == 'Azure OpenAI':
llm = AzureChatOpenAI(
openai_api_base=os.environ['openai_api_base'],
openai_api_version=os.environ['openai_api_version'],
deployment_name=os.environ['deployment_name'],
openai_api_key=os.environ['openai_api_key'],
openai_api_type="azure",
temperature=st.session_state.model_temp,
streaming=True
)
elif provider_name == 'AWS Bedrock':
bedrock_client = boto3.client(service_name="bedrock-runtime", region_name=os.environ['aws_region'],
aws_access_key_id=os.environ['aws_access_key'],
aws_secret_access_key=os.environ['aws_secret_key'])
llm = BedrockChat(
client=bedrock_client,
model_id=os.environ['aws_model_id'],
streaming=True,
model_kwargs={"temperature": st.session_state.model_temp})
elif provider_name == 'Ollama':
llm = ChatOllama(model=os.environ['ollama_chat_model'],
temperature=st.session_state.model_temp)
return llm
# Connect the LLM and the prompt together and receive the entire answer before proceeding
def bulk_response(llm, prompt):
answer = llm.invoke(prompt).content
return answer
# Connect the LLM and the prompt together and yield responses uniformly in 2ms intervals
def yield_response(llm, prompt):
for word in llm.invoke(prompt).content.split(" "):
yield word + " "
time.sleep(0.02)
# aggregate the names of all reports stored in the index
def get_sources(index):
aggregation_query = {
"size": 0,
"query": {
"match_all": {
}
},
"aggs": {
"sources": {
"terms": {
"field": "source_name.keyword",
"size": 1000
}
}
}
}
sources = es.search(index=index, body=aggregation_query)
buckets = sources['aggregations']['sources']['buckets']
source_list = []
for bucket in buckets:
key = bucket['key']
source_list.append(key)
return source_list
# hybrid search operation combining a spare vector sematic search with lexical keyword search and a hard filter to search only data withih the selected document
def report_search(index, question, source_name):
model_id = os.environ['transformer_model']
query = {
"rrf": {
"retrievers": [
{
"standard": {
"filter": {
"term": {
"source_name.keyword": source_name
}
}
}
},
{
"standard": {
"query": {
"term": {
"text": question
}
}
}
},
{
"standard": {
"query": {
"text_expansion": {
"expanded_text": {
"model_id": model_id,
"model_text": question
}
}
}
}
}
],
"rank_window_size": 20,
"rank_constant": 1
}
}
field_list = ['page', 'text', '_score']
results = es.search(index=index, retriever=query, size=10, fields=field_list)
response_data = [{"_score": hit["_score"], **hit["_source"]} for hit in results["hits"]["hits"]]
documents = []
# Check if there are hits
if "hits" in results and "total" in results["hits"]:
total_hits = results["hits"]["total"]
# Check if there are any hits with a value greater than 0
if isinstance(total_hits, dict) and "value" in total_hits and total_hits["value"] > 0:
for hit in response_data:
doc_data = {field: hit[field] for field in field_list if field in hit}
documents.append(doc_data)
return documents
# We use the search results to create the context for the LLM, eliminating unnecessary fields and lower relevance results
def create_context_docs(results):
for record in results:
if "_score" in record:
del record["_score"]
truncated_results = results[:5]
result = ""
for item in truncated_results:
result += f"Page {item['page']} : {item['text']}\n"
result = result.replace("{", "").replace("}", "")
context_documents = truncate_text(result, 10000)
return context_documents
# Once we've gathered all the inputs and the context and construct a prompt based on the chosen pattern
def prompt_builder(question, pattern, results=None, answer=None, reflection=None):
if results:
context_documents = create_context_docs(results)
if pattern == 'zero-shot-rag':
prompt_file = 'prompts/generic_rag_prompt.txt'
with open(prompt_file, "r") as file:
prompt_contents_template = file.read()
prompt = prompt_contents_template.format(question=question, context_documents=context_documents)
augmented_prompt = prompt
elif pattern == 'few-shot-rag':
prompt_file = 'prompts/few_shot_rag_prompt.txt'
with open(prompt_file, "r") as file:
prompt_contents_template = file.read()
prompt = prompt_contents_template.format(question=question, context_documents=context_documents)
augmented_prompt = prompt
elif pattern == 'reflection-rag':
prompt_file = 'prompts/reflection_rag_prompt.txt'
with open(prompt_file, "r") as file:
prompt_contents_template = file.read()
prompt = prompt_contents_template.format(question=question, answer=answer,
context_documents=context_documents)
augmented_prompt = prompt
elif pattern == 'guided-rag':
prompt_file = 'prompts/generic_rag_with_reflection_prompt.txt'
with open(prompt_file, "r") as file:
prompt_contents_template = file.read()
prompt = prompt_contents_template.format(question=question, context_documents=context_documents,
reflection=reflection, answer=answer)
augmented_prompt = prompt
elif pattern == 'auto-prompt-engineer':
prompt_file = 'prompts/auto_prompt_engineer.txt'
with open(prompt_file, "r") as file:
prompt_contents_template = file.read()
prompt = prompt_contents_template.format(question=question, context=context_documents)
augmented_prompt = prompt
elif pattern == 'generated-prompt':
prompt_file = 'prompts/guided-rag-with-generated-prompt.txt'
with open(prompt_file, "r") as file:
prompt_contents_template = file.read()
prompt = prompt_contents_template.format(question=question, context_documents=context_documents,
reflection=reflection)
augmented_prompt = prompt
messages = [
SystemMessage(
content="You are a question/answer assistant. Your job is to answer user questions, using context "
"provided in this prompt. The context is the result of a sematic search on a document "
"and therefore represents extracts from the document that contains the information you need to "
"answer the question"),
HumanMessage(content=augmented_prompt)
]
return messages
# We use this function to write the LLM interaction to an Elasticsearch logging index
def log_llm_interaction(question, prompt, response, sent_time, received_time, report_name):
log_id = uuid.uuid4()
dt_latency = received_time - sent_time
actual_latency = dt_latency.total_seconds()
str_prompt = str(prompt)
body = {
"@timestamp": datetime.now(tz=timezone.utc),
"report_name": report_name,
"question": question,
"answer": response,
"provider": st.session_state.provider_name,
"model": st.session_state.model_name,
"model_temp": st.session_state.model_temp,
"timestamp_sent": sent_time,
"timestamp_received": received_time,
"prompt_cost": calculate_cost(str_prompt, 'prompt'),
"response_cost": calculate_cost(response, 'response'),
"llm_latency": actual_latency,
"pattern_name": st.session_state['pattern_name']
}
es.index(index=logging_index, id=log_id, document=body)
return
# This function logs the generated answer from an LLM which we will evaluate and generate metrics from
def log_benchmark_test(question, ground_truth, context, answer, report_name):
string_context = create_context_docs(context)
log_id = uuid.uuid4()
body = {
"@timestamp": datetime.now(tz=timezone.utc),
"report_name": report_name,
"question": question,
"ground_truth": ground_truth,
"answer": answer,
"contexts": string_context,
"provider": st.session_state.provider_name,
"model": st.session_state.model_name,
"pattern_name": st.session_state['pattern_name']
}
es.index(index=benchmarking_index, id=log_id, document=body)
return
# Based on the datasource/document chosen, we pull the set of questions for benchmarking
def get_questions_answers(index, source):
query = {
"match": {
"source_name": source
}
}
field_list = ['question', 'ground_truth']
results = es.search(index=index, query=query, size=100, fields=field_list, track_scores=True)
response_data = [{"_id": hit["_id"], "_score": hit["_score"], **hit["_source"]} for hit in results["hits"]["hits"]]
documents = []
# Check if there are hits
if "hits" in results and "total" in results["hits"]:
total_hits = results["hits"]["total"]
# Check if there are any hits with a value greater than 0
if isinstance(total_hits, dict) and "value" in total_hits and total_hits["value"] > 0:
for hit in response_data:
doc_data = {field: hit[field] for field in field_list if field in hit}
doc_data["_id"] = hit["_id"] # Include the document ID in the document data
documents.append(doc_data)
return documents
model_list = []
for m in model_provider_map:
model_name = m['model_name']
model_list.append(model_name)
pattern_list = []
for p in pattern_explainer_map:
pattern_name = p['pattern_name']
pattern_list.append(pattern_name)
if 'pattern_name' and 'model_name' not in st.session_state:
st.session_state.pattern_name = pattern_list[0]
st.session_state.model_name = model_list[0]
# Run benchmarking questions through the chosen RAG pipeline
def execute_benchmark(questions_answers):
total_count = len(questions_answers)
counter = 1
with st.status("looping through questions", expanded=True) as status:
for qa in questions_answers:
question = qa['question']
ground_truth = qa['ground_truth']
status_label = f"processing question {counter} of {total_count}: {question}"
status.update(label=status_label, state="running")
results = report_search(source_index, question, report_source)
if st.session_state.pattern_name == 'zero-shot-rag' or st.session_state.pattern_name == 'few-shot-rag':
prompt_construct = prompt_builder(question=question, results=results,
pattern=st.session_state.pattern_name)
llm = llm_response(st.session_state.provider_name)
sent_time = datetime.now(tz=timezone.utc)
answer = bulk_response(llm=llm, prompt=prompt_construct)
received_time = datetime.now(tz=timezone.utc)
log_llm_interaction(question, prompt_construct, answer, sent_time, received_time,
report_source)
log_benchmark_test(question=question, ground_truth=ground_truth, context=results,
report_name=report_source,
answer=answer)
elif st.session_state.pattern_name == 'reflection-rag':
prompt_construct1 = prompt_builder(question=question, results=results,
pattern=st.session_state.pattern_name)
llm = llm_response(st.session_state.provider_name)
sent_time = datetime.now(tz=timezone.utc)
answer1 = bulk_response(llm=llm, prompt=prompt_construct1)
received_time = datetime.now(tz=timezone.utc)
log_llm_interaction(question, prompt_construct1, answer1, sent_time, received_time,
report_source)
prompt_construct2 = prompt_builder(question=question, results=results, pattern='reflection-rag',
answer=answer1)
sent_time = datetime.now(tz=timezone.utc)
answer2 = bulk_response(llm=llm, prompt=prompt_construct2)
received_time = datetime.now(tz=timezone.utc)
log_llm_interaction(question, prompt_construct2, answer2, sent_time, received_time,
report_source)
prompt_construct3 = prompt_builder(question=question, results=results, pattern='guided-rag',
answer=answer1, reflection=answer2)
sent_time = datetime.now(tz=timezone.utc)
answer3 = bulk_response(llm=llm, prompt=prompt_construct3)
received_time = datetime.now(tz=timezone.utc)
log_llm_interaction(question, prompt_construct3, answer3, sent_time, received_time,
report_source)
log_benchmark_test(question=question, ground_truth=ground_truth, context=results,
report_name=report_source,
answer=answer3)
elif st.session_state.pattern_name == 'auto-prompt-engineer':
prompt_construct1 = prompt_builder(question=question, pattern=st.session_state.pattern_name,
results=results)
llm = llm_response(st.session_state.provider_name)
sent_time = datetime.now(tz=timezone.utc)
answer1 = bulk_response(llm=llm, prompt=prompt_construct1)
received_time = datetime.now(tz=timezone.utc)
log_llm_interaction(question, prompt_construct1, answer1, sent_time, received_time,
report_source)
prompt_construct2 = prompt_builder(question=question, pattern='generated-prompt', results=results,
reflection=answer1)
sent_time = datetime.now(tz=timezone.utc)
answer2 = bulk_response(llm=llm, prompt=prompt_construct2)
received_time = datetime.now(tz=timezone.utc)
log_llm_interaction(question, prompt_construct2, answer2, sent_time, received_time,
report_source)
log_benchmark_test(question=question, ground_truth=ground_truth, context=results,
report_name=report_source,
answer=answer2)
counter = counter + 1
status.update(label="all questions processed", state="complete")
return
# Define the sidebar for this page
st.sidebar.page_link("app.py", label="Home")
st.sidebar.page_link("pages/import.py", label="Manage reports/documents")
st.sidebar.page_link("pages/benchmark_data_setup.py", label="Manage benchmark questions")
st.sidebar.page_link("pages/benchmark.py", label="Run a benchmark test")
st.sidebar.page_link("pages/setup.py", label="Setup your Elastic environment")
st.sidebar.page_link(os.environ['kibana_url'], label="Kibana")
st.title("RAG workbench")
# We now handle the form and layout and mix it in with some application logic (not ideal but this is Streamlit)
col1, col2 = st.columns([1, 3])
with col1:
st.session_state.pattern_name = st.selectbox('Choose a prompt template', pattern_list,
index=pattern_list.index(st.session_state.pattern_name))
st.session_state.model_name = st.selectbox('Choose a model', model_list,
index=model_list.index(st.session_state.model_name))
st.session_state.provider_name = get_provider_from_model(st.session_state.model_name)
report_source = st.selectbox("Choose your source document", get_sources(source_index))
model_temp_options = [i / 100 for i in range(0, 101, 5)]
st.session_state['model_temp'] = st.select_slider('Select your model temperature:', options=model_temp_options,
value=st.session_state.model_temp)
if report_source:
benchmark_questions = get_questions_answers(benchmarking_qa_index, report_source)
benchmark = st.button("Generate data for a benchmark test")
else:
benchmark = ""
with col2:
question = st.text_input("Search your document with a question")
submit = st.button("Run the RAG pipeline")
pattern_description = get_pattern_from_name(st.session_state.pattern_name)
st.markdown("*The following RAG pattern will be applied:*")
st.markdown(f"*{pattern_description}*")
if submit:
# Run the search for context
results = report_search(source_index, question, report_source)
# Write the results to a dataframe so that they can be presented as supporting results below the LLM response
df_results = pd.DataFrame(results)
# Connect to the relevant LLM
llm = llm_response(st.session_state.provider_name)
# Execute the relevant prompt pattern
if results:
if st.session_state.pattern_name == 'zero-shot-rag':
st.write("assistant: 🤖")
with st.status("reaching out to the llm...", expanded=True) as status:
prompt_construct = prompt_builder(question=question, results=results,
pattern=st.session_state.pattern_name)
sent_time = datetime.now(tz=timezone.utc)
response = st.write_stream(
yield_response(llm=llm, prompt=prompt_construct))
received_time = datetime.now(tz=timezone.utc)
status.update(label="response generated", state="complete")
# Log the interaction
log_llm_interaction(question, prompt_construct, response, sent_time, received_time,
report_source)
elif st.session_state.pattern_name == 'few-shot-rag':
st.write("assistant: 🤖")
with st.status("reaching out to the llm...", expanded=True) as status:
prompt_construct = prompt_builder(question=question, results=results,
pattern=st.session_state.pattern_name)
sent_time = datetime.now(tz=timezone.utc)
response = st.write_stream(
yield_response(llm=llm, prompt=prompt_construct))
received_time = datetime.now(tz=timezone.utc)
status.update(label="response generated", state="complete")
# Log the interaction
log_llm_interaction(question, prompt_construct, response, sent_time, received_time,
report_source)
elif st.session_state.pattern_name == 'reflection-rag':
# initiate first prompt
prompt_construct = prompt_builder(question=question, results=results, pattern='zero-shot-rag')
st.write("assistant: 🤖")
with st.status("generating the initial response...", expanded=False) as status:
sent_time = datetime.now(tz=timezone.utc)
response1 = st.write_stream(
yield_response(llm=llm, prompt=prompt_construct))
received_time = datetime.now(tz=timezone.utc)
status.update(label="response generated", state="complete")
# Log the interaction
log_llm_interaction(question, prompt_construct, response1, sent_time, received_time,
report_source)
# now evaluate the original prompt
prompt_construct = prompt_builder(question=question, results=results, pattern='reflection-rag',
answer=response1)
# output the editorial response
st.write("editor:✍️")
with st.status("reviewing the initial response...", expanded=False) as status:
sent_time = datetime.now(tz=timezone.utc)
response2 = st.write_stream(
yield_response(llm=llm, prompt=prompt_construct))
received_time = datetime.now(tz=timezone.utc)
status.update(label="response reviewed", state="complete")
# Log the interaction
log_llm_interaction(question, prompt_construct, response2, sent_time, received_time,
report_source)
# now process the recommendations
prompt_construct = prompt_builder(question=question, results=results, pattern='guided-rag',
answer=response1, reflection=response2)
# output the final resposne
st.write("assistant: 🤖")
with st.status("updating the response...", expanded=False) as status:
sent_time = datetime.now(tz=timezone.utc)
response3 = st.write_stream(
yield_response(llm=llm, prompt=prompt_construct))
received_time = datetime.now(tz=timezone.utc)
status.update(label="response completed", state="complete")
# Log the interaction
log_llm_interaction(question, prompt_construct, response3, sent_time, received_time,
report_source)
elif st.session_state.pattern_name == 'auto-prompt-engineer':
st.write("prompt engineer: 🤖")
prompt_construct = prompt_builder(question=question, pattern=st.session_state.pattern_name,
results=results)
# output the generated prompt
with st.status("building a prompt...", expanded=False) as status:
sent_time = datetime.now(tz=timezone.utc)
response1 = st.write_stream(
yield_response(llm=llm, prompt=prompt_construct))
received_time = datetime.now(tz=timezone.utc)
# Log the interaction
log_llm_interaction(question, prompt_construct, response1, sent_time, received_time,
report_source)
st.write("assistant: 🤖")
prompt_construct = prompt_builder(question=question, pattern='generated-prompt', results=results,
reflection=response1)
# output the final answer
with st.status("attempting to answer the question...", expanded=True) as status:
sent_time = datetime.now(tz=timezone.utc)
response2 = st.write_stream(
yield_response(llm=llm, prompt=prompt_construct))
received_time = datetime.now(tz=timezone.utc)
# Log the interaction
log_llm_interaction(question, prompt_construct, response2, sent_time, received_time,
report_source)
status.update(label="response generated", state="complete", expanded=True)
st.dataframe(df_results)
else:
st.write("your search yielded zero results")
elif benchmark:
execute_benchmark(benchmark_questions)