Spaces:
Sleeping
Sleeping
File size: 5,653 Bytes
6855b1e 8922c23 f58ccf2 6855b1e 8922c23 6855b1e 843e3ae 6855b1e 843e3ae 6855b1e 3152165 6855b1e de33d74 6855b1e bf0597d 8922c23 6855b1e f58ccf2 bcae708 6855b1e f58ccf2 6855b1e 906b1c0 |
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 |
import logging
import os
from subprocess import check_call, STDOUT
from time import asctime
import gradio as gr
from llama_index.core import Document, VectorStoreIndex
from generate_response import generate_chat_response_with_history, set_llm, is_search_query, condense_question, \
generate_chat_response_with_history_rag_return_response
from web_search import search
API_KEY_PATH = "../keys/gpt_api_key.txt"
logger = logging.getLogger("agent_logger")
sourced = False
query = False
rag_similarity = False
def google_search_chat(message, history):
condensed_question = condense_question(message, history)
if is_search_query(condensed_question):
search_results = search(message, condensed_question)
relevant_content = ""
sources = ""
for index, result in enumerate(search_results):
relevant_content = relevant_content + "\n" + ''.join(result['text'])
sources = sources + f'\n {index + 1}. ' + result['url'] # python is zero-indexed
if relevant_content != "":
documents = [Document(text=relevant_content)]
index = VectorStoreIndex.from_documents(documents)
response = generate_chat_response_with_history_rag_return_response(index, message, history)
# similar_str = "not calculated"
# faithfulness_str = "not calculated"
#
# if rag_similarity:
# sim_evaluator = SemanticSimilarityEvaluator()
# faith_evaluator = FaithfulnessEvaluator(llm=get_llm())
# # condensed_context = condense_context(relevant_content)
# # logger.info("Calculating similarity...")
# # similar = sim_evaluator.evaluate(response=str(response),
# # reference=condensed_context)
# logger.info("Calculating faithfulness...")
# faithfulness = faith_evaluator.evaluate_response(query=condensed_question, response=response)
# # similar_str = str(round((similar.score * 100), 2)) + "%"
# faithfulness_str = "Yes" if faithfulness.passing else "No"
#
# logger.info(f'**Search Query:** {condensed_question} \n **Faithfulness:** {faithfulness_str} \n '
# f'**Similarity:** {similar_str} \n **Sources used:** \n {sources}')
response_text = []
string_output = ""
for text in response.response_gen:
response_text.append(text)
string_output = ''.join(response_text)
yield string_output
# if not sourced:
# pass
# if sourced and not query and not rag_similarity:
# yield string_output + f'\n\n --- \n **Sources used:** \n {sources}'
# if sourced and query and not rag_similarity:
# yield (string_output
# + f'\n\n --- \n **Search Query:** {condensed_question} '
# f'\n **Sources used:** \n {sources}')
# if rag_similarity:
# yield (string_output
# + f'\n\n --- \n **Search Query:** {condensed_question} \n '
# # f'**Similarity of response to the sources [ℹ️]'
# # f'(https://en.wikipedia.org/wiki/Semantic_similarity):** {similar_str} \n'
# f'**Is response in source documents?**: {faithfulness_str}'
# f'\n **Sources used:** \n {sources}')
logger.info(f'Assistant Response: {string_output}')
else:
logger.info(
f'Assistant Response: Sorry, no search results found.')
yield "Sorry, no search results found."
else:
yield from generate_chat_response_with_history(message, history)
if __name__ == '__main__':
logging.root.setLevel(logging.INFO)
filehandler = logging.FileHandler(f'agent_log_{asctime().replace(" ", "").lower().replace(":", "")}.log',
'a')
formatter = logging.Formatter('%(asctime)-15s::%(levelname)s::%(filename)s::%(funcName)s::%(lineno)d::%(message)s')
filehandler.setFormatter(formatter)
logger = logging.getLogger("agent_logger")
for hdlr in logger.handlers[:]: # remove the existing file handlers
if isinstance(hdlr, logging.FileHandler):
logger.removeHandler(hdlr)
logger.addHandler(filehandler) # set the new handler
logger.setLevel(logging.INFO)
logging.getLogger('selenium.webdriver.common').setLevel(logging.DEBUG)
api_key = os.getenv('gpt_api_key')
# GPT - 4 Turbo. The latest GPT - 4 model intended to reduce cases of “laziness” where the model doesn’t complete
# a task. Returns a maximum of 4,096 output tokens. Link:
# https://openai.com/blog/new-embedding-models-and-api-updates
set_llm(key=api_key, model="gpt-4-0125-preview", temperature=0)
# apt-get install -y libglib2.0-0=2.50.3-2 \
# libnss3=2:3.26.2-1.1+deb9u1 \
# libgconf-2-4=3.2.6-4+b1 \
# libfontconfig1=2.11.0-6.7+b1
# check_call(['apt-get', 'install', '-y', 'libglib2.0-0 libnss3 libgconf-2-4 libfontconfig1'],
# stdout=open(os.devnull, 'wb'), stderr=STDOUT)
logger.info("Launching Gradio ChatInterface for searchbot...")
demo = gr.ChatInterface(fn=google_search_chat,
title="Search Assistant", retry_btn=None, undo_btn=None, clear_btn=None,
theme="soft")
demo.launch(auth=('convo', 'session2024'))
|