File size: 7,083 Bytes
2cd4061 6dc1ca6 2cd4061 6dc1ca6 2cd4061 fd2037f 2cd4061 fd2037f 2cd4061 6dc1ca6 fd2037f 2cd4061 fd2037f 2cd4061 fd2037f 2cd4061 fd2037f 2cd4061 fd2037f 2cd4061 fd2037f 6dc1ca6 2cd4061 6dc1ca6 2cd4061 6dc1ca6 2cd4061 |
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 |
import gradio as gr
from huggingface_hub import InferenceClient
import json
from PIL import Image
from bs4 import BeautifulSoup
import requests
def extract_text_from_webpage(html_content):
soup = BeautifulSoup(html_content, 'html.parser')
for tag in soup(["script", "style", "header", "footer"]):
tag.extract()
return soup.get_text(strip=True)
def search(query):
term = query
all_results = []
max_chars_per_page = 8000
with requests.Session() as session:
resp = session.get(
url="https://www.google.com/search",
headers={"User-Agent": "Mozilla/5.0 (Windows NT 10.0; Win64; x64; rv:109.0) Gecko/20100101 Firefox/111.0"},
params={"q": term, "num": 3},
timeout=5
)
resp.raise_for_status()
soup = BeautifulSoup(resp.text, "html.parser")
result_block = soup.find_all("div", attrs={"class": "g"})
for result in result_block:
link = result.find("a", href=True)
link = link["href"]
try:
webpage = session.get(link, headers={"User-Agent": "Mozilla/5.0 (Windows NT 10.0; Win64; x64; rv:109.0) Gecko/20100101 Firefox/111.0"}, timeout=5)
webpage.raise_for_status()
visible_text = extract_text_from_webpage(webpage.text)
if len(visible_text) > max_chars_per_page:
visible_text = visible_text[:max_chars_per_page]
all_results.append({"link": link, "text": visible_text})
except requests.exceptions.RequestException:
all_results.append({"link": link, "text": None})
return all_results
# Initialize inference clients for different models
client_gemma = InferenceClient("mistralai/Mistral-7B-Instruct-v0.3")
client_mixtral = InferenceClient("NousResearch/Nous-Hermes-2-Mixtral-8x7B-DPO")
client_llama = InferenceClient("meta-llama/Meta-Llama-3-8B-Instruct")
# Define the main chat function
def respond(message, history):
func_caller = []
user_prompt = message
functions_metadata = [
{"type": "function", "function": {"name": "web_search", "description": "Search query on google", "parameters": {"type": "object", "properties": {"query": {"type": "string", "description": "web search query"}}, "required": ["query"]}}},
]
for msg in history:
func_caller.append({"role": "user", "content": f"{str(msg[0])}"})
func_caller.append({"role": "assistant", "content": f"{str(msg[1])}"})
message_text = message
func_caller.append({"role": "user", "content": f'[SYSTEM]You are a helpful assistant. You have access to the following functions: \n {str(functions_metadata)}\n\nTo use these functions respond with:\n<functioncall> {{ "name": "function_name", "arguments": {{ "arg_1": "value_1", "arg_1": "value_1", ... }} }} </functioncall> [USER] {message_text}'})
response = client_gemma.chat_completion(func_caller, max_tokens=200)
response = str(response)
try:
response = response[int(response.find("{")):int(response.rindex("</"))]
except:
response = response[int(response.find("{")):(int(response.rfind("}"))+1)]
response = response.replace("\\n", "")
response = response.replace("\\'", "'")
response = response.replace('\\"', '"')
response = response.replace('\\', '')
print(f"\n{response}")
try:
json_data = json.loads(str(response))
if json_data["name"] == "web_search":
query = json_data["arguments"]["query"]
web_results = search(query)
web2 = ' '.join([f"Link: {res['link']}\nText: {res['text']}\n\n" for res in web_results])
messages = f"system\nYou are OpenCHAT mini a helpful assistant made by KingNish. You are provided with WEB results from which you can find informations to answer users query in Structured and More better way. You do not say Unnecesarry things Only say thing which is important and relevant. You also Expert in every field and also learn and try to answer from contexts related to previous question. Try your best to give best response possible to user. You also try to show emotions using Emojis and reply like human, use short forms, friendly tone and emotions."
for msg in history:
messages += f"\nuser\n{str(msg[0])}"
messages += f"\nassistant\n{str(msg[1])}"
messages+=f"\nuser\n{message_text}\nweb_result\n{web2}\nassistant\n"
stream = client_mixtral.text_generation(messages, max_new_tokens=2000, do_sample=True, stream=True, details=True, return_full_text=False)
output = ""
for response in stream:
if not response.token.text == "":
output += response.token.text
yield output
else:
messages = f"system\nYou are OpenCHAT mini a helpful assistant made by KingNish. You answers users query like human friend. You are also Expert in every field and also learn and try to answer from contexts related to previous question. Try your best to give best response possible to user. You also try to show emotions using Emojis and reply like human, use short forms, friendly tone and emotions."
for msg in history:
messages += f"\nuser\n{str(msg[0])}"
messages += f"\nassistant\n{str(msg[1])}"
messages+=f"\nuser\n{message_text}\nassistant\n"
stream = client_llama.text_generation(messages, max_new_tokens=2000, do_sample=True, stream=True, details=True, return_full_text=False)
output = ""
for response in stream:
if not response.token.text == "":
output += response.token.text
yield output
except:
messages = f"system\nYou are OpenCHAT mini a helpful assistant made by KingNish. You answers users query like human friend. You are also Expert in every field and also learn and try to answer from contexts related to previous question. Try your best to give best response possible to user. You also try to show emotions using Emojis and reply like human, use short forms, friendly tone and emotions."
for msg in history:
messages += f"\nuser\n{str(msg[0])}"
messages += f"\nassistant\n{str(msg[1])}"
messages+=f"\nuser\n{message_text}\nassistant\n"
stream = client_llama.text_generation(messages, max_new_tokens=2000, do_sample=True, stream=True, details=True, return_full_text=False)
output = ""
for response in stream:
if not response.token.text == "":
output += response.token.text
yield output
# Define the Gradio demo
demo = gr.ChatInterface(
fn=respond,
chatbot=gr.Chatbot(show_copy_button=True, likeable=True, layout="panel"),
description="Ask anything and get responses based on web searches and AI models.",
textbox=gr.MultimodalTextbox(),
multimodal=True,
concurrency_limit=200,
)
# Launch the Gradio demo
demo.launch() |