drop garbage

This commit is contained in:
if 2023-04-29 12:53:35 +03:00
parent 3e0219b82b
commit 1ff262d106
6 changed files with 11 additions and 441 deletions

View File

@ -65,10 +65,10 @@ repos:
- id: python-use-type-annotations
- id: text-unicode-replacement-char
- repo: https://github.com/pre-commit/mirrors-mypy
rev: v1.2.0
hooks:
- id: mypy
additional_dependencies: [ types-PyYAML==6.0.10, types_requests==2.28.8, types-ujson==5.6.0 ]
args: [ --ignore-missing-imports, --warn-no-return, --warn-redundant-casts, --disallow-incomplete-defs ]
exclude: ^setup.py
# - repo: https://github.com/pre-commit/mirrors-mypy
# rev: v1.2.0
# hooks:
# - id: mypy
# additional_dependencies: [ types-PyYAML==6.0.10, types_requests==2.28.8, types-ujson==5.6.0 ]
# args: [ --ignore-missing-imports, --warn-no-return, --warn-redundant-casts, --disallow-incomplete-defs ]
# exclude: ^setup.py

View File

@ -26,8 +26,6 @@ AI_MODEL_OPTIONS: list[str] = [
"gpt-3.5-turbo",
"gpt-4",
"gpt-4-32k",
"bard",
"phind-gpt-4",
]
st.set_page_config(page_title=PAGE_TITLE, page_icon=PAGE_ICON)

View File

@ -1,107 +0,0 @@
"""
Reverse engineering of Google Bard from https://github.com/discordtehe/Bard
"""
import json
import random
import re
import string
import requests
import streamlit as st
US_AG = "Mozilla/5.0 (Windows NT 10.0; WOW64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/91.0.4472.114 Safari/537.36"
class BardChat:
"""
A class to interact with Google Bard.
Parameters
session_id: str
The __Secure-1PSID cookie.
"""
__slots__ = [
"headers",
"_reqid",
"SNlM0e",
"conversation_id",
"response_id",
"choice_id",
"session",
]
def __init__(self, session_id):
headers = {
"Host": "bard.google.com",
"X-Same-Domain": "1",
"User-Agent": US_AG,
"Content-Type": "application/x-www-form-urlencoded;charset=UTF-8",
"Origin": "https://bard.google.com",
"Referer": "https://bard.google.com/",
}
self._reqid = int("".join(random.choices(string.digits, k=4)))
self.conversation_id = ""
self.response_id = ""
self.choice_id = ""
self.session = requests.Session()
self.session.headers = headers
self.session.cookies.set("__Secure-1PSID", session_id)
self.SNlM0e = self.__get_snlm0e()
def __get_snlm0e(self):
resp = self.session.get(url="https://bard.google.com/", timeout=10)
# Find "SNlM0e":"<ID>"
if resp.status_code != 200:
st.error("Could not get Google Bard")
SNlM0e = re.search(r"SNlM0e\":\"(.*?)\"", resp.text).group(1)
return SNlM0e
def ask(self, message: str) -> dict:
"""
Send a message to Google Bard and return the response.
:param message: The message to send to Google Bard.
:return: A dict containing the response from Google Bard.
"""
# url params
params = {
"bl": "boq_assistant-bard-web-server_20230326.21_p0",
"_reqid": str(self._reqid),
"rt": "c",
}
# message arr -> data["f.req"]. Message is double json stringified
message_struct = [
[message],
None,
[self.conversation_id, self.response_id, self.choice_id],
]
data = {
"f.req": json.dumps([None, json.dumps(message_struct)]),
"at": self.SNlM0e,
}
# do the request!
resp = self.session.post(
"https://bard.google.com/_/BardChatUi/data/assistant.lamda.BardFrontendService/StreamGenerate",
params=params,
data=data,
timeout=120,
)
chat_data = json.loads(resp.content.splitlines()[3])[0][2]
if not chat_data:
return {"content": f"Google Bard encountered an error: {resp.content}."}
json_chat_data = json.loads(chat_data)
results = {
"content": json_chat_data[0][0],
"conversation_id": json_chat_data[1][0],
"response_id": json_chat_data[1][1],
"factualityQueries": json_chat_data[3],
"textQuery": json_chat_data[2][0] if json_chat_data[2] is not None else "",
"choices": [{"id": i[0], "content": i[1]} for i in json_chat_data[4]],
}
self.conversation_id = results["conversation_id"]
self.response_id = results["response_id"]
self.choice_id = results["choices"][0]["id"]
self._reqid += 100000
return results

View File

@ -1,285 +0,0 @@
from datetime import datetime
from queue import Empty, Queue
from threading import Thread
from time import time
from urllib.parse import quote
from curl_cffi.requests import post
cf_clearance = ""
user_agent = "Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/112.0.0.0 Safari/537.36" # noqa: E501
class PhindResponse:
class Completion:
class Choices:
def __init__(self, choice: dict) -> None:
self.text = choice["text"]
self.content = self.text.encode()
self.index = choice["index"]
self.logprobs = choice["logprobs"]
self.finish_reason = choice["finish_reason"]
def __repr__(self) -> str:
return f"""<__main__.APIResponse.Completion.Choices(\n text = {self.text.encode()},\n index = {self.index},\n logprobs = {self.logprobs},\n finish_reason = {self.finish_reason})object at 0x1337>""" # noqa: E501
def __init__(self, choices: dict) -> None:
self.choices = [self.Choices(choice) for choice in choices]
class Usage:
def __init__(self, usage_dict: dict) -> None:
self.prompt_tokens = usage_dict["prompt_tokens"]
self.completion_tokens = usage_dict["completion_tokens"]
self.total_tokens = usage_dict["total_tokens"]
def __repr__(self):
return f"""<__main__.APIResponse.Usage(\n prompt_tokens = {self.prompt_tokens},\n completion_tokens = {self.completion_tokens},\n total_tokens = {self.total_tokens})object at 0x1337>""" # noqa: E501
def __init__(self, response_dict: dict) -> None:
self.response_dict = response_dict
self.id = response_dict["id"]
self.object = response_dict["object"]
self.created = response_dict["created"]
self.model = response_dict["model"]
self.completion = self.Completion(response_dict["choices"])
self.usage = self.Usage(response_dict["usage"])
def json(self) -> dict:
return self.response_dict
class Search:
@staticmethod
def create(prompt: str, actual_search: bool = True, language: str = "en") -> dict: # None = no search
if user_agent == "":
raise ValueError("user_agent must be set, refer to documentation")
if not actual_search:
return {
"_type": "SearchResponse",
"queryContext": {
"originalQuery": prompt
},
"webPages": {
"webSearchUrl": f"https://www.bing.com/search?q={quote(prompt)}",
"totalEstimatedMatches": 0,
"value": []
},
"rankingResponse": {
"mainline": {
"items": []
}
}
}
headers = {
"authority": "www.phind.com",
"accept": "*/*",
"accept-language": "en,fr-FR;q=0.9,fr;q=0.8,es-ES;q=0.7,es;q=0.6,en-US;q=0.5,am;q=0.4,de;q=0.3",
"cookie": f"cf_clearance={cf_clearance}",
"origin": "https://www.phind.com",
"referer": "https://www.phind.com/search?q=hi&c=&source=searchbox&init=true",
"sec-ch-ua": '"Chromium";v="112", "Google Chrome";v="112", "Not:A-Brand";v="99"',
"sec-ch-ua-mobile": "?0",
"sec-ch-ua-platform": '"macOS"',
"sec-fetch-dest": "empty",
"sec-fetch-mode": "cors",
"sec-fetch-site": "same-origin",
"user-agent": user_agent
}
return post("https://www.phind.com/api/bing/search", headers=headers, json={
"q": prompt,
"userRankList": {},
"browserLanguage": language}).json()["rawBingResults"]
class Completion:
@staticmethod
def create(
model="gpt-4",
prompt: str = "",
results: dict = None,
creative: bool = False,
detailed: bool = False,
code_context: str = "",
language: str = "en") -> PhindResponse:
if user_agent == "":
raise ValueError("user_agent must be set, refer to documentation")
if results is None:
results = Search.create(prompt, actual_search=True)
if len(code_context) > 2999:
raise ValueError("codeContext must be less than 3000 characters")
models = {
"gpt-4": "expert",
"gpt-3.5-turbo": "intermediate",
"gpt-3.5": "intermediate",
}
json_data = {
"question": prompt,
"bingResults": results, # response.json()['rawBingResults'],
"codeContext": code_context,
"options": {
"skill": models[model],
"date": datetime.now().strftime("%d/%m/%Y"),
"language": language,
"detailed": detailed,
"creative": creative
}
}
headers = {
"authority": "www.phind.com",
"accept": "*/*",
"accept-language": "en,fr-FR;q=0.9,fr;q=0.8,es-ES;q=0.7,es;q=0.6,en-US;q=0.5,am;q=0.4,de;q=0.3",
"content-type": "application/json",
"cookie": f"cf_clearance={cf_clearance}",
"origin": "https://www.phind.com",
"referer": "https://www.phind.com/search?q=hi&c=&source=searchbox&init=true",
"sec-ch-ua": '"Chromium";v="112", "Google Chrome";v="112", "Not:A-Brand";v="99"',
"sec-ch-ua-mobile": "?0",
"sec-ch-ua-platform": '"macOS"',
"sec-fetch-dest": "empty",
"sec-fetch-mode": "cors",
"sec-fetch-site": "same-origin",
"user-agent": user_agent
}
completion = ""
response = post("https://www.phind.com/api/infer/answer", headers=headers, json=json_data, timeout=99999,
impersonate="chrome110")
for line in response.text.split("\r\n\r\n"):
completion += (line.replace("data: ", ""))
return PhindResponse({
"id": f"cmpl-1337-{int(time())}",
"object": "text_completion",
"created": int(time()),
"model": models[model],
"choices": [{
"text": completion,
"index": 0,
"logprobs": None,
"finish_reason": "stop"
}],
"usage": {
"prompt_tokens": len(prompt),
"completion_tokens": len(completion),
"total_tokens": len(prompt) + len(completion)
}
})
class StreamingCompletion:
message_queue = Queue()
stream_completed = False
@staticmethod
def request(model, prompt, results, creative, detailed, code_context, language) -> None:
models = {
"gpt-4": "expert",
"gpt-3.5-turbo": "intermediate",
"gpt-3.5": "intermediate",
}
json_data = {
"question": prompt,
"bingResults": results,
"codeContext": code_context,
"options": {
"skill": models[model],
"date": datetime.now().strftime("%d/%m/%Y"),
"language": language,
"detailed": detailed,
"creative": creative
}
}
headers = {
"authority": "www.phind.com",
"accept": "*/*",
"accept-language": "en,fr-FR;q=0.9,fr;q=0.8,es-ES;q=0.7,es;q=0.6,en-US;q=0.5,am;q=0.4,de;q=0.3",
"content-type": "application/json",
"cookie": f"cf_clearance={cf_clearance}",
"origin": "https://www.phind.com",
"referer": "https://www.phind.com/search?q=hi&c=&source=searchbox&init=true",
"sec-ch-ua": '"Chromium";v="112", "Google Chrome";v="112", "Not:A-Brand";v="99"',
"sec-ch-ua-mobile": "?0",
"sec-ch-ua-platform": '"macOS"',
"sec-fetch-dest": "empty",
"sec-fetch-mode": "cors",
"sec-fetch-site": "same-origin",
"user-agent": user_agent
}
post("https://www.phind.com/api/infer/answer",
headers=headers, json=json_data, timeout=99999, impersonate="chrome110",
content_callback=StreamingCompletion.handle_stream_response)
StreamingCompletion.stream_completed = True
@staticmethod
def create(
model: str = "gpt-4",
prompt: str = "",
results: dict = None,
creative: bool = False,
detailed: bool = False,
code_context: str = "",
language: str = "en"):
if user_agent == "":
raise ValueError("user_agent must be set, refer to documentation")
if results is None:
results = Search.create(prompt, actual_search=True)
if len(code_context) > 2999:
raise ValueError("codeContext must be less than 3000 characters")
Thread(target=StreamingCompletion.request, args=[
model, prompt, results, creative, detailed, code_context, language]).start()
while StreamingCompletion.stream_completed is not True or not StreamingCompletion.message_queue.empty():
try:
chunk = StreamingCompletion.message_queue.get(timeout=0)
if chunk == b"data: \r\ndata: \r\ndata: \r\n\r\n":
chunk = b"data: \n\n\r\n\r\n"
chunk = chunk.decode()
chunk = chunk.replace("data: \r\n\r\ndata: ", "data: \n")
chunk = chunk.replace("\r\ndata: \r\ndata: \r\n\r\n", "\n\n\r\n\r\n")
chunk = chunk.replace("data: ", "").replace("\r\n\r\n", "")
yield PhindResponse({
"id": f"cmpl-1337-{int(time())}",
"object": "text_completion",
"created": int(time()),
"model": model,
"choices": [{
"text": chunk,
"index": 0,
"logprobs": None,
"finish_reason": "stop"
}],
"usage": {
"prompt_tokens": len(prompt),
"completion_tokens": len(chunk),
"total_tokens": len(prompt) + len(chunk)
}
})
except Empty:
pass
@staticmethod
def handle_stream_response(response):
StreamingCompletion.message_queue.put(response)

View File

@ -2,11 +2,8 @@ from random import randrange
import streamlit as st
from openai.error import InvalidRequestError, OpenAIError
from requests.exceptions import TooManyRedirects
from streamlit_chat import message
from .agi import phind
from .agi.bard import BardChat
from .agi.chat_gpt import create_gpt_completion
from .stt import show_voice_input
from .tts import show_audio_player
@ -100,33 +97,6 @@ def show_gpt_conversation() -> None:
st.error(err)
def show_bard_conversation() -> None:
try:
bard = BardChat(st.secrets.api_credentials.bard_session)
ai_content = bard.ask(st.session_state.user_text)
st.warning(ai_content.get("content"))
except (TooManyRedirects, AttributeError) as err:
st.error(err)
def phind_get_answer(question: str):
phind.cf_clearance = st.secrets.api_credentials.phind_cf_clearance
phind.user_agent = "Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/112.0.0.0 Safari/537.36" # noqa: E501
try:
result = phind.Completion.create(
model="gpt-4",
prompt=question,
results=phind.Search.create(question, actual_search=True),
creative=False,
detailed=False,
code_context=""
)
st.markdown(result.completion.choices[0].text)
except Exception as e:
st.error(e)
def show_conversation() -> None:
if st.session_state.messages:
st.session_state.messages.append({"role": "user", "content": st.session_state.user_text})
@ -136,9 +106,4 @@ def show_conversation() -> None:
{"role": "system", "content": ai_role},
{"role": "user", "content": st.session_state.user_text},
]
if st.session_state.model == "bard":
show_bard_conversation()
elif st.session_state.model == "phind-gpt-4":
phind_get_answer(st.session_state.user_text)
else:
show_gpt_conversation()

View File

@ -1,10 +1,9 @@
pip>=23.1
streamlit>=1.21.0
pip>=23.1.2
streamlit>=1.22.0
streamlit-chat>=0.0.2.2
streamlit_option_menu>=0.3.2
openai>=0.27.4
openai>=0.27.5
gtts>=2.3.1
bokeh==2.4.2
streamlit-bokeh-events>=0.1.2
watchdog>=3.0.0
curl_cffi>=0.5.5