Spaces:
Sleeping
Sleeping
File size: 16,292 Bytes
9c8a515 |
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 |
import gradio as gr
#import urllib.request
import requests
import bs4
import lxml
import os
#import subprocess
from huggingface_hub import InferenceClient,HfApi
import random
import json
import datetime
from pypdf import PdfReader
import uuid
#from query import tasks
from agent import (
PREFIX,
SAVE_MEMORY,
COMPRESS_DATA_PROMPT,
COMPRESS_DATA_PROMPT_SMALL,
LOG_PROMPT,
LOG_RESPONSE,
)
client = InferenceClient(
"mistralai/Mixtral-8x7B-Instruct-v0.1"
)
reponame="Omnibus/tmp"
save_data=f'https://huggingface.co/datasets/{reponame}/raw/main/'
token_self = os.environ['HF_TOKEN']
api=HfApi(token=token_self)
def find_all(purpose,task,history, url, result):
return_list=[]
print (url)
print (f"trying URL:: {url}")
try:
if url != "" and url != None:
out = []
source = requests.get(url)
if source.status_code ==200:
soup = bs4.BeautifulSoup(source.content,'lxml')
rawp=(f'RAW TEXT RETURNED: {soup.text}')
cnt=0
cnt+=len(rawp)
out.append(rawp)
out.append("HTML fragments: ")
q=("a","p","span","content","article")
for p in soup.find_all("a"):
out.append([{"LINK TITLE":p.get('title'),"URL":p.get('href'),"STRING":p.string}])
c=0
out = str(out)
rl = len(out)
print(f'rl:: {rl}')
for i in str(out):
if i == " " or i=="," or i=="\n" or i=="/" or i=="." or i=="<":
c +=1
print (f'c:: {c}')
#if c > MAX_HISTORY:
#print("compressing...")
#rawp = compress_data(c,purpose,task,out,result)
#result += rawp
rawp=out
return True, rawp
else:
return False, "Enter Valid URL"
except Exception as e:
print (e)
return False, f'Error: {e}'
#else:
# history = "observation: The search query I used did not return a valid response"
return "MAIN", None, history, task
def read_txt(txt_path):
text=""
with open(txt_path,"r") as f:
text = f.read()
f.close()
print (text)
return text
def read_pdf(pdf_path):
text=""
reader = PdfReader(f'{pdf_path}')
number_of_pages = len(reader.pages)
for i in range(number_of_pages):
page = reader.pages[i]
text = f'{text}\n{page.extract_text()}'
print (text)
return text
error_box=[]
def read_pdf_online(url):
uid=uuid.uuid4()
print(f"reading {url}")
response = requests.get(url, stream=True)
print(response.status_code)
text=""
#################
#####################
try:
if response.status_code == 200:
with open("test.pdf", "wb") as f:
f.write(response.content)
#f.close()
#out = Path("./data.pdf")
#print (out)
reader = PdfReader("test.pdf")
number_of_pages = len(reader.pages)
print(number_of_pages)
for i in range(number_of_pages):
page = reader.pages[i]
text = f'{text}\n{page.extract_text()}'
print(f"PDF_TEXT:: {text}")
return text
else:
text = response.status_code
error_box.append(url)
print(text)
return text
except Exception as e:
print (e)
return e
VERBOSE = True
MAX_HISTORY = 100
MAX_DATA = 20000
def format_prompt(message, history):
prompt = "<s>"
for user_prompt, bot_response in history:
prompt += f"[INST] {user_prompt} [/INST]"
prompt += f" {bot_response}</s> "
prompt += f"[INST] {message} [/INST]"
return prompt
def run_gpt(
prompt_template,
stop_tokens,
max_tokens,
seed,
**prompt_kwargs,
):
print(seed)
timestamp=datetime.datetime.now()
generate_kwargs = dict(
temperature=0.9,
max_new_tokens=max_tokens,
top_p=0.95,
repetition_penalty=1.0,
do_sample=True,
seed=seed,
)
content = PREFIX.format(
timestamp=timestamp,
purpose="Compile the provided data and complete the users task"
) + prompt_template.format(**prompt_kwargs)
if VERBOSE:
print(LOG_PROMPT.format(content))
#formatted_prompt = format_prompt(f"{system_prompt}, {prompt}", history)
#formatted_prompt = format_prompt(f'{content}', history)
stream = client.text_generation(content, **generate_kwargs, stream=True, details=True, return_full_text=False)
resp = ""
for response in stream:
resp += response.token.text
#yield resp
if VERBOSE:
print(LOG_RESPONSE.format(resp))
return resp
def compress_data(c, instruct, history):
seed=random.randint(1,1000000000)
print (c)
#tot=len(purpose)
#print(tot)
divr=int(c)/MAX_DATA
divi=int(divr)+1 if divr != int(divr) else int(divr)
chunk = int(int(c)/divr)
print(f'chunk:: {chunk}')
print(f'divr:: {divr}')
print (f'divi:: {divi}')
out = []
#out=""
s=0
e=chunk
print(f'e:: {e}')
new_history=""
#task = f'Compile this data to fulfill the task: {task}, and complete the purpose: {purpose}\n'
for z in range(divi):
print(f's:e :: {s}:{e}')
hist = history[s:e]
resp = run_gpt(
COMPRESS_DATA_PROMPT_SMALL,
stop_tokens=["observation:", "task:", "action:", "thought:"],
max_tokens=8192,
seed=seed,
direction=instruct,
knowledge="",
history=hist,
)
out.append(resp)
#new_history = resp
print (resp)
#out+=resp
e=e+chunk
s=s+chunk
return out
def compress_data_og(c, instruct, history):
seed=random.randint(1,1000000000)
print (c)
#tot=len(purpose)
#print(tot)
divr=int(c)/MAX_DATA
divi=int(divr)+1 if divr != int(divr) else int(divr)
chunk = int(int(c)/divr)
print(f'chunk:: {chunk}')
print(f'divr:: {divr}')
print (f'divi:: {divi}')
out = []
#out=""
s=0
e=chunk
print(f'e:: {e}')
new_history=""
#task = f'Compile this data to fulfill the task: {task}, and complete the purpose: {purpose}\n'
for z in range(divi):
print(f's:e :: {s}:{e}')
hist = history[s:e]
resp = run_gpt(
COMPRESS_DATA_PROMPT,
stop_tokens=["observation:", "task:", "action:", "thought:"],
max_tokens=8192,
seed=seed,
direction=instruct,
knowledge=new_history,
history=hist,
)
new_history = resp
print (resp)
out+=resp
e=e+chunk
s=s+chunk
'''
resp = run_gpt(
COMPRESS_DATA_PROMPT,
stop_tokens=["observation:", "task:", "action:", "thought:"],
max_tokens=8192,
seed=seed,
direction=instruct,
knowledge=new_history,
history="All data has been recieved.",
)'''
print ("final" + resp)
#history = "observation: {}\n".format(resp)
return resp
def summarize(inp,history,report_check,sum_mem_check,data=None,files=None,url=None,pdf_url=None,pdf_batch=None):
json_box=[]
if inp == "":
inp = "Process this data"
history.clear()
history = [(inp,"Working on it...")]
yield "",history,error_box,json_box
if pdf_batch.startswith("http"):
c=0
data=""
for i in str(pdf_batch):
if i==",":
c+=1
print (f'c:: {c}')
try:
for i in range(c+1):
batch_url = pdf_batch.split(",",c)[i]
bb = read_pdf_online(batch_url)
data=f'{data}\nFile Name URL ({batch_url}):\n{bb}'
except Exception as e:
print(e)
#data=f'{data}\nError reading URL ({batch_url})'
if pdf_url.startswith("http"):
print("PDF_URL")
out = read_pdf_online(pdf_url)
data=out
if url.startswith("http"):
val, out = find_all(url)
if not val:
data="Error"
rawp = str(out)
else:
data=out
if files:
for i, file in enumerate(files):
try:
print (file)
if file.endswith(".pdf"):
zz=read_pdf(file)
print (zz)
data=f'{data}\nFile Name ({file}):\n{zz}'
elif file.endswith(".txt"):
zz=read_txt(file)
print (zz)
data=f'{data}\nFile Name ({file}):\n{zz}'
except Exception as e:
data=f'{data}\nError opening File Name ({file})'
print (e)
if data != "Error" and data != "":
print(inp)
out = str(data)
rl = len(out)
print(f'rl:: {rl}')
c=1
for i in str(out):
if i == " " or i=="," or i=="\n":
c +=1
print (f'c:: {c}')
if sum_mem_check=="Memory":
save_memory(inp,out)
rawp = "Complete"
if sum_mem_check=="Summarize":
json_out = compress_data(c,inp,out)
out = str(json_out)
if report_check:
rl = len(out)
print(f'rl:: {rl}')
c=1
for i in str(out):
if i == " " or i=="," or i=="\n":
c +=1
print (f'c2:: {c}')
rawp = compress_data_og(c,inp,out)
else:
rawp = out
else:
rawp = "Provide a valid data source"
history.clear()
history.append((inp,rawp))
yield "", history,error_box,json_out
SAVE_MEMORY = """
You are attempting to complete the task
task: {task}
Data:
{history}
Instructions:
Compile and categorize the data above into a JSON dictionary string
Include ALL text, datapoints, titles, descriptions, and source urls indexed into an easy to search JSON format
Your final response should be only the final formatted JSON string enclosed in brackets, and nothing else.
Required keys:
"keywords":["short", "list", "of", "keywords", "relevant", "to", "this", "entry"]
"title":"title of entry"
"description":"description of entry"
"content":"full content of data about entry"
"url":"https://url.source"
"""
def save_memory(purpose, history):
uid=uuid.uuid4()
history=str(history)
c=0
inp = str(history)
rl = len(inp)
print(f'rl:: {rl}')
for i in str(inp):
if i == " " or i=="," or i=="\n" or i=="/" or i=="." or i=="<":
c +=1
print (f'c:: {c}')
seed=random.randint(1,1000000000)
print (c)
#tot=len(purpose)
#print(tot)
divr=int(c)/MAX_DATA
divi=int(divr)+1 if divr != int(divr) else int(divr)
chunk = int(int(c)/divr)
print(f'chunk:: {chunk}')
print(f'divr:: {divr}')
print (f'divi:: {divi}')
#out = []
#out=""
s=0
e=chunk
print(f'e:: {e}')
new_history=""
task = f'Index this Data\n'
for z in range(divi):
print(f's:e :: {s}:{e}')
hist = inp[s:e]
resp = run_gpt(
SAVE_MEMORY,
stop_tokens=["observation:", "task:", "action:", "thought:"],
max_tokens=4096,
seed=seed,
purpose=purpose,
task=task,
history=hist,
).strip('\n')
#new_history = resp
print (resp)
#out+=resp
e=e+chunk
s=s+chunk
print ("final1" + resp)
try:
resp='[{'+resp.split('[{')[1].split('</s>')[0]
print ("final2\n" + resp)
print(f"keywords:: {resp['keywords']}")
except Exception as e:
resp = resp
print(e)
timestamp=str(datetime.datetime.now())
timename=timestamp.replace(" ","--").replace(":","-").replace(".","-")
json_object=resp
#json_object = json.dumps(out_box)
#json_object = json.dumps(out_box,indent=4)
with open(f"tmp-{uid}.json", "w") as outfile:
outfile.write(json_object)
api.upload_file(
path_or_fileobj=f"tmp-{uid}.json",
path_in_repo=f"/mem-test2/{timename}.json",
repo_id=reponame,
#repo_id=save_data.split('datasets/',1)[1].split('/raw',1)[0],
token=token_self,
repo_type="dataset",
)
lines = resp.strip().strip("\n").split("\n")
r = requests.get(f'{save_data}mem-test2/main.json')
print(f'status code main:: {r.status_code}')
if r.status_code==200:
lod = json.loads(r.text)
#lod = eval(lod)
print (f'lod:: {lod}')
else:
lod = []
for i,line in enumerate(lines):
key_box=[]
print(f'LINE:: {line}')
if ":" in line:
print(f'line:: {line}')
if "keywords" in line[:16]:
print(f'trying:: {line}')
keyw=line.split(":")[1]
print (keyw)
print (keyw.split("[")[1].split("]")[0])
keyw=keyw.split("[")[1].split("]")[0]
for ea in keyw.split(","):
s1=""
ea=ea.strip().strip("\n")
for ev in ea:
if ev.isalnum():
s1+=ev
if ev == " ":
s1+=ev
#ea=s1
print(s1)
key_box.append(s1)
lod.append({"file_name":timename,"keywords":key_box})
json_object = json.dumps(lod, indent=4)
with open(f"tmp2-{uid}.json", "w") as outfile2:
outfile2.write(json_object)
api.upload_file(
path_or_fileobj=f"tmp2-{uid}.json",
path_in_repo=f"/mem-test2/main.json",
repo_id=reponame,
#repo_id=save_data.split('datasets/',1)[1].split('/raw',1)[0],
token=token_self,
repo_type="dataset",
)
#return [resp]
#################################
def clear_fn():
return "",[(None,None)]
with gr.Blocks() as app:
gr.HTML("""<center><h1>Mixtral 8x7B TLDR Summarizer + Web</h1><h3>Summarize Data of unlimited length</h3>""")
chatbot = gr.Chatbot(label="Mixtral 8x7B Chatbot",show_copy_button=True)
with gr.Row():
with gr.Column(scale=3):
prompt=gr.Textbox(label = "Instructions (optional)")
with gr.Column(scale=1):
report_check=gr.Checkbox(label="Return Report", value=True)
sum_mem_check=gr.Radio(label="Output",choices=["Summary","Memory"])
button=gr.Button()
#models_dd=gr.Dropdown(choices=[m for m in return_list],interactive=True)
with gr.Row():
stop_button=gr.Button("Stop")
clear_btn = gr.Button("Clear")
with gr.Row():
with gr.Tab("Text"):
data=gr.Textbox(label="Input Data (paste text)", lines=6)
with gr.Tab("File"):
file=gr.Files(label="Input File(s) (.pdf .txt)")
with gr.Tab("Raw HTML"):
url = gr.Textbox(label="URL")
with gr.Tab("PDF URL"):
pdf_url = gr.Textbox(label="PDF URL")
with gr.Tab("PDF Batch"):
pdf_batch = gr.Textbox(label="PDF URL Batch (comma separated)")
json_out=gr.JSON()
e_box=gr.Textbox()
#text=gr.JSON()
#inp_query.change(search_models,inp_query,models_dd)
clear_btn.click(clear_fn,None,[prompt,chatbot])
go=button.click(summarize,[prompt,chatbot,report_check,sum_mem_check,data,file,url,pdf_url,pdf_batch],[prompt,chatbot,e_box,json_out])
stop_button.click(None,None,None,cancels=[go])
app.queue(default_concurrency_limit=20).launch(show_api=False) |