File size: 18,728 Bytes
acc4ffe
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
# example call script
# https://dev.azure.com/visionbio/objectdetection/_git/objectdetection?path=/verify/langimg.py&version=GBehazar/langchain&_a=contents

import re
import io
import os
import ssl
from typing import Optional, Tuple
import datetime
import sys
import gradio as gr
import requests
import json
from threading import Lock
from langchain import ConversationChain, LLMChain
from langchain.agents import load_tools, initialize_agent, Tool
from langchain.tools.bing_search.tool import BingSearchRun, BingSearchAPIWrapper
from langchain.chains.conversation.memory import ConversationBufferMemory
from langchain.llms import OpenAI
from langchain.chains import PALChain
from langchain.llms import AzureOpenAI
from langchain.utilities import ImunAPIWrapper, ImunMultiAPIWrapper
from openai.error import AuthenticationError, InvalidRequestError, RateLimitError
import argparse

# header_key = os.environ["CVFIAHMED_KEY"]
OPENAI_API_KEY = os.environ["OPENAI_API_KEY"]
TOOLS_LIST = ['pal-math', 'imun']  #'google-search','news-api','tmdb-api','open-meteo-api'
TOOLS_DEFAULT_LIST = ['pal-math', 'imun']
BUG_FOUND_MSG = "Congratulations, you've found a bug in this application!"
AUTH_ERR_MSG = "Please paste your OpenAI key from openai.com to use this application. "
MAX_TOKENS = 512


############ GLOBAL CHAIN ###########
# chain = None
# memory = None
#####################################
############ GLOBAL IMAGE_COUNT #####
IMAGE_COUNT=0
#####################################
############## ARGS #################
AGRS = None
#####################################


# Temporarily address Wolfram Alpha SSL certificate issue
ssl._create_default_https_context = ssl._create_unverified_context


def get_caption_onnx_api(imgf):

    headers = {
        'Content-Type': 'application/octet-stream',
        'Ocp-Apim-Subscription-Key': header_key,
    }

    params = {
        'features': 'description',
        'model-version': 'latest',
        'language': 'en',
        'descriptionExclude': 'Celebrities,Landmarks',
    }

    with open(imgf, 'rb') as f:
        data = f.read()

    response = requests.post('https://cvfiahmed.cognitiveservices.azure.com/vision/v2022-07-31-preview/operations/imageanalysis:analyze', params=params, headers=headers, data=data)

    return json.loads(response.content)['descriptionResult']['values'][0]['text']

def reset_memory(history):
    # global memory
    # memory.clear()
    print ("clearning memory, loading langchain...")
    load_chain()
    history = []
    return history, history


def load_chain(history):
    global ARGS
    # global chain
    # global memory
    # memory = None

    if ARGS.openAIModel == 'openAIGPT35':
        # openAI GPT 3.5
        llm = OpenAI(temperature=0, max_tokens=MAX_TOKENS)
    elif ARGS.openAIModel == 'azureChatGPT':
    # for Azure OpenAI ChatGPT
    # Azure OpenAI param name 'deployment_name': 'text-davinci-002', 'model_name': 'text-davinci-002', 'temperature': 0.7, 'max_tokens': 256, 'top_p': 1, 'frequency_penalty': 0, 'presence_penalty': 0, 'n': 1, 'best_of': 1
        # llm = AzureOpenAI(deployment_name="text-chat-davinci-002", model_name="text-chat-davinci-002", temperature=1, top_p=0.9, max_tokens=MAX_TOKENS)
        llm = AzureOpenAI(deployment_name="text-chat-davinci-002", model_name="text-chat-davinci-002", temperature=0, max_tokens=MAX_TOKENS)
        # llm = AzureOpenAI(deployment_name="gpt-35-turbo-version-0301", model_name="gpt-35-turbo (version 0301)", temperature=0, max_tokens=MAX_TOKENS)
    elif ARGS.openAIModel == 'azureTextDavinci003':
    # for Azure OpenAI ChatGPT
    # Azure OpenAI param name 'deployment_name': 'text-davinci-002', 'model_name': 'text-davinci-002', 'temperature': 0.7, 'max_tokens': 256, 'top_p': 1, 'frequency_penalty': 0, 'presence_penalty': 0, 'n': 1, 'best_of': 1
        llm = AzureOpenAI(deployment_name="text-davinci-003", model_name="text-davinci-003", temperature=0, max_tokens=MAX_TOKENS)

    # tool_names = TOOLS_DEFAULT_LIST
    # tools = load_tools(tool_names, llm=llm)
    memory = ConversationBufferMemory(memory_key="chat_history")

    #############################
    # loading tools

    imun_dense = ImunAPIWrapper(
        imun_url="https://ehazarwestus.cognitiveservices.azure.com/computervision/imageanalysis:analyze",
        params="api-version=2023-02-01-preview&model-version=latest&features=denseCaptions",
        imun_subscription_key=os.environ["IMUN_SUBSCRIPTION_KEY2"])

    imun = ImunAPIWrapper()
    imun = ImunMultiAPIWrapper(imuns=[imun, imun_dense])

    imun_celeb = ImunAPIWrapper(
        imun_url="https://cvfiahmed.cognitiveservices.azure.com/vision/v3.2/models/celebrities/analyze",
        params="")

    imun_read = ImunAPIWrapper(
        imun_url="https://vigehazar.cognitiveservices.azure.com/formrecognizer/documentModels/prebuilt-read:analyze",
        params="api-version=2022-08-31",
        imun_subscription_key=os.environ["IMUN_OCR_SUBSCRIPTION_KEY"])

    imun_receipt = ImunAPIWrapper(
        imun_url="https://vigehazar.cognitiveservices.azure.com/formrecognizer/documentModels/prebuilt-receipt:analyze",
        params="api-version=2022-08-31",
        imun_subscription_key=os.environ["IMUN_OCR_SUBSCRIPTION_KEY"])

    imun_businesscard = ImunAPIWrapper(
        imun_url="https://vigehazar.cognitiveservices.azure.com/formrecognizer/documentModels/prebuilt-businessCard:analyze",
        params="api-version=2022-08-31",
        imun_subscription_key=os.environ["IMUN_OCR_SUBSCRIPTION_KEY"])

    imun_layout = ImunAPIWrapper(
        imun_url="https://vigehazar.cognitiveservices.azure.com/formrecognizer/documentModels/prebuilt-layout:analyze",
        params="api-version=2022-08-31",
        imun_subscription_key=os.environ["IMUN_OCR_SUBSCRIPTION_KEY"])

    bing = BingSearchAPIWrapper(k=2)

    def edit_photo(query: str) -> str:
        endpoint = "http://10.123.124.92:7863/"
        query = query.strip()
        url_idx = query.rfind(" ")
        img_url = query[url_idx + 1:].strip()
        if img_url.endswith((".", "?")):
            img_url = img_url[:-1]
        if not img_url.startswith(("http://", "https://")):
            return "Invalid image URL"
        img_url = img_url.replace("0.0.0.0", "10.123.124.92")
        instruction = query[:url_idx]
        # This should be some internal IP to wherever the server runs
        job = {"image_path": img_url, "instruction": instruction}
        response = requests.post(endpoint, json=job)
        if response.status_code != 200:
            return "Could not finish the task try again later!"
        return "Here is the edited image " + endpoint + response.json()["edited_image"]

    # these tools should not step on each other's toes
    tools = [
        Tool(
            name="PAL-MATH",
            func=PALChain.from_math_prompt(llm).run,
            description=(
            "A wrapper around calculator. "
            "A language model that is really good at solving complex word math problems."
            "Input should be a fully worded hard word math problem."
            )
        ),
        Tool(
            name = "Image Understanding",
            func=imun.run,
            description=(
            "A wrapper around Image Understanding. "
            "Useful for when you need to understand what is inside an image (objects, texts, people)."
            "Input should be an image url, or path to an image file (e.g. .jpg, .png)."
            )
        ),
        Tool(
            name = "OCR Understanding",
            func=imun_read.run,
            description=(
            "A wrapper around OCR Understanding (Optical Character Recognition). "
            "Useful after Image Understanding tool has found text or handwriting is present in the image tags."
            "This tool can find the actual text, written name, or product name in the image."
            "Input should be an image url, or path to an image file (e.g. .jpg, .png)."
            )
        ),
        Tool(
            name = "Receipt Understanding",
            func=imun_receipt.run,
            description=(
            "A wrapper receipt understanding. "
            "Useful after Image Understanding tool has recognized a receipt in the image tags."
            "This tool can find the actual receipt text, prices and detailed items."
            "Input should be an image url, or path to an image file (e.g. .jpg, .png)."
            )
        ),
        Tool(
            name = "Business Card Understanding",
            func=imun_businesscard.run,
            description=(
            "A wrapper around business card understanding. "
            "Useful after Image Understanding tool has recognized businesscard in the image tags."
            "This tool can find the actual business card text, name, address, email, website on the card."
            "Input should be an image url, or path to an image file (e.g. .jpg, .png)."
            )
        ),
        Tool(
            name = "Layout Understanding",
            func=imun_layout.run,
            description=(
            "A wrapper around layout and table understanding. "
            "Useful after Image Understanding tool has recognized businesscard in the image tags."
            "This tool can find the actual business card text, name, address, email, website on the card."
            "Input should be an image url, or path to an image file (e.g. .jpg, .png)."
            )
        ),
        Tool(
            name = "Celebrity Understanding",
            func=imun_celeb.run,
            description=(
            "A wrapper around celebrity understanding. "
            "Useful after Image Understanding tool has recognized people in the image tags that could be celebrities."
            "This tool can find the name of celebrities in the image."
            "Input should be an image url, or path to an image file (e.g. .jpg, .png)."
            )
        ),
        BingSearchRun(api_wrapper=bing),
        Tool(
            name = "Photo Editing",
            func=edit_photo,
            description=(
            "A wrapper around photo editing. "
            "Useful to edit an image with a given instruction."
            "Input should be an image url, or path to an image file (e.g. .jpg, .png)."
            )
        ),
    ]

    # chain = initialize_agent(tools, llm, agent="conversational-react-description", verbose=True, memory=memory)
    # chain = initialize_agent(tools, llm, agent="conversational-assistant", verbose=True, memory=memory, return_intermediate_steps=True)
    chain = initialize_agent(tools, llm, agent="conversational-assistant", verbose=True, memory=memory, return_intermediate_steps=True, max_iterations=4)
    print("langchain reloaded")
    history = []
    history.append(("Show me what you got!", "Hi Human, I am ready to serve!"))    
    return history, history, chain


def run_chain(chain, inp):
    # global chain
    
    output = ""
    try:
        output = chain.conversation(input=inp, keep_short=ARGS.noIntermediateConv)
        # output = chain.run(input=inp)
    except AuthenticationError as ae:
        output = AUTH_ERR_MSG + str(datetime.datetime.now()) + ". " + str(ae)
        print("output", output)
    except RateLimitError as rle:
        output = "\n\nRateLimitError: " + str(rle)
    except ValueError as ve:
        output = "\n\nValueError: " + str(ve)
    except InvalidRequestError as ire:
        output = "\n\nInvalidRequestError: " + str(ire)
    except Exception as e:
        output = "\n\n" + BUG_FOUND_MSG + ":\n\n" + str(e)

    return output


class ChatWrapper:

    def __init__(self):
        self.lock = Lock()

    def __call__(
            self, inp: str, history: Optional[Tuple[str, str]], chain: Optional[ConversationChain]
    ):
        """Execute the chat functionality."""
        self.lock.acquire()
        try:
            print("\n==== date/time: " + str(datetime.datetime.now()) + " ====")
            print("inp: " + inp)
            history = history or []
            # If chain is None, that is because no API key was provided.
            output = "Please paste your OpenAI key from openai.com to use this app. " + str(datetime.datetime.now())
            
            ########################
            # multi line 
            outputs = run_chain(chain, inp)

            outputs = process_chain_output(outputs)

            print (" len(outputs) {}".format(len(outputs)))
            for i, output in enumerate(outputs):
                if i==0:
                    history.append((inp, output))    
                else:
                    history.append((None, output))
            

        except Exception as e:
            raise e
        finally:
            self.lock.release()

        print (history)
        return history, history, ""

# upload image 
def add_image(state, chain, image):
    global IMAGE_COUNT
    global ARGS
    IMAGE_COUNT = IMAGE_COUNT + 1
    state = state or []

    # cap_onnx = get_caption_onnx_api(image.name)
    # cap_onnx = "The image shows " + cap_onnx
    # state = state + [(f"![](/file={image.name})", cap_onnx)]

    # : f"Image {N} http://0.0.0.0:7860/file={image.name}"
    # Image_N
    # wget http://0.0.0.0:7860/file=/tmp/bananabdzk2eqi.jpg
    # url_input_for_chain = "Image_{} http://0.0.0.0:7860/file={}".format(IMAGE_COUNT, image.name)
    url_input_for_chain = "http://0.0.0.0:{}/file={}".format(ARGS.port, image.name)

    # !!!!!! quick HACK to refer to image in this server for image editing pruprose
    url_input_for_chain = url_input_for_chain.replace("0.0.0.0", "10.123.124.92")


    ########################
    # multi line 
    outputs = run_chain(chain, url_input_for_chain)

    outputs = process_chain_output(outputs)

    print (" len(outputs) {}".format(len(outputs)))
    for i, output in enumerate(outputs):
        if i==0:
            # state.append((f"![](/file={image.name})", output))    
            state.append(((image.name,), output))
        else:
            state.append((None, output))
    


    print (state)
    return state, state

def replace_with_image_markup(text):
    img_url = None
    text= text.strip()
    url_idx = text.rfind(" ")
    img_url = text[url_idx + 1:].strip()
    if img_url.endswith((".", "?")):
        img_url = img_url[:-1]

    # if img_url is not None:
    #     img_url = f"![](/file={img_url})"
    return img_url

def process_chain_output(outputs):
    global ARGS
    # print("outputs {}".format(outputs))
    if isinstance(outputs, str): # single line output
        outputs = [outputs]
    elif isinstance(outputs, list): # multi line output
        if ARGS.noIntermediateConv: # remove the items with assistant in it. 
            cleanOutputs = []
            for output in outputs:
                # print("inside loop outputs {}".format(output))
                # found an edited image url to embed
                img_url = None
                # print ("type list: {}".format(output))
                if "assistant: here is the edited image " in output.lower():
                    img_url = replace_with_image_markup(output)
                    cleanOutputs.append("Assistant: Here is the edited image")
                if img_url is not None:
                    cleanOutputs.append((img_url,))
                cleanOutputs.append(output)
                # cleanOutputs = cleanOutputs + output+ "."
            outputs = cleanOutputs
    
    # make it bold
    # outputs = "<b>{}</b>".format(outputs)
    return outputs


def init_and_kick_off():
    global ARGS
    # initalize chatWrapper
    chat = ChatWrapper()

    # with gr.Blocks(css=".gradio-container {background-color: lightgray}") as block:
    # with gr.Blocks(css="#resetbtn {background-color: #4CAF50; color: red;}  #chatbot {height: 700px; overflow: auto;}") as block:
    with gr.Blocks() as block:  
        llm_state = gr.State()
        history_state = gr.State()
        chain_state = gr.State()

        
        
        reset_btn = gr.Button(value="!!!CLICK to wake up the AI!!!", variant="secondary", elem_id="resetbtn").style(full_width=False)

        with gr.Row():
            chatbot = gr.Chatbot(elem_id="chatbot").style(height=620)

        with gr.Row():
            with gr.Column(scale=0.75):
                message = gr.Textbox(label="What's on your mind??",
                                        placeholder="What's the answer to life, the universe, and everything?",
                                        lines=1)
            with gr.Column(scale=0.15):
                submit = gr.Button(value="Send", variant="secondary").style(full_width=False)
            with gr.Column(scale=0.10, min_width=0):
                btn = gr.UploadButton("📁", file_types=["image", "video", "audio"])
        
        with gr.Row():
            with gr.Column(scale=0.90):
                gr.HTML("""
                    <p>This application, developed by Cognitive Service Team Microsoft, demonstrates all cognitive service APIs in a conversational agent
                    </p>""")
            # with gr.Column(scale=0.10):
            #     reset_btn = gr.Button(value="Initiate Chat", variant="secondary", elem_id="resetbtn").style(full_width=False)

        message.submit(chat, inputs=[message, history_state, chain_state],
                    outputs=[chatbot, history_state, message])

        submit.click(chat, inputs=[message, history_state, chain_state],
                    outputs=[chatbot, history_state, message])

        btn.upload(add_image, inputs=[history_state, chain_state, btn], outputs=[history_state, chatbot])
        # reset_btn.click(reset_memory, inputs=[history_state], outputs=[chatbot, history_state])

        # openai_api_key_textbox.change(set_openai_api_key,
        #                           inputs=[openai_api_key_textbox],
        #                           outputs=[chain_state])
        # load the chain    
        reset_btn.click(load_chain, inputs=[history_state], outputs=[chatbot, history_state, chain_state])
        


    # # load the chain    
    # load_chain()

    # launch the app
    block.launch(server_name="0.0.0.0", server_port = ARGS.port)

if __name__ == '__main__':
    parser = argparse.ArgumentParser()

    parser.add_argument('--port', type=int, required=False, default=7860)
    parser.add_argument('--openAIModel', type=str, required=False, default='openAIGPT35')
    parser.add_argument('--noIntermediateConv', default=False, action='store_true', help='if this flag is turned on no intermediate conversation should be shown')

    global ARGS
    ARGS = parser.parse_args()

    init_and_kick_off()
    
# python app_test.py --port 7862 --openAIModel 'azureChatGPT'