Spaces:
Running
Running
File size: 28,600 Bytes
899e654 3655a2f f6874c9 bcfa1bb 3a9a095 f6874c9 bf38597 1f46917 bf38597 1f46917 b9c4863 bf38597 b9c4863 f6874c9 4d70c71 1f46917 bf38597 f6874c9 34275fb f6874c9 81ff87e 27ed3be bf38597 0f106cc a3d9c23 0f106cc ed0e073 4d70c71 ed0e073 4d70c71 ed0e073 f41b9ef ed0e073 bf38597 f6874c9 5799683 f6874c9 5799683 f6874c9 3655a2f f6874c9 5799683 f6874c9 27ed3be f6874c9 27ed3be 3655a2f f6874c9 3655a2f 5799683 4d70c71 5799683 4d70c71 5799683 4d70c71 5799683 5135341 27ed3be 5135341 27ed3be 5799683 3655a2f 5135341 3655a2f f6874c9 5135341 f6874c9 5852d5d f6874c9 5852d5d 3655a2f bf38597 bffdc7a 4b342a9 bffdc7a 48303e3 d89748a f6874c9 d89748a f6874c9 d89748a bcfa1bb ed0e073 b9c4863 bcfa1bb bf38597 a28bf8c bf38597 a28bf8c bf38597 a28bf8c bf38597 2a276f1 a28bf8c bf38597 3655a2f a28bf8c 3655a2f bf38597 a28bf8c bf38597 a28bf8c 3a9a095 81ff87e 3a9a095 bf38597 009769a bf38597 3655a2f bf38597 bcfa1bb 3655a2f bcfa1bb b9c4863 bcfa1bb b9c4863 f6874c9 4d70c71 ed0e073 9451027 bcfa1bb 1e0c08a 4d70c71 ed0e073 5799683 bcfa1bb 3655a2f bcfa1bb 4d70c71 81ff87e 27ed3be 3655a2f bbb69f6 4d70c71 ed0e073 4d70c71 06d209a 4d70c71 81ff87e 27ed3be 4d70c71 bbb69f6 ed0e073 4d70c71 27ed3be ed0e073 f41b9ef 27ed3be f41b9ef ed0e073 27ed3be 4d70c71 ed0e073 4d70c71 bbb69f6 4d70c71 bf38597 4d70c71 bcfa1bb 5799683 bbb69f6 bcfa1bb 1d8e2d0 9451027 bf38597 bcfa1bb 5799683 b974d61 bf38597 5799683 b9c4863 bf38597 4d70c71 9451027 5799683 bcfa1bb 27ed3be a48c771 bcfa1bb b9c4863 a48c771 899e654 9451027 bf38597 b9c4863 bcfa1bb 1f46917 bed08e5 1f46917 bcfa1bb 1f46917 f6874c9 |
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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 |
from starlette.responses import JSONResponse, FileResponse, HTMLResponse
from gradio.data_classes import FileData, GradioModel
from sse_starlette.sse import EventSourceResponse
from typing import (List, Tuple, Optional)
from fastapi import FastAPI, Request
import gradio as gr
import mimetypes
import threading
import requests
import argparse
import aiohttp
import uvicorn
import random
import string
import base64
import json
import time
import math
import sys
import os
# --- === CONFIG === ---
ENV_HANDLE = "env"#or "url on env"
IMAGE_HANDLE = "url"# or "base64"
API_BASE = "env"# or "openai"
GRADIO_URL_PATH = "https://quardo-gpt-4o-mini.hf.space/"
api_key = os.environ['API_API_KEY']
oai_api_key = os.environ['OPENAI_API_KEY']
base_url = os.environ.get('OPENAI_BASE_URL', "https://api.openai.com/v1")
# Will not add O1-mini, and O1-preview into the default, as it requeires TIER-5 sub on OpenAI's API.
# But if you wanna add O1 just remove this comment line and comment the other
# def_models = '["gpt-4", "gpt-4-0125-preview", "gpt-4-0613", "gpt-4-1106-preview", "gpt-4-turbo", "gpt-4-turbo-2024-04-09", "gpt-4-turbo-preview", "chatgpt-4o-latest", "gpt-4o", "gpt-4o-2024-05-13", "gpt-4o-2024-08-06", "gpt-4o-mini", "gpt-4o-mini-2024-07-18", "o1-mini", "o1-mini-2024-09-12", "o1-preview", "o1-preview-2024-09-12"]'
def_models = '["gpt-4", "gpt-4-0125-preview", "gpt-4-0613", "gpt-4-1106-preview", "gpt-4-turbo", "gpt-4-turbo-2024-04-09", "gpt-4-turbo-preview", "chatgpt-4o-latest", "gpt-4o", "gpt-4o-2024-05-13", "gpt-4o-2024-08-06", "gpt-4o-mini", "gpt-4o-mini-2024-07-18"]'
vision_models = {"gpt-4-turbo", "chatgpt-4o-latest", "gpt-4o", "gpt-4o-2024-05-13", "gpt-4o-2024-08-06", "gpt-4o-mini", "gpt-4o-mini-2024-07-18"}
calcPrompt = """[System: You have ability to calculate math problems (formated on python), via tools provided to you by system.
To call a tool you need to write a json in a empty line; like writing it at the end of message.
To use calculator; you need to follow this example JSON:
{"tool": "calc", "isCall": true, "prompt": "math.pi * 5"}
> 'tool' variable is used to define which tool you are calling
> 'isCall' used to confirm that you are calling that function and not showing it for example
> 'prompt' the math that will be done via python.
Here's few more example so you can under stand better
To show as an example>
{"tool": "calc", "isCall": false, "prompt": "math.sqrt(16)"}
{"tool": "calc", "isCall": false, "prompt": "math.pow(2, 3)"}
{"tool": "calc", "isCall": false, "prompt": "math.sin(math.pi / 2)"}
To actually use the tool>
{"tool": "calc", "isCall": true, "prompt": "math.factorial(5)"}
{"tool": "calc", "isCall": true, "prompt": "math.log(100, 10)"}
{"tool": "calc", "isCall": true, "prompt": "math.cos(0)"}
In chat use examples:
1.
Please, wait while I calculate 2+2...
{"tool": "calc", "isCall": false, "prompt": "2+2"}
2.
Plase, wait while I calculate the square root of 25...
{"tool": "calc", "isCall": true, "prompt": "math.sqrt(25)"}
3.
This is how I perform calculations:
{"tool": "calc", "isCall": false, "prompt": "math.pow(3, 2)"}
4. (Do not do this, this would block the user from seeing the result.)
Alright! Here's the result of a complex calculation involving trigonometry and logarithms. ```
{"tool": "calc", "isCall": true, "prompt": "math.sin(math.pi / 4) + math.log(10, 10)"}
]""";
searchPrompt = """[System: You have ability to search queries on a search engine, via tools provided to you by system.
(Warning: Each search call can take up to 30 or more seconds. Only one search function can be called per round. If a response has already been received, the system will answer based on that response. If the query needs to be searched again, the system will ask the user if they want to requery.)
To call a tool you need to write a json in a empty line; like writing it at the end of message.
To look up queries; you need to follow this example JSON:
{"tool": "search", "isCall": true, "prompt": "What is the latest news on climate change?"}
> 'tool' variable is used to define which tool you are calling
> 'isCall' used to confirm that you are calling that function and not showing it for example
> 'prompt' the query that will be searched.
Here's a few more examples so you can understand better
To show as an example>
{"tool": "search", "isCall": false, "prompt": "How to bake a chocolate cake?"}
{"tool": "search", "isCall": false, "prompt": "What are the symptoms of the flu?"}
{"tool": "search", "isCall": false, "prompt": "Best practices for remote work"}
To actually use the tool>
{"tool": "search", "isCall": true, "prompt": "How to invest in stocks?"}
{"tool": "search", "isCall": true, "prompt": "What is the current status of the Mars rover?"}
{"tool": "search", "isCall": true, "prompt": "Latest advancements in AI technology"}
In chat use examples:
1.
Please, wait while I search for the latest trends in technology...
{"tool": "search", "isCall": false, "prompt": "Latest trends in technology"}
2.
Please, wait while I search for the best ways to improve mental health...
{"tool": "search", "isCall": true, "prompt": "Best ways to improve mental health"}
3.
This is how I perform searches:
{"tool": "search", "isCall": false, "prompt": "How to start a garden?"}
4. (Do not do this, this would block the user from seeing the result.)
Alright! Here's the result of a search on the impact of social media on teenagers. ```
{"tool": "search", "isCall": true, "prompt": "Impact of social media on teenagers"}
]""";
# --- === CONFIG === ---
def loadENV():
def worker():
while True:
if ENV_HANDLE == "url on env":
try:
response = requests.get(os.environ["ENV_URL"])
response.raise_for_status()
env_data = response.json()
for key, value in env_data.items():
os.environ[key] = value
handleApiKeys()
checkModels()
loadModels()
except Exception as e:
print(f"Error loading environment variables: {e}")
time.sleep(180)
if ENV_HANDLE == "url on env":
try:
response = requests.get(os.environ["ENV_URL"])
response.raise_for_status()
env_data = response.json()
for key, value in env_data.items():
os.environ[key] = value
handleApiKeys()
checkModels()
loadModels()
except Exception as e:
print(f"Error loading environment variables: {e}")
threading.Thread(target=worker, daemon=True).start()
def checkModels():
global base_url
if API_BASE == "env":
try:
response = requests.get(f"{base_url}/models", headers={"Authorization": f"Bearer {get_api_key()}"})
response.raise_for_status()
if not ('data' in response.json()):
base_url = "https://api.openai.com/v1"
api_key = oai_api_key
except Exception as e:
print(f"Error testing API endpoint: {e}")
else:
base_url = "https://api.openai.com/v1"
api_key = oai_api_key
def loadModels():
global models, modelList
try:
models = json.loads(os.environ.get('OPENAI_API_MODELS', def_models))
except json.JSONDecodeError:
models = json.loads(def_models)
models = sorted(models)
modelList = {
"object": "list",
"data": [{"id": v, "object": "model", "created": 0, "owned_by": "system"} for v in models]
}
def handleApiKeys():
global api_key
if ',' in api_key:
output = []
for key in api_key.split(','):
try:
response = requests.get(f"{base_url}/models", headers={"Authorization": f"Bearer {key}"})
response.raise_for_status()
if ('data' in response.json()):
output.append(key)
except Exception as e:
print((F"API key {key} is not valid or an actuall error happend {e}"))
if len(output)==1:
raise RuntimeError("No API key is working")
api_key = ",".join(output)
else:
try:
response = requests.get(f"{base_url}/models", headers={"Authorization": f"Bearer {api_key}"})
response.raise_for_status()
if not ('data' in response.json()):
raise RuntimeError("Current API key is not valid")
except Exception as e:
raise RuntimeError(f"Current API key is not valid or an actual error happened: {e}")
def safe_eval(expression):
print(expression)
allowed_names = {name: obj for name, obj in math.__dict__.items() if not name.startswith("__")}
allowed_names['math'] = math
code = compile(expression, "<string>", "eval")
for name in code.co_names:
if name not in allowed_names and name != 'math':
raise NameError(f"Use of {name} is not allowed")
return eval(code, {"__builtins__": {}}, allowed_names)
def get_api_key(call='api_key'):
if call == 'api_key':
key = api_key
elif call == 'oai_api_key':
key = oai_api_key
else:
key = api_key
if ',' in key:
return random.choice(key.split(','))
return key
def encodeChat(messages):
output = []
for message in messages:
role = message['role']
name = f" [{message['name']}]" if 'name' in message else ''
content = message['content']
formatted_message = f"<|im_start|>{role}{name}\n{content}<|end_of_text|>"
output.append(formatted_message)
return "\n".join(output)
def moderate(messages):
try:
response = requests.post(
f"{base_url}/moderations",
headers={
"Content-Type": "application/json",
"Authorization": f"Bearer {get_api_key(call='api_key')}"
},
json={"input": encodeChat(messages)}
)
response.raise_for_status()
moderation_result = response.json()
except requests.exceptions.RequestException as e:
print(f"Error during moderation request to {base_url}: {e}")
try:
response = requests.post(
"https://api.openai.com/v1/moderations",
headers={
"Content-Type": "application/json",
"Authorization": f"Bearer {get_api_key(call='oai_api_key')}"
},
json={"input": encodeChat(messages)}
)
response.raise_for_status()
moderation_result = response.json()
except requests.exceptions.RequestException as e:
print(f"Error during moderation request to fallback URL: {e}")
return False
try:
if any(result["flagged"] for result in moderation_result["results"]):
return moderation_result
except KeyError:
if moderation_result["flagged"]:
return moderation_result
return False
async def streamChat(params):
if params.get("model") in ["o1-mini", "o1-mini-2024-09-12", "o1-preview", "o1-preview-2024-09-12"]:
if "temperature" in params:
del params["temperature"]
if "top_p" in params:
del params["top_p"]
if "max_tokens" in params:
params["max_completion_tokens"] = params.pop("max_tokens")
for message in params.get("messages", []):
if message["role"] == "system":
params["messages"].remove(message)
params["stream"] = False;
async with aiohttp.ClientSession() as session:
try:
async with session.post(f"{base_url}/chat/completions", headers={"Authorization": f"Bearer {get_api_key(call='api_key')}", "Content-Type": "application/json"}, json=params) as r:
r.raise_for_status()
response_data = await r.json()
yield {"choices": [{"delta": {"content": response_data["choices"][0]["message"]["content"]}}]}
except aiohttp.ClientError:
try:
async with session.post("https://api.openai.com/v1/chat/completions", headers={"Authorization": f"Bearer {get_api_key(call='oai_api_key')}", "Content-Type": "application/json"}, json=params) as r:
r.raise_for_status()
response_data = await r.json()
yield {"choices": [{"delta": {"content": response_data["choices"][0]["message"]["content"]}}]}
except aiohttp.ClientError:
return
else:
async with aiohttp.ClientSession() as session:
try:
async with session.post(f"{base_url}/chat/completions", headers={"Authorization": f"Bearer {get_api_key(call='api_key')}", "Content-Type": "application/json"}, json=params) as r:
r.raise_for_status()
async for line in r.content:
if line:
line_str = line.decode('utf-8')
if line_str.startswith("data: "):
line_str = line_str[6:].strip()
if line_str == "[DONE]":
continue
try:
message = json.loads(line_str)
yield message
except json.JSONDecodeError:
continue
except aiohttp.ClientError:
try:
async with session.post("https://api.openai.com/v1/chat/completions", headers={"Authorization": f"Bearer {get_api_key(call='oai_api_key')}", "Content-Type": "application/json"}, json=params) as r:
r.raise_for_status()
async for line in r.content:
if line:
line_str = line.decode('utf-8')
if line_str.startswith("data: "):
line_str = line_str[6:].strip()
if line_str == "[DONE]":
continue
try:
message = json.loads(line_str)
yield message
except json.JSONDecodeError:
continue
except aiohttp.ClientError:
return
def searchEngine(query):
### This /search endpoint is custom made, OpenAI does not have it.
### If you dupelicate this space, please either try to find another API or make one yourself.
response = requests.get(f"{base_url}/search?query={requests.utils.quote(query)}")
response.raise_for_status()
response_data = response.json()
return response_data.get("choices", [{}])[0].get("message", {}).get("content", "")
def rnd(length=8):
letters = string.ascii_letters + string.digits
return ''.join(random.choice(letters) for i in range(length))
def handleMultimodalData(model, role, data):
if isinstance(data, dict) and 'files' in data and len(data['files']) > 0:
result, handler, hasFoundFile = [], ["[System: This message contains files; the system will be splitting it.]"], False
for file in data['files']:
if file['mime_type'].startswith("image/") and model in vision_models:
if IMAGE_HANDLE == "base64":
with open(file['path'], "rb") as image_file:
encoded_image = base64.b64encode(image_file.read()).decode('utf-8')
result.append({"type": "image_url", "image_url": {"url": "data:" + file['mime_type'] + ";base64," + encoded_image}})
else:
result.append({"type": "image_url", "image_url": {"url": GRADIO_URL_PATH + "file=" + requests.utils.quote(str(file['path']))}})
if file['mime_type'].startswith("text/") or file['mime_type'].startswith("application/"):
hasFoundFile = True
try:
with open(file['path'], "rb") as data_file:
file_content = data_file.read().decode('utf-8')
handler.append("<|file_start|>" + file['orig_name'] + "\n" + file_content + "<|file_end|>")
except UnicodeDecodeError:
continue
if hasFoundFile:
handler.append(data.get('text', ''))
return {"role": role, "content": [{"type": "text", "text": "\n\n".join(handler)}] + result}
else:
return {"role": role, "content": [{"type": "text", "text": data.get('text', '')}] + result}
elif isinstance(data, Tuple):
file_path = data[0]
mime_type, _ = mimetypes.guess_type(file_path)
content = getattr(data, 'text', str(data))
if mime_type and mime_type.startswith("image/") and model in vision_models:
if IMAGE_HANDLE == "base64":
with open(file_path, "rb") as image_file:
encoded_image = base64.b64encode(image_file.read()).decode('utf-8')
return {"role": role, "content": [{"type": "image_url", "image_url": {"url": "data:" + mime_type + ";base64," + encoded_image}}]}
else:
return {"role": role, "content": [{"type": "image_url", "image_url": {"url": GRADIO_URL_PATH + "file=" + requests.utils.quote(file_path)}}]}
elif mime_type and (mime_type.startswith("text/") or mime_type.startswith("application/")):
try:
with open(file_path, "rb") as data_file:
file_content = data_file.read().decode('utf-8')
return {"role": role, "content": [{"type": "text", "text": "<|file_start|>" + os.path.basename(file_path) + "\n" + file_content + "<|file_end|>"}]}
except UnicodeDecodeError:
return {"role": role, "content": [{"type": "text", "text": content}]}
else:
return {"role": role, "content": content}
else:
return {"role": role, "content": getattr(data, 'text', str(data))}
class FileMessage(GradioModel):
file: FileData
alt_text: Optional[str] = None
class MultimodalMessage(GradioModel):
text: Optional[str] = None
files: Optional[List[FileMessage]]
async def respond(
message,
history: List[Tuple[
Optional[MultimodalMessage],
Optional[MultimodalMessage],
]],
system_message,
model_name,
max_tokens,
temperature,
top_p,
seed,
random_seed,
calcBeta,
searchBeta,
betterSystemPrompt
):
messages = [];
if calcBeta:
messages.append({"role": "system", "content": calcPrompt});
if searchBeta:
messages.append({"role": "system", "content": searchPrompt});
if betterSystemPrompt:
messages.append({"role": "system", "content": f"You are a helpful assistant. You are an OpenAI GPT model named {model_name}. The current time is {time.strftime('%Y-%m-%d %H:%M:%S')}. Please adhere to OpenAI's usage policies and guidelines. Ensure your responses are accurate, respectful, and within the scope of OpenAI's rules."});
else:
messages.append({"role": "system", "content": system_message});
for val in history:
if val[0] is not None:
user_message = handleMultimodalData(model_name, "user", val[0])
if user_message:
messages.append(user_message)
if val[1] is not None:
assistant_message = handleMultimodalData(model_name, "assistant", val[1])
if assistant_message:
messages.append(assistant_message)
if message:
user_message = handleMultimodalData(model_name, "user", message)
if user_message:
messages.append(user_message)
mode = moderate(messages)
if mode:
reasons = []
categories = mode[0].get('categories', {}) if isinstance(mode, list) else mode.get('categories', {})
for category, flagged in categories.items():
if flagged:
reasons.append(category)
if reasons:
yield "[MODERATION] I'm sorry, but I can't assist with that.\n\nReasons:\n```\n" + "\n".join([f"{i+1}. {reason}" for i, reason in enumerate(reasons)]) + "\n```"
else:
yield "[MODERATION] I'm sorry, but I can't assist with that."
return
async def handleResponse(completion, prefix="", didSearchedAlready=False):
response = ""
isRequeryNeeded = False
async for token in completion:
response += token['choices'][0]['delta'].get("content", token['choices'][0]['delta'].get("refusal", ""))
yield f"{prefix}{response}"
mode = moderate(messages + [{"role": "assistant", "content": response}])
if mode:
reasons = []
categories = mode[0].get('categories', {}) if isinstance(mode, list) else mode.get('categories', {})
for category, flagged in categories.items():
if flagged:
reasons.append(category)
if reasons:
yield "[MODERATION] I'm sorry, but I can't assist with that.\n\nReasons:\n```\n" + "\n".join([f"{i+1}. {reason}" for i, reason in enumerate(reasons)]) + "\n```"
else:
yield "[MODERATION] I'm sorry, but I can't assist with that."
return
for line in response.split('\n'):
try:
data = json.loads(line)
if isinstance(data, dict) and data.get("tool") == "calc" and data.get("isCall") and "prompt" in data:
isRequeryNeeded = True
try:
result = safe_eval(data["prompt"])
response = response.replace(line, f'[System: `{data["prompt"]}` === `{result}`]')
except Exception as e:
response = response.replace(line, f'[System: Error in calculation; `{e}`]')
yield f"{prefix}{response}"
elif isinstance(data, dict) and data.get("tool") == "search" and data.get("isCall") and "prompt" in data:
isRequeryNeeded = True
if didSearchedAlready:
response = response.replace(line, f'[System: One search per response is allowed; due to how long and resource it takes; query: `{data["prompt"]}]`]')
else:
try:
result = searchEngine(data["prompt"])
result_escaped = result.replace('`', '\\`')
response = response.replace(line, f'[System: `{data["prompt"]}` ===\n```\n{result_escaped}\n```\n]')
didSearchedAlready = True
except Exception as e:
response = response.replace(line, f'[System: Error in search function; `{e}`]')
yield f"{prefix}{response}"
yield f"{prefix}{response}"
except (json.JSONDecodeError, AttributeError, Exception):
continue
if isRequeryNeeded:
messages.append({"role": "assistant", "content": response})
async for res in handleResponse(streamChat({
"model": model_name,
"messages": messages,
"max_tokens": max_tokens,
"temperature": temperature,
"top_p": top_p,
"seed": (random.randint(0, 2**32) if random_seed else seed),
"user": rnd(),
"stream": True
}), f"{prefix}{response}\n\n", didSearchedAlready):
yield res
async for res in handleResponse(streamChat({
"model": model_name,
"messages": messages,
"max_tokens": max_tokens,
"temperature": temperature,
"top_p": top_p,
"seed": (random.randint(0, 2**32) if random_seed else seed),
"user": rnd(),
"stream": True
})):
yield res
handleApiKeys();loadModels();checkModels();loadENV();
lastUpdateMessage = "Back online yet again. Removed support for image generation."
demo = gr.ChatInterface(
respond,
title="gpt-4o-mini",
description=f"A OpenAI API proxy!<br/>View API docs [here](/api/v1/docs) <strong>[Yes you can use this as an API in a simpler manner]</strong>.<br/><strong>[Last update: {lastUpdateMessage}]</strong> Also you can only submit images to vision models; txt/code/etc. to all models.",
multimodal=True,
additional_inputs=[
gr.Textbox(value="You are a helpful assistant. You are an OpenAI GPT model. Please adhere to OpenAI's usage policies and guidelines. Ensure your responses are accurate, respectful, and within the scope of OpenAI's rules.", label="System message"),
gr.Dropdown(choices=models, value="gpt-4o-mini", label="Model"),
gr.Slider(minimum=1, maximum=4096, value=4096, step=1, label="Max new tokens"),
gr.Slider(minimum=0.1, maximum=2.0, value=0.7, step=0.05, label="Temperature"),
gr.Slider(minimum=0.05, maximum=1.0, value=0.95, step=0.05, label="Top-p (nucleus sampling)"),
gr.Slider(minimum=0, maximum=2**32, value=0, step=1, label="Seed"),
gr.Checkbox(label="Randomize Seed", value=True),
gr.Checkbox(label="FakeTool [Calculator beta]", value=True),
gr.Checkbox(label="FakeTool [Search engine beta (Warning; each query takes up to 30 seconds)]", value=True),
gr.Checkbox(label="Better system prompt (ignores the system prompt set by user.)", value=True),
],
css="footer{display:none !important}",
head="""<script>if(!confirm("By using our application, which integrates with OpenAI's API, you acknowledge and agree to the following terms regarding the data you provide:\\n\\n1. Data Collection: This application may log the following data through the Gradio endpoint or the API endpoint: message requests (including messages, responses, model settings, and images sent along with the messages), images that were generated (including only the prompt and the image), search tool calls (including query, search results, summaries, and output responses), and moderation checks (including input and output).\\n2. Data Retention and Removal: Data is retained until further notice or until a specific request for removal is made.\\n3. Data Usage: The collected data may be used for various purposes, including but not limited to, administrative review of logs, AI training, and publication as a dataset.\\n4. Privacy: Please avoid sharing any personal information.\\n\\nBy continuing to use our application, you explicitly consent to the collection, use, and potential sharing of your data as described above. If you disagree with our data collection, usage, and sharing practices, we advise you not to use our application."))location.href="/declined";</script>"""
)
app = FastAPI()
@app.get("/declined")
def test():
return HTMLResponse(content="""
<html>
<head>
<title>Declined</title>
</head>
<body>
<p>Ok, you can go back to Hugging Face. I just didn't have any idea how to handle decline so you are redirected here.</p><br/>
<a href="/">Go back</button>
</body>
</html>
""")
@app.get("/api/v1/docs")
def html():
return FileResponse("index.html")
app = gr.mount_gradio_app(app, demo, path="/")
class ArgParser(argparse.ArgumentParser):
def __init__(self, *args, **kwargs):
super(ArgParser, self).__init__(*args, **kwargs)
self.add_argument("-s", "--server", type=str, default="0.0.0.0")
self.add_argument("-p", "--port", type=int, default=7860)
self.add_argument("-d", "--dev", default=False, action="store_true")
self.args = self.parse_args(sys.argv[1:])
if __name__ == "__main__":
args = ArgParser().args
if args.dev:
uvicorn.run("__main__:app", host=args.server, port=args.port, reload=True)
else:
uvicorn.run("__main__:app", host=args.server, port=args.port, reload=False)
|