Spaces:
Running
Running
File size: 10,845 Bytes
b9c4863 bf38597 b9c4863 bcfa1bb bf38597 1f46917 bf38597 1f46917 b9c4863 bf38597 b9c4863 1f46917 bf38597 bcfa1bb b9c4863 bcfa1bb b9c4863 bf38597 b9c4863 bcfa1bb bf38597 bcfa1bb b9c4863 bcfa1bb b9c4863 bf38597 bcfa1bb bf38597 bcfa1bb bf38597 bcfa1bb bf38597 bcfa1bb bf38597 bcfa1bb bf38597 bcfa1bb b9c4863 bf38597 bcfa1bb bed08e5 b9c4863 bf38597 bcfa1bb bf38597 bcfa1bb b9c4863 bf38597 bcfa1bb b9c4863 bf38597 b9c4863 bf38597 b9c4863 bf38597 b9c4863 bf38597 b9c4863 bf38597 b9c4863 bcfa1bb 1f46917 bed08e5 1f46917 bcfa1bb 1f46917 |
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 |
from sse_starlette.sse import EventSourceResponse
from starlette.responses import JSONResponse, FileResponse
from fastapi import FastAPI, Request
import gradio as gr
import requests
import argparse
import aiohttp
import uvicorn
import random
import string
import base64
import json
import sys
import os
# --- === CONFIG === ---
IMAGE_HANDLE = "url"# or "base64"
API_BASE = "openai"# or "env"
api_key = os.environ['OPENAI_API_KEY']
base_url = os.environ.get('OPENAI_BASE_URL', "https://api.openai.com/v1")
# --- === CONFIG === ---
if API_BASE == "env":
try:
response = requests.get(f"{base_url}/models", headers={"Authorization": f"Bearer {api_key}"})
response.raise_for_status()
models = response.json()
if not ('data' in models):
base_url = "https://api.openai.com/v1"
except Exception as e:
print(f"Error testing API endpoint: {e}")
else:
base_url = "https://api.openai.com/v1"
async def streamChat(params):
async with aiohttp.ClientSession() as session:
async with session.post(f"{base_url}/chat/completions", headers={"Authorization": f"Bearer {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
def rnd(length=8):
letters = string.ascii_letters + string.digits
return ''.join(random.choice(letters) for i in range(length))
def getModels():
response = requests.get(f"{base_url}/models", headers={"Authorization": f"Bearer {api_key}",})
response.raise_for_status()
models = response.json()
return sorted([
model['id'] for model in models['data']
if 'gpt' in model['id'] and model['id'] not in {"gpt-3.5-turbo-instruct", "gpt-3.5-turbo-instruct-0914"}
])
def handleMultimodalData(model, role, data):
if type(data) == str:
return {"role": role, "content": str(data)}
elif isinstance(data, str):
return {"role": role, "content": data.text}
elif hasattr(data, 'files') and data.files and len(data.files) > 0 and model in {"gpt-4-1106-vision-preview", "gpt-4-vision-preview", "gpt-4-turbo", "gpt-4o", "gpt-4o-2024-05-13", "gpt-4o-mini", "gpt-4o-mini-2024-07-18"}:
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/"):
if IMAGE_HANDLE == "base64":
with open(file.path, "rb") as image_file:
result.append({"type": "image_url", "image_url": {"url": "data:" + file.mime_type + ";base64," + base64.b64encode(image_file.read()).decode('utf-8')}})
image_file.close()
else:
result.append({"type": "image_url", "image_url": {"url": file.url}})
if file.mime_type.startswith("text/") or file.mime_type.startswith("application/"):
hasFoundFile = True
with open(file.path, "rb") as data_file:
handler.append("<|file_start|>" + file.orig_name + "\n" + data_file.read().decode('utf-8') + "<|file_end|>")
if hasFoundFile:
handler.append(data.text)
return {"role": role, "content": [{"type": "text", "text": "\n\n".join(handler)}] + result}
else:
return {"role": role, "content": [{"type": "text", "text": data.text}] + result}
elif hasattr(data, 'files') and data.files and len(data.files) > 0 and not (model in {"gpt-4-1106-vision-preview", "gpt-4-vision-preview", "gpt-4-turbo", "gpt-4o", "gpt-4o-2024-05-13", "gpt-4o-mini", "gpt-4o-mini-2024-07-18"}):
handler, hasFoundFile = ["[System: This message contains files; the system will be splitting it.]"], False
for file in data.files:
if file.mime_type.startswith("text/") or file.mime_type.startswith("application/"):
hasFoundFile = True
with open(file.path, "rb") as data_file:
handler.append("<|file_start|>" + file.orig_name + "\n" + data_file.read().decode('utf-8') + "<|file_end|>")
if hasFoundFile:
handler.append(data.text)
return {"role": role, "content": "\n\n".join(handler)}
else:
return {"role": role, "content": data.text}
else:
if isinstance(data, tuple):
return {"role": role, "content": str(data)}
return {"role": role, "content": getattr(data, 'text', str(data))}
async def respond(
message,
history: list[tuple[str, str]],
system_message,
model_name,
max_tokens,
temperature,
top_p,
seed,
random_seed
):
messages = [{"role": "system", "content": "If user submits any file that file will be visible only that turn. This is not due to privacy related things but rather due to developer's lazyness; Ask user to upload the file again if they ask a follow-up question without the data."}, {"role": "system", "content": system_message}]
for val in history:
if val[0]:
messages.append(handleMultimodalData(model_name,"user",val[0]))
if val[1]:
messages.append(handleMultimodalData(model_name,"assistant",val[1]))
messages.append(handleMultimodalData(model_name,"user",message))
response = ""
completion = 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
})
async for token in completion:
response += token['choices'][0]['delta'].get("content", "")
yield response
demo = gr.ChatInterface(
respond,
title="GPT-4O-mini",
description="A simple proxy to OpenAI!<br/>You can use this space as a proxy! click [here](/api/v1/docs) to view the documents.<br/>Also you can only submit images to vision/4o models but can submit txt/code/etc. files to all models.<br/>###### Also the file queries are only shown to model for 1 round cuz gradio.",
multimodal=True,
additional_inputs=[
gr.Textbox(value="You are a helpful assistant.", label="System message"),
gr.Dropdown(choices=getModels(), value="gpt-4o-mini-2024-07-18", 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),
],
)
app = FastAPI()
@app.get("/api/v1/docs")
def html():
return FileResponse("index.html")
@app.get("/api/v1/models")
async def test_endpoint():
response = requests.get(f"{base_url}/models", headers={"Authorization": f"Bearer {api_key}"})
response.raise_for_status()
models = response.json()
models['data'] = sorted(
[model for model in models['data'] if 'gpt' in model['id'] and model['id'] not in {"gpt-3.5-turbo-instruct", "gpt-3.5-turbo-instruct-0914"}],
key=lambda x: x['id']
)
return JSONResponse(content=models)
@app.post("/api/v1/chat/completions")
async def chat_completion(request: Request):
try:
body = await request.json()
if not body.get("messages") or not body.get("model"):
return JSONResponse(content={"error": { "code": "MISSING_VALUE", "message": "Both 'messages' and 'model' are required fields."}}, status_code=400)
params = {
key: value for key, value in {
"model": body.get("model"),
"messages": body.get("messages"),
"max_tokens": body.get("max_tokens"),
"temperature": body.get("temperature"),
"top_p": body.get("top_p"),
"frequency_penalty": body.get("frequency_penalty"),
"logit_bias": body.get("logit_bias"),
"logprobs": body.get("logprobs"),
"top_logprobs": body.get("top_logprobs"),
"n": body.get("n"),
"presence_penalty": body.get("presence_penalty"),
"response_format": body.get("response_format"),
"seed": body.get("seed"),
"service_tier": body.get("service_tier"),
"stop": body.get("stop"),
"stream": body.get("stream"),
"stream_options": body.get("stream_options"),
"tools": body.get("tools"),
"tool_choice": body.get("tool_choice"),
"parallel_tool_calls": body.get("parallel_tool_calls"),
"user": rnd(),
}.items() if value is not None
}
if body.get("stream"):
async def event_generator():
async for event in streamChat(params):
yield json.dumps(event)
return EventSourceResponse(event_generator())
else:
response = requests.post(f"{base_url}/chat/completions", headers={"Authorization": f"Bearer {api_key}", "Content-Type": "application/json"}, json=params)
response.raise_for_status()
completion = response.json()
return JSONResponse(content=completion)
except Exception as e:
return JSONResponse(content={"error": { "code": "SERVER_ERROR", "message": str(e)}}, status_code=400)
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) |