Update app.py
Browse files
app.py
CHANGED
|
@@ -32,30 +32,6 @@ handler.setLevel(logging.INFO)
|
|
| 32 |
app.logger.addHandler(handler)
|
| 33 |
app.logger.setLevel(logging.INFO)
|
| 34 |
|
| 35 |
-
SYSTEM_ASSISTANT = """作为 Stable Diffusion Prompt 提示词专家,您将从关键词中创建提示,通常来自 Danbooru 等数据库。
|
| 36 |
-
提示通常描述图像,使用常见词汇,按重要性排列,并用逗号分隔。避免使用"-"或".",但可以接受空格和自然语言。避免词汇重复。
|
| 37 |
-
为了强调关键词,请将其放在括号中以增加其权重。例如,"(flowers)"将'flowers'的权重增加1.1倍,而"(((flowers)))"将其增加1.331倍。使用"(flowers:1.5)"将'flowers'的权重增加1.5倍。只为重要的标签增加权重。
|
| 38 |
-
提示包括三个部分:**前缀** (质量标签+风格词+效果器)+ **主题** (图像的主要焦点)+ **场景** (背景、环境)。
|
| 39 |
-
* 前缀影响图像质量。像"masterpiece"、"best quality"、"4k"这样的标签可以提高图像的细节。像"illustration"、"lensflare"这样的风格词定义图像的风格。像"bestlighting"、"lensflare"、"depthoffield"这样的效果器会影响光照和深度。
|
| 40 |
-
* 主题是图像的主要焦点,如角色或场景。对主题进行详细描述可以确保图像丰富而详细。增加主题的权重以增强其清晰度。对于角色,描述面部、头发、身体、服装、姿势等特征。
|
| 41 |
-
* 场景描述环境。没有场景,图像的背景是平淡的,主题显得过大。某些主题本身包含场景(例如建筑物、风景)。像"花草草地"、"阳光"、"河流"这样的环境词可以丰富场景。你的任务是设计图像生成的提示。请按照以下步骤进行操作:
|
| 42 |
-
1. 我会发送给您一个图像场景。需要你生成详细的图像描述
|
| 43 |
-
2. 图像描述必须是英文,输出为Positive Prompt。
|
| 44 |
-
示例:
|
| 45 |
-
我发送:二战时期的护士。
|
| 46 |
-
您回复只回复:
|
| 47 |
-
A WWII-era nurse in a German uniform, holding a wine bottle and stethoscope, sitting at a table in white attire, with a table in the background, masterpiece, best quality, 4k, illustration style, best lighting, depth of field, detailed character, detailed environment.
|
| 48 |
-
"""
|
| 49 |
-
|
| 50 |
-
RATIO_MAP = {
|
| 51 |
-
"1:1": "1024x1024",
|
| 52 |
-
"1:2": "1024x2048",
|
| 53 |
-
"3:2": "1536x1024",
|
| 54 |
-
"4:3": "1536x2048",
|
| 55 |
-
"16:9": "2048x1152",
|
| 56 |
-
"9:16": "1152x2048"
|
| 57 |
-
}
|
| 58 |
-
|
| 59 |
# 模型映射
|
| 60 |
MODEL_MAPPING = {
|
| 61 |
"flux.1-schnell": {
|
|
@@ -84,12 +60,37 @@ MODEL_MAPPING = {
|
|
| 84 |
}
|
| 85 |
}
|
| 86 |
|
| 87 |
-
|
| 88 |
-
|
| 89 |
-
|
| 90 |
-
|
| 91 |
-
|
| 92 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 93 |
|
| 94 |
def get_random_token(auth_header):
|
| 95 |
if not auth_header:
|
|
@@ -153,7 +154,7 @@ def index():
|
|
| 153 |
</style>
|
| 154 |
</head>
|
| 155 |
<body>
|
| 156 |
-
<h1>Text-to-Image API with SiliconFlow!</h1>
|
| 157 |
|
| 158 |
<h2>Usage:</h2>
|
| 159 |
<ol>
|
|
@@ -166,6 +167,7 @@ def index():
|
|
| 166 |
</ul>
|
| 167 |
</li>
|
| 168 |
</ol>
|
|
|
|
| 169 |
<h2>Example Request:</h2>
|
| 170 |
<pre><code>
|
| 171 |
{
|
|
@@ -178,46 +180,13 @@ def index():
|
|
| 178 |
]
|
| 179 |
}
|
| 180 |
</code></pre>
|
|
|
|
| 181 |
<p>For more details, please refer to the API documentation.</p>
|
| 182 |
</body>
|
| 183 |
</html>
|
| 184 |
"""
|
| 185 |
return usage, 200
|
| 186 |
|
| 187 |
-
@app.route('/ai/v1/models', methods=['GET'])
|
| 188 |
-
def get_models():
|
| 189 |
-
try:
|
| 190 |
-
# 验证身份
|
| 191 |
-
auth_cookie = getAuthCookie(request)
|
| 192 |
-
if not auth_cookie:
|
| 193 |
-
return jsonify({"error": "Unauthorized"}), 401
|
| 194 |
-
|
| 195 |
-
# 返回模型列表
|
| 196 |
-
models_list = [
|
| 197 |
-
{
|
| 198 |
-
"id": model_id,
|
| 199 |
-
"object": "model",
|
| 200 |
-
"created": int(time.time()),
|
| 201 |
-
"owned_by": info["provider"],
|
| 202 |
-
"permission": [],
|
| 203 |
-
"root": model_id,
|
| 204 |
-
"parent": None
|
| 205 |
-
}
|
| 206 |
-
for model_id, info in MODEL_MAPPING.items()
|
| 207 |
-
]
|
| 208 |
-
|
| 209 |
-
# 记录日志
|
| 210 |
-
app.logger.info(f'{request.remote_addr} - GET /ai/v1/models - 200')
|
| 211 |
-
|
| 212 |
-
return jsonify({
|
| 213 |
-
"object": "list",
|
| 214 |
-
"data": models_list
|
| 215 |
-
})
|
| 216 |
-
|
| 217 |
-
except Exception as error:
|
| 218 |
-
app.logger.error(f"Error: {str(error)}")
|
| 219 |
-
return jsonify({"error": "Authentication failed", "details": str(error)}), 401
|
| 220 |
-
|
| 221 |
@app.route('/ai/v1/chat/completions', methods=['POST'])
|
| 222 |
def handle_request():
|
| 223 |
try:
|
|
@@ -295,6 +264,25 @@ def handle_request():
|
|
| 295 |
app.logger.error(f"Error: {str(e)}")
|
| 296 |
return jsonify({"error": f"Internal Server Error: {str(e)}"}), 500
|
| 297 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 298 |
def stream_response(unique_id, image_data, original_prompt, translated_prompt, size, created, model, system_fingerprint, use_original):
|
| 299 |
return Response(stream_with_context(generate_stream(unique_id, image_data, original_prompt, translated_prompt, size, created, model, system_fingerprint, use_original)), content_type='text/event-stream')
|
| 300 |
|
|
|
|
| 32 |
app.logger.addHandler(handler)
|
| 33 |
app.logger.setLevel(logging.INFO)
|
| 34 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 35 |
# 模型映射
|
| 36 |
MODEL_MAPPING = {
|
| 37 |
"flux.1-schnell": {
|
|
|
|
| 60 |
}
|
| 61 |
}
|
| 62 |
|
| 63 |
+
SYSTEM_ASSISTANT = """作为 Stable Diffusion Prompt 提示词专家,您将从关键词中创建提示,通常来自 Danbooru 等数据库。
|
| 64 |
+
提示通常描述图像,使用常见词汇,按重要性排列,并用逗号分隔。避免使用"-"或".",但可以接受空格和自然语言。避免词汇重复。
|
| 65 |
+
|
| 66 |
+
为了强调关键词,请将其放在括号中以增加其权重。例如,"(flowers)"将'flowers'的权重增加1.1倍,而"(((flowers)))"将其增加1.331倍。使用"(flowers:1.5)"将'flowers'的权重增加1.5倍。只为重要的标签增加权重。
|
| 67 |
+
|
| 68 |
+
提示包括三个部分:**前缀**(质量标签+风格词+效果器)+ **主题**(图像的主要焦点)+ **场景**(背景、环境)。
|
| 69 |
+
|
| 70 |
+
* 前缀影响图像质量。像"masterpiece"、"best quality"、"4k"这样的标签可以提高图像的细节。像"illustration"、"lensflare"这样的风格词定义图像的风格。像"bestlighting"、"lensflare"、"depthoffield"这样的效果器会影响光照和深度。
|
| 71 |
+
|
| 72 |
+
* 主题是图像的主要焦点,如角色或场景。对主题进行详细描述可以确保图像丰富而详细。增加主题的权重以增强其清晰度。对于角色,描述面部、头发、身体、服装、姿势等特征。
|
| 73 |
+
|
| 74 |
+
* 场景描述环境。没有场景,图像的背景是平淡的,主题显得过大。某些主题本身包含场景(例如建筑物、风景)。像"花草草地"、"阳光"、"河流"这样的环境词可以丰富场景。你的任务是设计图像生成的提示。请按照以下步骤进行操作:
|
| 75 |
+
|
| 76 |
+
1. 我会发送给您一个图像场景。需要你生成详细的图像描述
|
| 77 |
+
2. 图像描述必须是英文,输出为Positive Prompt。
|
| 78 |
+
|
| 79 |
+
示例:
|
| 80 |
+
|
| 81 |
+
我发送:二战时期的护士。
|
| 82 |
+
您回复只回复:
|
| 83 |
+
A WWII-era nurse in a German uniform, holding a wine bottle and stethoscope, sitting at a table in white attire, with a table in the background, masterpiece, best quality, 4k, illustration style, best lighting, depth of field, detailed character, detailed environment.
|
| 84 |
+
"""
|
| 85 |
+
|
| 86 |
+
RATIO_MAP = {
|
| 87 |
+
"1:1": "1024x1024",
|
| 88 |
+
"1:2": "1024x2048",
|
| 89 |
+
"3:2": "1536x1024",
|
| 90 |
+
"4:3": "1536x2048",
|
| 91 |
+
"16:9": "2048x1152",
|
| 92 |
+
"9:16": "1152x2048"
|
| 93 |
+
}
|
| 94 |
|
| 95 |
def get_random_token(auth_header):
|
| 96 |
if not auth_header:
|
|
|
|
| 154 |
</style>
|
| 155 |
</head>
|
| 156 |
<body>
|
| 157 |
+
<h1>Welcome to the Text-to-Image API with SiliconFlow!</h1>
|
| 158 |
|
| 159 |
<h2>Usage:</h2>
|
| 160 |
<ol>
|
|
|
|
| 167 |
</ul>
|
| 168 |
</li>
|
| 169 |
</ol>
|
| 170 |
+
|
| 171 |
<h2>Example Request:</h2>
|
| 172 |
<pre><code>
|
| 173 |
{
|
|
|
|
| 180 |
]
|
| 181 |
}
|
| 182 |
</code></pre>
|
| 183 |
+
|
| 184 |
<p>For more details, please refer to the API documentation.</p>
|
| 185 |
</body>
|
| 186 |
</html>
|
| 187 |
"""
|
| 188 |
return usage, 200
|
| 189 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 190 |
@app.route('/ai/v1/chat/completions', methods=['POST'])
|
| 191 |
def handle_request():
|
| 192 |
try:
|
|
|
|
| 264 |
app.logger.error(f"Error: {str(e)}")
|
| 265 |
return jsonify({"error": f"Internal Server Error: {str(e)}"}), 500
|
| 266 |
|
| 267 |
+
@app.route('/ai/v1/models', methods=['GET'])
|
| 268 |
+
def get_models():
|
| 269 |
+
models_list = [
|
| 270 |
+
{
|
| 271 |
+
"id": key,
|
| 272 |
+
"object": "model",
|
| 273 |
+
"owned_by": value["provider"],
|
| 274 |
+
"mapping": value["mapping"]
|
| 275 |
+
}
|
| 276 |
+
for key, value in MODEL_MAPPING.items()
|
| 277 |
+
]
|
| 278 |
+
|
| 279 |
+
response = {
|
| 280 |
+
"object": "list",
|
| 281 |
+
"data": models_list
|
| 282 |
+
}
|
| 283 |
+
|
| 284 |
+
return jsonify(response)
|
| 285 |
+
|
| 286 |
def stream_response(unique_id, image_data, original_prompt, translated_prompt, size, created, model, system_fingerprint, use_original):
|
| 287 |
return Response(stream_with_context(generate_stream(unique_id, image_data, original_prompt, translated_prompt, size, created, model, system_fingerprint, use_original)), content_type='text/event-stream')
|
| 288 |
|