wenyongda/OpenDify
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

去除缓存区

Browse Source
This commit is contained in:
wenyongda 2025-06-23 10:34:47 +08:00
parent 5350860f78
commit ccf4e4cedc
1 changed files with 99 additions and 97 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

196
main.py
View File

@ -56,7 +56,7 @@ class DifyModelManager:
headers=headers, headers=headers,
params={"user": "default_user"} params={"user": "default_user"}
) )
if response.status_code == 200: if response.status_code == 200:
app_info = response.json() app_info = response.json()
return app_info.get("name", "Unknown App") return app_info.get("name", "Unknown App")
@ -71,7 +71,7 @@ class DifyModelManager:
"""刷新所有应用信息""" """刷新所有应用信息"""
self.name_to_api_key.clear() self.name_to_api_key.clear()
self.api_key_to_name.clear() self.api_key_to_name.clear()
for api_key in self.api_keys: for api_key in self.api_keys:
app_name = await self.fetch_app_info(api_key) app_name = await self.fetch_app_info(api_key)
if app_name: if app_name:
@ -112,14 +112,14 @@ def get_api_key(model_name):
def transform_openai_to_dify(openai_request, endpoint): def transform_openai_to_dify(openai_request, endpoint):
"""将OpenAI格式的请求转换为Dify格式""" """将OpenAI格式的请求转换为Dify格式"""
if endpoint == "/chat/completions": if endpoint == "/chat/completions":
messages = openai_request.get("messages", []) messages = openai_request.get("messages", [])
stream = openai_request.get("stream", False) stream = openai_request.get("stream", False)
# 尝试从历史消息中提取conversation_id # 尝试从历史消息中提取conversation_id
conversation_id = None conversation_id = None
# 提取system消息内容 # 提取system消息内容
system_content = "" system_content = ""
system_messages = [msg for msg in messages if msg.get("role") == "system"] system_messages = [msg for msg in messages if msg.get("role") == "system"]
@ -127,7 +127,7 @@ def transform_openai_to_dify(openai_request, endpoint):
system_content = system_messages[0].get("content", "") system_content = system_messages[0].get("content", "")
# 记录找到的system消息 # 记录找到的system消息
logger.info(f"Found system message: {system_content[:100]}{'...' if len(system_content) > 100 else ''}") logger.info(f"Found system message: {system_content[:100]}{'...' if len(system_content) > 100 else ''}")
if CONVERSATION_MEMORY_MODE == 2: # 零宽字符模式 if CONVERSATION_MEMORY_MODE == 2: # 零宽字符模式
if len(messages) > 1: if len(messages) > 1:
# 遍历历史消息，找到最近的assistant消息 # 遍历历史消息，找到最近的assistant消息
@ -138,15 +138,15 @@ def transform_openai_to_dify(openai_request, endpoint):
conversation_id = decode_conversation_id(content) conversation_id = decode_conversation_id(content)
if conversation_id: if conversation_id:
break break
# 获取最后一条用户消息 # 获取最后一条用户消息
user_query = messages[-1]["content"] if messages and messages[-1].get("role") != "system" else "" user_query = messages[-1]["content"] if messages and messages[-1].get("role") != "system" else ""
# 如果有system消息且是首次对话(没有conversation_id)，则将system内容添加到用户查询前 # 如果有system消息且是首次对话(没有conversation_id)，则将system内容添加到用户查询前
if system_content and not conversation_id: if system_content and not conversation_id:
user_query = f"系统指令: {system_content}\n\n用户问题: {user_query}" user_query = f"系统指令: {system_content}\n\n用户问题: {user_query}"
logger.info(f"[零宽字符模式] 首次对话，添加system内容到查询前") logger.info(f"[零宽字符模式] 首次对话，添加system内容到查询前")
dify_request = { dify_request = {
"inputs": {}, "inputs": {},
"query": user_query, "query": user_query,
@ -157,12 +157,12 @@ def transform_openai_to_dify(openai_request, endpoint):
else: # history_message模式(默认) else: # history_message模式(默认)
# 获取最后一条用户消息 # 获取最后一条用户消息
user_query = messages[-1]["content"] if messages and messages[-1].get("role") != "system" else "" user_query = messages[-1]["content"] if messages and messages[-1].get("role") != "system" else ""
# 构造历史消息 # 构造历史消息
if len(messages) > 1: if len(messages) > 1:
history_messages = [] history_messages = []
has_system_in_history = False has_system_in_history = False
# 检查历史消息中是否已经包含system消息 # 检查历史消息中是否已经包含system消息
for msg in messages[:-1]: # 除了最后一条消息 for msg in messages[:-1]: # 除了最后一条消息
role = msg.get("role", "") role = msg.get("role", "")
@ -171,12 +171,12 @@ def transform_openai_to_dify(openai_request, endpoint):
if role == "system": if role == "system":
has_system_in_history = True has_system_in_history = True
history_messages.append(f"{role}: {content}") history_messages.append(f"{role}: {content}")
# 如果历史中没有system消息但现在有system消息，则添加到历史的最前面 # 如果历史中没有system消息但现在有system消息，则添加到历史的最前面
if system_content and not has_system_in_history: if system_content and not has_system_in_history:
history_messages.insert(0, f"system: {system_content}") history_messages.insert(0, f"system: {system_content}")
logger.info(f"[history_message模式] 添加system内容到历史消息前") logger.info(f"[history_message模式] 添加system内容到历史消息前")
# 将历史消息添加到查询中 # 将历史消息添加到查询中
if history_messages: if history_messages:
history_context = "\n\n".join(history_messages) history_context = "\n\n".join(history_messages)
@ -184,7 +184,7 @@ def transform_openai_to_dify(openai_request, endpoint):
elif system_content: # 没有历史消息但有system消息 elif system_content: # 没有历史消息但有system消息
user_query = f"系统指令: {system_content}\n\n用户问题: {user_query}" user_query = f"系统指令: {system_content}\n\n用户问题: {user_query}"
logger.info(f"[history_message模式] 首次对话，添加system内容到查询前") logger.info(f"[history_message模式] 首次对话，添加system内容到查询前")
dify_request = { dify_request = {
"inputs": {}, "inputs": {},
"query": user_query, "query": user_query,
@ -193,21 +193,21 @@ def transform_openai_to_dify(openai_request, endpoint):
} }
return dify_request return dify_request
return None return None
def transform_dify_to_openai(dify_response, model="claude-3-5-sonnet-v2", stream=False): def transform_dify_to_openai(dify_response, model="claude-3-5-sonnet-v2", stream=False):
"""将Dify格式的响应转换为OpenAI格式""" """将Dify格式的响应转换为OpenAI格式"""
if not stream: if not stream:
# 首先获取回答内容，支持不同的响应模式 # 首先获取回答内容，支持不同的响应模式
answer = "" answer = ""
mode = dify_response.get("mode", "") mode = dify_response.get("mode", "")
# 普通聊天模式 # 普通聊天模式
if "answer" in dify_response: if "answer" in dify_response:
answer = dify_response.get("answer", "") answer = dify_response.get("answer", "")
# 如果是Agent模式，需要从agent_thoughts中提取回答 # 如果是Agent模式，需要从agent_thoughts中提取回答
elif "agent_thoughts" in dify_response: elif "agent_thoughts" in dify_response:
# Agent模式下通常最后一个thought包含最终答案 # Agent模式下通常最后一个thought包含最终答案
@ -216,12 +216,12 @@ def transform_dify_to_openai(dify_response, model="claude-3-5-sonnet-v2", stream
for thought in agent_thoughts: for thought in agent_thoughts:
if thought.get("thought"): if thought.get("thought"):
answer = thought.get("thought", "") answer = thought.get("thought", "")
# 只在零宽字符会话记忆模式时处理conversation_id # 只在零宽字符会话记忆模式时处理conversation_id
if CONVERSATION_MEMORY_MODE == 2: if CONVERSATION_MEMORY_MODE == 2:
conversation_id = dify_response.get("conversation_id", "") conversation_id = dify_response.get("conversation_id", "")
history = dify_response.get("conversation_history", []) history = dify_response.get("conversation_history", [])
# 检查历史消息中是否已经有会话ID # 检查历史消息中是否已经有会话ID
has_conversation_id = False has_conversation_id = False
if history: if history:
@ -231,14 +231,14 @@ def transform_dify_to_openai(dify_response, model="claude-3-5-sonnet-v2", stream
if decode_conversation_id(content) is not None: if decode_conversation_id(content) is not None:
has_conversation_id = True has_conversation_id = True
break break
# 只在新会话且历史消息中没有会话ID时插入 # 只在新会话且历史消息中没有会话ID时插入
if conversation_id and not has_conversation_id: if conversation_id and not has_conversation_id:
logger.info(f"[Debug] Inserting conversation_id: {conversation_id}, history_length: {len(history)}") logger.info(f"[Debug] Inserting conversation_id: {conversation_id}, history_length: {len(history)}")
encoded = encode_conversation_id(conversation_id) encoded = encode_conversation_id(conversation_id)
answer = answer + encoded answer = answer + encoded
logger.info(f"[Debug] Response content after insertion: {repr(answer)}") logger.info(f"[Debug] Response content after insertion: {repr(answer)}")
return { return {
"id": dify_response.get("message_id", ""), "id": dify_response.get("message_id", ""),
"object": "chat.completion", "object": "chat.completion",
@ -277,11 +277,11 @@ def encode_conversation_id(conversation_id):
"""将conversation_id编码为不可见的字符序列""" """将conversation_id编码为不可见的字符序列"""
if not conversation_id: if not conversation_id:
return "" return ""
# 使用Base64编码减少长度 # 使用Base64编码减少长度
import base64 import base64
encoded = base64.b64encode(conversation_id.encode()).decode() encoded = base64.b64encode(conversation_id.encode()).decode()
# 使用8种不同的零宽字符表示3位数字 # 使用8种不同的零宽字符表示3位数字
# 这样可以将编码长度进一步减少 # 这样可以将编码长度进一步减少
char_map = { char_map = {
@ -294,7 +294,7 @@ def encode_conversation_id(conversation_id):
'6': '\u2061', # 函数应用 '6': '\u2061', # 函数应用
'7': '\u2062', # 不可见乘号 '7': '\u2062', # 不可见乘号
} }
# 将Base64字符串转换为八进制数字 # 将Base64字符串转换为八进制数字
result = [] result = []
for c in encoded: for c in encoded:
@ -312,14 +312,14 @@ def encode_conversation_id(conversation_id):
val = 63 val = 63
else: # '=' else: # '='
val = 0 val = 0
# 每个Base64字符可以产生2个3位数字 # 每个Base64字符可以产生2个3位数字
first = (val >> 3) & 0x7 first = (val >> 3) & 0x7
second = val & 0x7 second = val & 0x7
result.append(char_map[str(first)]) result.append(char_map[str(first)])
if c != '=': # 不编码填充字符的后半部分 if c != '=': # 不编码填充字符的后半部分
result.append(char_map[str(second)]) result.append(char_map[str(second)])
return ''.join(result) return ''.join(result)
def decode_conversation_id(content): def decode_conversation_id(content):
@ -336,17 +336,17 @@ def decode_conversation_id(content):
'\u2061': '6', # 函数应用 '\u2061': '6', # 函数应用
'\u2062': '7', # 不可见乘号 '\u2062': '7', # 不可见乘号
} }
# 提取最后一段零宽字符序列 # 提取最后一段零宽字符序列
space_chars = [] space_chars = []
for c in reversed(content): for c in reversed(content):
if c not in char_to_val: if c not in char_to_val:
break break
space_chars.append(c) space_chars.append(c)
if not space_chars: if not space_chars:
return None return None
# 将零宽字符转换回Base64字符串 # 将零宽字符转换回Base64字符串
space_chars.reverse() space_chars.reverse()
base64_chars = [] base64_chars = []
@ -357,7 +357,7 @@ def decode_conversation_id(content):
val = (first << 3) | second val = (first << 3) | second
else: else:
val = first << 3 val = first << 3
# 转换回Base64字符 # 转换回Base64字符
if val < 26: if val < 26:
base64_chars.append(chr(val + ord('A'))) base64_chars.append(chr(val + ord('A')))
@ -369,17 +369,17 @@ def decode_conversation_id(content):
base64_chars.append('+') base64_chars.append('+')
else: else:
base64_chars.append('/') base64_chars.append('/')
# 添加Base64填充 # 添加Base64填充
padding = len(base64_chars) % 4 padding = len(base64_chars) % 4
if padding: if padding:
base64_chars.extend(['='] * (4 - padding)) base64_chars.extend(['='] * (4 - padding))
# 解码Base64字符串 # 解码Base64字符串
import base64 import base64
base64_str = ''.join(base64_chars) base64_str = ''.join(base64_chars)
return base64.b64decode(base64_str).decode() return base64.b64decode(base64_str).decode()
except Exception as e: except Exception as e:
logger.debug(f"Failed to decode conversation_id: {e}") logger.debug(f"Failed to decode conversation_id: {e}")
return None return None
@ -424,9 +424,9 @@ def chat_completions():
# 继续处理原始逻辑 # 继续处理原始逻辑
openai_request = request.get_json() openai_request = request.get_json()
logger.info(f"Received request: {json.dumps(openai_request, ensure_ascii=False)}") logger.info(f"Received request: {json.dumps(openai_request, ensure_ascii=False)}")
model = openai_request.get("model", "claude-3-5-sonnet-v2") model = openai_request.get("model", "claude-3-5-sonnet-v2")
# 验证模型是否支持 # 验证模型是否支持
api_key = get_api_key(model) api_key = get_api_key(model)
if not api_key: if not api_key:
@ -439,9 +439,9 @@ def chat_completions():
"code": "model_not_found" "code": "model_not_found"
} }
}, 404 }, 404
dify_request = transform_openai_to_dify(openai_request, "/chat/completions") dify_request = transform_openai_to_dify(openai_request, "/chat/completions")
if not dify_request: if not dify_request:
logger.error("Failed to transform request") logger.error("Failed to transform request")
return { return {
@ -463,11 +463,11 @@ def chat_completions():
if stream: if stream:
def generate(): def generate():
client = httpx.Client(timeout=None) client = httpx.Client(timeout=None)
def flush_chunk(chunk_data): def flush_chunk(chunk_data):
"""Helper function to flush chunks immediately""" """Helper function to flush chunks immediately"""
return chunk_data.encode('utf-8') return chunk_data.encode('utf-8')
def calculate_delay(buffer_size): def calculate_delay(buffer_size):
""" """
根据缓冲区大小动态计算延迟 根据缓冲区大小动态计算延迟
@ -481,7 +481,7 @@ def chat_completions():
return 0.01 # 20ms延迟 return 0.01 # 20ms延迟
else: # 内容很少，使用较慢的速度 else: # 内容很少，使用较慢的速度
return 0.02 # 30ms延迟 return 0.02 # 30ms延迟
def send_char(char, message_id): def send_char(char, message_id):
"""Helper function to send single character""" """Helper function to send single character"""
openai_chunk = { openai_chunk = {
@ -499,10 +499,10 @@ def chat_completions():
} }
chunk_data = f"data: {json.dumps(openai_chunk)}\n\n" chunk_data = f"data: {json.dumps(openai_chunk)}\n\n"
return flush_chunk(chunk_data) return flush_chunk(chunk_data)
# 初始化缓冲区 # 初始化缓冲区
output_buffer = [] output_buffer = []
try: try:
with client.stream( with client.stream(
'POST', 'POST',
@ -517,74 +517,76 @@ def chat_completions():
) as response: ) as response:
generate.message_id = None generate.message_id = None
buffer = "" buffer = ""
for raw_bytes in response.iter_raw(): for raw_bytes in response.iter_raw():
if not raw_bytes: if not raw_bytes:
continue continue
try: try:
buffer += raw_bytes.decode('utf-8') buffer += raw_bytes.decode('utf-8')
while '\n' in buffer: while '\n' in buffer:
line, buffer = buffer.split('\n', 1) line, buffer = buffer.split('\n', 1)
line = line.strip() line = line.strip()
if not line or not line.startswith('data: '): if not line or not line.startswith('data: '):
continue continue
try: try:
json_str = line[6:] json_str = line[6:]
dify_chunk = json.loads(json_str) dify_chunk = json.loads(json_str)
if dify_chunk.get("event") == "message" and "answer" in dify_chunk: if dify_chunk.get("event") == "message" and "answer" in dify_chunk:
current_answer = dify_chunk["answer"] current_answer = dify_chunk["answer"]
if not current_answer: if not current_answer:
continue continue
message_id = dify_chunk.get("message_id", "") message_id = dify_chunk.get("message_id", "")
if not generate.message_id: if not generate.message_id:
generate.message_id = message_id generate.message_id = message_id
# 将当前批次的字符添加到输出缓冲区 # # 将当前批次的字符添加到输出缓冲区
for char in current_answer: # for char in current_answer:
output_buffer.append((char, generate.message_id)) # output_buffer.append((char, generate.message_id))
# 根据缓冲区大小动态调整输出速度 # # 根据缓冲区大小动态调整输出速度
while output_buffer: # while output_buffer:
char, msg_id = output_buffer.pop(0) # char, msg_id = output_buffer.pop(0)
yield send_char(char, msg_id) # yield send_char(char, msg_id)
# 根据剩余缓冲区大小计算延迟 # # 根据剩余缓冲区大小计算延迟
delay = calculate_delay(len(output_buffer)) # delay = calculate_delay(len(output_buffer))
time.sleep(delay) # time.sleep(delay)
yield send_char(current_answer, message_id)
# 立即继续处理下一个请求 # 立即继续处理下一个请求
continue continue
# 处理Agent模式的消息事件 # 处理Agent模式的消息事件
elif dify_chunk.get("event") == "agent_message" and "answer" in dify_chunk: elif dify_chunk.get("event") == "agent_message" and "answer" in dify_chunk:
current_answer = dify_chunk["answer"] current_answer = dify_chunk["answer"]
if not current_answer: if not current_answer:
continue continue
message_id = dify_chunk.get("message_id", "") message_id = dify_chunk.get("message_id", "")
if not generate.message_id: if not generate.message_id:
generate.message_id = message_id generate.message_id = message_id
# 将当前批次的字符添加到输出缓冲区 # # 将当前批次的字符添加到输出缓冲区
for char in current_answer: # for char in current_answer:
output_buffer.append((char, generate.message_id)) # output_buffer.append((char, generate.message_id))
# 根据缓冲区大小动态调整输出速度 # # 根据缓冲区大小动态调整输出速度
while output_buffer: # while output_buffer:
char, msg_id = output_buffer.pop(0) # char, msg_id = output_buffer.pop(0)
yield send_char(char, msg_id) # yield send_char(char, msg_id)
# 根据剩余缓冲区大小计算延迟 # # 根据剩余缓冲区大小计算延迟
delay = calculate_delay(len(output_buffer)) # delay = calculate_delay(len(output_buffer))
time.sleep(delay) # time.sleep(delay)
yield send_char(current_answer, message_id)
# 立即继续处理下一个请求 # 立即继续处理下一个请求
continue continue
# 处理Agent的思考过程，记录日志但不输出给用户 # 处理Agent的思考过程，记录日志但不输出给用户
elif dify_chunk.get("event") == "agent_thought": elif dify_chunk.get("event") == "agent_thought":
thought_id = dify_chunk.get("id", "") thought_id = dify_chunk.get("id", "")
@ -592,7 +594,7 @@ def chat_completions():
tool = dify_chunk.get("tool", "") tool = dify_chunk.get("tool", "")
tool_input = dify_chunk.get("tool_input", "") tool_input = dify_chunk.get("tool_input", "")
observation = dify_chunk.get("observation", "") observation = dify_chunk.get("observation", "")
logger.info(f"[Agent Thought] ID: {thought_id}, Tool: {tool}") logger.info(f"[Agent Thought] ID: {thought_id}, Tool: {tool}")
if thought: if thought:
logger.info(f"[Agent Thought] Thought: {thought}") logger.info(f"[Agent Thought] Thought: {thought}")
@ -600,35 +602,35 @@ def chat_completions():
logger.info(f"[Agent Thought] Tool Input: {tool_input}") logger.info(f"[Agent Thought] Tool Input: {tool_input}")
if observation: if observation:
logger.info(f"[Agent Thought] Observation: {observation}") logger.info(f"[Agent Thought] Observation: {observation}")
# 获取message_id以关联思考和最终输出 # 获取message_id以关联思考和最终输出
message_id = dify_chunk.get("message_id", "") message_id = dify_chunk.get("message_id", "")
if not generate.message_id and message_id: if not generate.message_id and message_id:
generate.message_id = message_id generate.message_id = message_id
continue continue
# 处理消息中的文件(如图片)，记录日志但不直接输出给用户 # 处理消息中的文件(如图片)，记录日志但不直接输出给用户
elif dify_chunk.get("event") == "message_file": elif dify_chunk.get("event") == "message_file":
file_id = dify_chunk.get("id", "") file_id = dify_chunk.get("id", "")
file_type = dify_chunk.get("type", "") file_type = dify_chunk.get("type", "")
file_url = dify_chunk.get("url", "") file_url = dify_chunk.get("url", "")
logger.info(f"[Message File] ID: {file_id}, Type: {file_type}, URL: {file_url}") logger.info(f"[Message File] ID: {file_id}, Type: {file_type}, URL: {file_url}")
continue continue
elif dify_chunk.get("event") == "message_end": elif dify_chunk.get("event") == "message_end":
# 快速输出剩余内容 # 快速输出剩余内容
while output_buffer: while output_buffer:
char, msg_id = output_buffer.pop(0) char, msg_id = output_buffer.pop(0)
yield send_char(char, msg_id) yield send_char(char, msg_id)
time.sleep(0.001) # 固定使用最小延迟快速输出剩余内容 time.sleep(0.001) # 固定使用最小延迟快速输出剩余内容
# 只在零宽字符会话记忆模式时处理conversation_id # 只在零宽字符会话记忆模式时处理conversation_id
if CONVERSATION_MEMORY_MODE == 2: if CONVERSATION_MEMORY_MODE == 2:
conversation_id = dify_chunk.get("conversation_id") conversation_id = dify_chunk.get("conversation_id")
history = dify_chunk.get("conversation_history", []) history = dify_chunk.get("conversation_history", [])
has_conversation_id = False has_conversation_id = False
if history: if history:
for msg in history: for msg in history:
@ -637,7 +639,7 @@ def chat_completions():
if decode_conversation_id(content) is not None: if decode_conversation_id(content) is not None:
has_conversation_id = True has_conversation_id = True
break break
# 只在新会话且历史消息中没有会话ID时插入 # 只在新会话且历史消息中没有会话ID时插入
if conversation_id and not has_conversation_id: if conversation_id and not has_conversation_id:
logger.info(f"[Debug] Inserting conversation_id in stream: {conversation_id}") logger.info(f"[Debug] Inserting conversation_id in stream: {conversation_id}")
@ -645,7 +647,7 @@ def chat_completions():
logger.info(f"[Debug] Stream encoded content: {repr(encoded)}") logger.info(f"[Debug] Stream encoded content: {repr(encoded)}")
for char in encoded: for char in encoded:
yield send_char(char, generate.message_id) yield send_char(char, generate.message_id)
final_chunk = { final_chunk = {
"id": generate.message_id, "id": generate.message_id,
"object": "chat.completion.chunk", "object": "chat.completion.chunk",
@ -659,11 +661,11 @@ def chat_completions():
} }
yield flush_chunk(f"data: {json.dumps(final_chunk)}\n\n") yield flush_chunk(f"data: {json.dumps(final_chunk)}\n\n")
yield flush_chunk("data: [DONE]\n\n") yield flush_chunk("data: [DONE]\n\n")
except json.JSONDecodeError as e: except json.JSONDecodeError as e:
logger.error(f"JSON decode error: {str(e)}") logger.error(f"JSON decode error: {str(e)}")
continue continue
except Exception as e: except Exception as e:
logger.error(f"Error processing chunk: {str(e)}") logger.error(f"Error processing chunk: {str(e)}")
continue continue
@ -692,7 +694,7 @@ def chat_completions():
json=dify_request, json=dify_request,
headers=headers headers=headers
) )
if response.status_code != 200: if response.status_code != 200:
error_msg = f"Dify API error: {response.text}" error_msg = f"Dify API error: {response.text}"
logger.error(f"Request failed: {error_msg}") logger.error(f"Request failed: {error_msg}")
@ -746,13 +748,13 @@ def chat_completions():
def list_models(): def list_models():
"""返回可用的模型列表""" """返回可用的模型列表"""
logger.info("Listing available models") logger.info("Listing available models")
# 刷新模型信息 # 刷新模型信息
asyncio.run(model_manager.refresh_model_info()) asyncio.run(model_manager.refresh_model_info())
# 获取可用模型列表 # 获取可用模型列表
available_models = model_manager.get_available_models() available_models = model_manager.get_available_models()
response = { response = {
"object": "list", "object": "list",
"data": available_models "data": available_models
@ -764,10 +766,10 @@ def list_models():
if __name__ == '__main__': if __name__ == '__main__':
if not VALID_API_KEYS: if not VALID_API_KEYS:
print("Warning: No API keys configured. Set the VALID_API_KEYS environment variable with comma-separated keys.") print("Warning: No API keys configured. Set the VALID_API_KEYS environment variable with comma-separated keys.")
# 启动时初始化模型信息 # 启动时初始化模型信息
asyncio.run(model_manager.refresh_model_info()) asyncio.run(model_manager.refresh_model_info())
host = os.getenv("SERVER_HOST", "127.0.0.1") host = os.getenv("SERVER_HOST", "127.0.0.1")
port = int(os.getenv("SERVER_PORT", 5000)) port = int(os.getenv("SERVER_PORT", 5000))
logger.info(f"Starting server on http://{host}:{port}") logger.info(f"Starting server on http://{host}:{port}")