feat: server multi models support (#799)

api/core/model_providers/models/entity/message.py

@@ -0,0 +1,53 @@
+import enum
+
+from langchain.schema import HumanMessage, AIMessage, SystemMessage, BaseMessage
+from pydantic import BaseModel
+
+
+class LLMRunResult(BaseModel):
+    content: str
+    prompt_tokens: int
+    completion_tokens: int
+
+
+class MessageType(enum.Enum):
+    HUMAN = 'human'
+    ASSISTANT = 'assistant'
+    SYSTEM = 'system'
+
+
+class PromptMessage(BaseModel):
+    type: MessageType = MessageType.HUMAN
+    content: str = ''
+
+
+def to_lc_messages(messages: list[PromptMessage]):
+    lc_messages = []
+    for message in messages:
+        if message.type == MessageType.HUMAN:
+            lc_messages.append(HumanMessage(content=message.content))
+        elif message.type == MessageType.ASSISTANT:
+            lc_messages.append(AIMessage(content=message.content))
+        elif message.type == MessageType.SYSTEM:
+            lc_messages.append(SystemMessage(content=message.content))
+
+    return lc_messages
+
+
+def to_prompt_messages(messages: list[BaseMessage]):
+    prompt_messages = []
+    for message in messages:
+        if isinstance(message, HumanMessage):
+            prompt_messages.append(PromptMessage(content=message.content, type=MessageType.HUMAN))
+        elif isinstance(message, AIMessage):
+            prompt_messages.append(PromptMessage(content=message.content, type=MessageType.ASSISTANT))
+        elif isinstance(message, SystemMessage):
+            prompt_messages.append(PromptMessage(content=message.content, type=MessageType.SYSTEM))
+    return prompt_messages
+
+
+def str_to_prompt_messages(texts: list[str]):
+    prompt_messages = []
+    for text in texts:
+        prompt_messages.append(PromptMessage(content=text))
+    return prompt_messages

api/core/model_providers/models/entity/model_params.py

@@ -0,0 +1,59 @@
+import enum
+from typing import Optional, TypeVar, Generic
+
+from langchain.load.serializable import Serializable
+from pydantic import BaseModel
+
+
+class ModelMode(enum.Enum):
+    COMPLETION = 'completion'
+    CHAT = 'chat'
+
+
+class ModelType(enum.Enum):
+    TEXT_GENERATION = 'text-generation'
+    EMBEDDINGS = 'embeddings'
+    SPEECH_TO_TEXT = 'speech2text'
+    IMAGE = 'image'
+    VIDEO = 'video'
+    MODERATION = 'moderation'
+
+    @staticmethod
+    def value_of(value):
+        for member in ModelType:
+            if member.value == value:
+                return member
+        raise ValueError(f"No matching enum found for value '{value}'")
+
+
+class ModelKwargs(BaseModel):
+    max_tokens: Optional[int]
+    temperature: Optional[float]
+    top_p: Optional[float]
+    presence_penalty: Optional[float]
+    frequency_penalty: Optional[float]
+
+
+class KwargRuleType(enum.Enum):
+    STRING = 'string'
+    INTEGER = 'integer'
+    FLOAT = 'float'
+
+
+T = TypeVar('T')
+
+
+class KwargRule(Generic[T], BaseModel):
+    enabled: bool = True
+    min: Optional[T] = None
+    max: Optional[T] = None
+    default: Optional[T] = None
+    alias: Optional[str] = None
+
+
+class ModelKwargsRules(BaseModel):
+    max_tokens: KwargRule = KwargRule[int](enabled=False)
+    temperature: KwargRule = KwargRule[float](enabled=False)
+    top_p: KwargRule = KwargRule[float](enabled=False)
+    presence_penalty: KwargRule = KwargRule[float](enabled=False)
+    frequency_penalty: KwargRule = KwargRule[float](enabled=False)

api/core/model_providers/models/entity/provider.py

@@ -0,0 +1,10 @@
+from enum import Enum
+
+
+class ProviderQuotaUnit(Enum):
+    TIMES = 'times'
+    TOKENS = 'tokens'
+
+
+class ModelFeature(Enum):
+    AGENT_THOUGHT = 'agent_thought'