admin/dify
1
0
Fork 0
Code Issues Pull Requests Actions 1 Packages Projects Releases Wiki Activity

feat: advanced prompt backend (#1301)

Browse Source 
Co-authored-by: takatost <takatost@gmail.com>
This commit is contained in:
Garfield Dai
2023-10-12 23:13:10 +08:00
committed by GitHub
parent 2d1cb076c6
commit 42a5b3ec17
61 changed files with 767 additions and 581 deletions
Download Patch File Download Diff File Expand all files Collapse all files

10
api/core/model_providers/models/entity/message.py
View File

@@ -13,13 +13,13 @@ class LLMRunResult(BaseModel):
class MessageType(enum.Enum):
HUMAN = 'human'
USER = 'user'
ASSISTANT = 'assistant'
SYSTEM = 'system'
class PromptMessage(BaseModel):
type: MessageType = MessageType.HUMAN
type: MessageType = MessageType.USER
content: str = ''
function_call: dict = None
@@ -27,7 +27,7 @@ class PromptMessage(BaseModel):
def to_lc_messages(messages: list[PromptMessage]):
lc_messages = []
for message in messages:
if message.type == MessageType.HUMAN:
if message.type == MessageType.USER:
lc_messages.append(HumanMessage(content=message.content))
elif message.type == MessageType.ASSISTANT:
additional_kwargs = {}
@@ -44,7 +44,7 @@ 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))
prompt_messages.append(PromptMessage(content=message.content, type=MessageType.USER))
elif isinstance(message, AIMessage):
message_kwargs = {
'content': message.content,
@@ -58,7 +58,7 @@ def to_prompt_messages(messages: list[BaseMessage]):
elif isinstance(message, SystemMessage):
prompt_messages.append(PromptMessage(content=message.content, type=MessageType.SYSTEM))
elif isinstance(message, FunctionMessage):
prompt_messages.append(PromptMessage(content=message.content, type=MessageType.HUMAN))
prompt_messages.append(PromptMessage(content=message.content, type=MessageType.USER))
return prompt_messages

119
api/core/model_providers/models/llm/base.py
View File

@@ -18,7 +18,7 @@ from core.model_providers.models.entity.message import PromptMessage, MessageTyp
from core.model_providers.models.entity.model_params import ModelType, ModelKwargs, ModelMode, ModelKwargsRules
from core.model_providers.providers.base import BaseModelProvider
from core.prompt.prompt_builder import PromptBuilder
from core.prompt.prompt_template import JinjaPromptTemplate
from core.prompt.prompt_template import PromptTemplateParser
from core.third_party.langchain.llms.fake import FakeLLM
import logging
@@ -232,7 +232,7 @@ class BaseLLM(BaseProviderModel):
:param message_type:
:return:
"""
if message_type == MessageType.HUMAN or message_type == MessageType.SYSTEM:
if message_type == MessageType.USER or message_type == MessageType.SYSTEM:
unit_price = self.price_config['prompt']
else:
unit_price = self.price_config['completion']
@@ -250,7 +250,7 @@ class BaseLLM(BaseProviderModel):
:param message_type:
:return: decimal.Decimal('0.0001')
"""
if message_type == MessageType.HUMAN or message_type == MessageType.SYSTEM:
if message_type == MessageType.USER or message_type == MessageType.SYSTEM:
unit_price = self.price_config['prompt']
else:
unit_price = self.price_config['completion']
@@ -265,7 +265,7 @@ class BaseLLM(BaseProviderModel):
:param message_type:
:return: decimal.Decimal('0.000001')
"""
if message_type == MessageType.HUMAN or message_type == MessageType.SYSTEM:
if message_type == MessageType.USER or message_type == MessageType.SYSTEM:
price_unit = self.price_config['unit']
else:
price_unit = self.price_config['unit']
@@ -330,6 +330,85 @@ class BaseLLM(BaseProviderModel):
prompt, stops = self._get_prompt_and_stop(prompt_rules, pre_prompt, inputs, query, context, memory)
return [PromptMessage(content=prompt)], stops
def get_advanced_prompt(self, app_mode: str,
app_model_config: str, inputs: dict,
query: str,
context: Optional[str],
memory: Optional[BaseChatMemory]) -> List[PromptMessage]:
model_mode = app_model_config.model_dict['mode']
conversation_histories_role = {}
raw_prompt_list = []
prompt_messages = []
if app_mode == 'chat' and model_mode == ModelMode.COMPLETION.value:
prompt_text = app_model_config.completion_prompt_config_dict['prompt']['text']
raw_prompt_list = [{
'role': MessageType.USER.value,
'text': prompt_text
}]
conversation_histories_role = app_model_config.completion_prompt_config_dict['conversation_histories_role']
elif app_mode == 'chat' and model_mode == ModelMode.CHAT.value:
raw_prompt_list = app_model_config.chat_prompt_config_dict['prompt']
elif app_mode == 'completion' and model_mode == ModelMode.CHAT.value:
raw_prompt_list = app_model_config.chat_prompt_config_dict['prompt']
elif app_mode == 'completion' and model_mode == ModelMode.COMPLETION.value:
prompt_text = app_model_config.completion_prompt_config_dict['prompt']['text']
raw_prompt_list = [{
'role': MessageType.USER.value,
'text': prompt_text
}]
else:
raise Exception("app_mode or model_mode not support")
for prompt_item in raw_prompt_list:
prompt = prompt_item['text']
# set prompt template variables
prompt_template = PromptTemplateParser(template=prompt)
prompt_inputs = {k: inputs[k] for k in prompt_template.variable_keys if k in inputs}
if '#context#' in prompt:
if context:
prompt_inputs['#context#'] = context
else:
prompt_inputs['#context#'] = ''
if '#query#' in prompt:
if query:
prompt_inputs['#query#'] = query
else:
prompt_inputs['#query#'] = ''
if '#histories#' in prompt:
if memory and app_mode == 'chat' and model_mode == ModelMode.COMPLETION.value:
memory.human_prefix = conversation_histories_role['user_prefix']
memory.ai_prefix = conversation_histories_role['assistant_prefix']
histories = self._get_history_messages_from_memory(memory, 2000)
prompt_inputs['#histories#'] = histories
else:
prompt_inputs['#histories#'] = ''
prompt = prompt_template.format(
prompt_inputs
)
prompt = re.sub(r'<\|.*?\|>', '', prompt)
prompt_messages.append(PromptMessage(type = MessageType(prompt_item['role']) ,content=prompt))
if memory and app_mode == 'chat' and model_mode == ModelMode.CHAT.value:
memory.human_prefix = MessageType.USER.value
memory.ai_prefix = MessageType.ASSISTANT.value
histories = self._get_history_messages_list_from_memory(memory, 2000)
prompt_messages.extend(histories)
if app_mode == 'chat' and model_mode == ModelMode.CHAT.value:
prompt_messages.append(PromptMessage(type = MessageType.USER ,content=query))
return prompt_messages
def prompt_file_name(self, mode: str) -> str:
if mode == 'completion':
return 'common_completion'
@@ -342,17 +421,17 @@ class BaseLLM(BaseProviderModel):
memory: Optional[BaseChatMemory]) -> Tuple[str, Optional[list]]:
context_prompt_content = ''
if context and 'context_prompt' in prompt_rules:
prompt_template = JinjaPromptTemplate.from_template(template=prompt_rules['context_prompt'])
prompt_template = PromptTemplateParser(template=prompt_rules['context_prompt'])
context_prompt_content = prompt_template.format(
context=context
{'context': context}
)
pre_prompt_content = ''
if pre_prompt:
prompt_template = JinjaPromptTemplate.from_template(template=pre_prompt)
prompt_inputs = {k: inputs[k] for k in prompt_template.input_variables if k in inputs}
prompt_template = PromptTemplateParser(template=pre_prompt)
prompt_inputs = {k: inputs[k] for k in prompt_template.variable_keys if k in inputs}
pre_prompt_content = prompt_template.format(
**prompt_inputs
prompt_inputs
)
prompt = ''
@@ -385,10 +464,8 @@ class BaseLLM(BaseProviderModel):
memory.ai_prefix = prompt_rules['assistant_prefix'] if 'assistant_prefix' in prompt_rules else 'Assistant'
histories = self._get_history_messages_from_memory(memory, rest_tokens)
prompt_template = JinjaPromptTemplate.from_template(template=prompt_rules['histories_prompt'])
histories_prompt_content = prompt_template.format(
histories=histories
)
prompt_template = PromptTemplateParser(template=prompt_rules['histories_prompt'])
histories_prompt_content = prompt_template.format({'histories': histories})
prompt = ''
for order in prompt_rules['system_prompt_orders']:
@@ -399,10 +476,8 @@ class BaseLLM(BaseProviderModel):
elif order == 'histories_prompt':
prompt += histories_prompt_content
prompt_template = JinjaPromptTemplate.from_template(template=query_prompt)
query_prompt_content = prompt_template.format(
query=query
)
prompt_template = PromptTemplateParser(template=query_prompt)
query_prompt_content = prompt_template.format({'query': query})
prompt += query_prompt_content
@@ -433,6 +508,16 @@ class BaseLLM(BaseProviderModel):
external_context = memory.load_memory_variables({})
return external_context[memory_key]
def _get_history_messages_list_from_memory(self, memory: BaseChatMemory,
max_token_limit: int) -> List[PromptMessage]:
"""Get memory messages."""
memory.max_token_limit = max_token_limit
memory.return_messages = True
memory_key = memory.memory_variables[0]
external_context = memory.load_memory_variables({})
memory.return_messages = False
return to_prompt_messages(external_context[memory_key])
def _get_prompt_from_messages(self, messages: List[PromptMessage],
model_mode: Optional[ModelMode] = None) -> Union[str | List[BaseMessage]]:
if not model_mode:

7
api/core/model_providers/providers/anthropic_provider.py
View File

@@ -9,7 +9,7 @@ from langchain.schema import HumanMessage
from core.helper import encrypter
from core.model_providers.models.base import BaseProviderModel
from core.model_providers.models.entity.model_params import ModelKwargsRules, KwargRule
from core.model_providers.models.entity.model_params import ModelKwargsRules, KwargRule, ModelMode
from core.model_providers.models.entity.provider import ModelFeature
from core.model_providers.models.llm.anthropic_model import AnthropicModel
from core.model_providers.models.llm.base import ModelType
@@ -34,10 +34,12 @@ class AnthropicProvider(BaseModelProvider):
{
'id': 'claude-instant-1',
'name': 'claude-instant-1',
'mode': ModelMode.CHAT.value,
},
{
'id': 'claude-2',
'name': 'claude-2',
'mode': ModelMode.CHAT.value,
'features': [
ModelFeature.AGENT_THOUGHT.value
]
@@ -46,6 +48,9 @@ class AnthropicProvider(BaseModelProvider):
else:
return []
def _get_text_generation_model_mode(self, model_name) -> str:
return ModelMode.CHAT.value
def get_model_class(self, model_type: ModelType) -> Type[BaseProviderModel]:
"""
Returns the model class.

17
api/core/model_providers/providers/azure_openai_provider.py
View File

@@ -12,7 +12,7 @@ from core.helper import encrypter
from core.model_providers.models.base import BaseProviderModel
from core.model_providers.models.embedding.azure_openai_embedding import AzureOpenAIEmbedding, \
AZURE_OPENAI_API_VERSION
from core.model_providers.models.entity.model_params import ModelType, ModelKwargsRules, KwargRule
from core.model_providers.models.entity.model_params import ModelType, ModelKwargsRules, KwargRule, ModelMode
from core.model_providers.models.entity.provider import ModelFeature
from core.model_providers.models.llm.azure_openai_model import AzureOpenAIModel
from core.model_providers.providers.base import BaseModelProvider, CredentialsValidateFailedError
@@ -61,6 +61,10 @@ class AzureOpenAIProvider(BaseModelProvider):
}
credentials = json.loads(provider_model.encrypted_config)
if provider_model.model_type == ModelType.TEXT_GENERATION.value:
model_dict['mode'] = self._get_text_generation_model_mode(credentials['base_model_name'])
if credentials['base_model_name'] in [
'gpt-4',
'gpt-4-32k',
@@ -77,12 +81,19 @@ class AzureOpenAIProvider(BaseModelProvider):
return model_list
def _get_text_generation_model_mode(self, model_name) -> str:
if model_name == 'text-davinci-003':
return ModelMode.COMPLETION.value
else:
return ModelMode.CHAT.value
def _get_fixed_model_list(self, model_type: ModelType) -> list[dict]:
if model_type == ModelType.TEXT_GENERATION:
models = [
{
'id': 'gpt-3.5-turbo',
'name': 'gpt-3.5-turbo',
'mode': ModelMode.CHAT.value,
'features': [
ModelFeature.AGENT_THOUGHT.value
]
@@ -90,6 +101,7 @@ class AzureOpenAIProvider(BaseModelProvider):
{
'id': 'gpt-3.5-turbo-16k',
'name': 'gpt-3.5-turbo-16k',
'mode': ModelMode.CHAT.value,
'features': [
ModelFeature.AGENT_THOUGHT.value
]
@@ -97,6 +109,7 @@ class AzureOpenAIProvider(BaseModelProvider):
{
'id': 'gpt-4',
'name': 'gpt-4',
'mode': ModelMode.CHAT.value,
'features': [
ModelFeature.AGENT_THOUGHT.value
]
@@ -104,6 +117,7 @@ class AzureOpenAIProvider(BaseModelProvider):
{
'id': 'gpt-4-32k',
'name': 'gpt-4-32k',
'mode': ModelMode.CHAT.value,
'features': [
ModelFeature.AGENT_THOUGHT.value
]
@@ -111,6 +125,7 @@ class AzureOpenAIProvider(BaseModelProvider):
{
'id': 'text-davinci-003',
'name': 'text-davinci-003',
'mode': ModelMode.COMPLETION.value,
}
]

6
api/core/model_providers/providers/baichuan_provider.py
View File

@@ -6,7 +6,7 @@ from langchain.schema import HumanMessage
from core.helper import encrypter
from core.model_providers.models.base import BaseProviderModel
from core.model_providers.models.entity.model_params import ModelKwargsRules, KwargRule, ModelType
from core.model_providers.models.entity.model_params import ModelKwargsRules, KwargRule, ModelType, ModelMode
from core.model_providers.models.llm.baichuan_model import BaichuanModel
from core.model_providers.providers.base import BaseModelProvider, CredentialsValidateFailedError
from core.third_party.langchain.llms.baichuan_llm import BaichuanChatLLM
@@ -21,6 +21,9 @@ class BaichuanProvider(BaseModelProvider):
Returns the name of a provider.
"""
return 'baichuan'
def _get_text_generation_model_mode(self, model_name) -> str:
return ModelMode.CHAT.value
def _get_fixed_model_list(self, model_type: ModelType) -> list[dict]:
if model_type == ModelType.TEXT_GENERATION:
@@ -28,6 +31,7 @@ class BaichuanProvider(BaseModelProvider):
{
'id': 'baichuan2-53b',
'name': 'Baichuan2-53B',
'mode': ModelMode.CHAT.value,
}
]
else:

27
api/core/model_providers/providers/base.py
View File

@@ -61,10 +61,19 @@ class BaseModelProvider(BaseModel, ABC):
ProviderModel.is_valid == True
).order_by(ProviderModel.created_at.asc()).all()
return [{
'id': provider_model.model_name,
'name': provider_model.model_name
} for provider_model in provider_models]
provider_model_list = []
for provider_model in provider_models:
provider_model_dict = {
'id': provider_model.model_name,
'name': provider_model.model_name
}
if model_type == ModelType.TEXT_GENERATION:
provider_model_dict['mode'] = self._get_text_generation_model_mode(provider_model.model_name)
provider_model_list.append(provider_model_dict)
return provider_model_list
@abstractmethod
def _get_fixed_model_list(self, model_type: ModelType) -> list[dict]:
@@ -76,6 +85,16 @@ class BaseModelProvider(BaseModel, ABC):
"""
raise NotImplementedError
@abstractmethod
def _get_text_generation_model_mode(self, model_name) -> str:
"""
get text generation model mode.
:param model_name:
:return:
"""
raise NotImplementedError
@abstractmethod
def get_model_class(self, model_type: ModelType) -> Type:
"""

7
api/core/model_providers/providers/chatglm_provider.py
View File

@@ -6,7 +6,7 @@ from langchain.llms import ChatGLM
from core.helper import encrypter
from core.model_providers.models.base import BaseProviderModel
from core.model_providers.models.entity.model_params import ModelKwargsRules, KwargRule, ModelType
from core.model_providers.models.entity.model_params import ModelKwargsRules, KwargRule, ModelType, ModelMode
from core.model_providers.models.llm.chatglm_model import ChatGLMModel
from core.model_providers.providers.base import BaseModelProvider, CredentialsValidateFailedError
from models.provider import ProviderType
@@ -27,15 +27,20 @@ class ChatGLMProvider(BaseModelProvider):
{
'id': 'chatglm2-6b',
'name': 'ChatGLM2-6B',
'mode': ModelMode.COMPLETION.value,
},
{
'id': 'chatglm-6b',
'name': 'ChatGLM-6B',
'mode': ModelMode.COMPLETION.value,
}
]
else:
return []
def _get_text_generation_model_mode(self, model_name) -> str:
return ModelMode.COMPLETION.value
def get_model_class(self, model_type: ModelType) -> Type[BaseProviderModel]:
"""
Returns the model class.

5
api/core/model_providers/providers/huggingface_hub_provider.py
View File

@@ -5,7 +5,7 @@ import requests
from huggingface_hub import HfApi
from core.helper import encrypter
from core.model_providers.models.entity.model_params import KwargRule, ModelKwargsRules, ModelType
from core.model_providers.models.entity.model_params import KwargRule, ModelKwargsRules, ModelType, ModelMode
from core.model_providers.models.llm.huggingface_hub_model import HuggingfaceHubModel
from core.model_providers.providers.base import BaseModelProvider, CredentialsValidateFailedError
@@ -29,6 +29,9 @@ class HuggingfaceHubProvider(BaseModelProvider):
def _get_fixed_model_list(self, model_type: ModelType) -> list[dict]:
return []
def _get_text_generation_model_mode(self, model_name) -> str:
return ModelMode.COMPLETION.value
def get_model_class(self, model_type: ModelType) -> Type[BaseProviderModel]:
"""
Returns the model class.

9
api/core/model_providers/providers/localai_provider.py
View File

@@ -6,7 +6,7 @@ from langchain.schema import HumanMessage
from core.helper import encrypter
from core.model_providers.models.embedding.localai_embedding import LocalAIEmbedding
from core.model_providers.models.entity.model_params import ModelKwargsRules, ModelType, KwargRule
from core.model_providers.models.entity.model_params import ModelKwargsRules, ModelType, KwargRule, ModelMode
from core.model_providers.models.llm.localai_model import LocalAIModel
from core.model_providers.providers.base import BaseModelProvider, CredentialsValidateFailedError
@@ -27,6 +27,13 @@ class LocalAIProvider(BaseModelProvider):
def _get_fixed_model_list(self, model_type: ModelType) -> list[dict]:
return []
def _get_text_generation_model_mode(self, model_name) -> str:
credentials = self.get_model_credentials(model_name, ModelType.TEXT_GENERATION)
if credentials['completion_type'] == 'chat_completion':
return ModelMode.CHAT.value
else:
return ModelMode.COMPLETION.value
def get_model_class(self, model_type: ModelType) -> Type[BaseProviderModel]:
"""
Returns the model class.

7
api/core/model_providers/providers/minimax_provider.py
View File

@@ -7,7 +7,7 @@ from langchain.schema import HumanMessage
from core.helper import encrypter
from core.model_providers.models.base import BaseProviderModel
from core.model_providers.models.embedding.minimax_embedding import MinimaxEmbedding
from core.model_providers.models.entity.model_params import ModelKwargsRules, KwargRule, ModelType
from core.model_providers.models.entity.model_params import ModelKwargsRules, KwargRule, ModelType, ModelMode
from core.model_providers.models.llm.minimax_model import MinimaxModel
from core.model_providers.providers.base import BaseModelProvider, CredentialsValidateFailedError
from core.third_party.langchain.llms.minimax_llm import MinimaxChatLLM
@@ -29,10 +29,12 @@ class MinimaxProvider(BaseModelProvider):
{
'id': 'abab5.5-chat',
'name': 'abab5.5-chat',
'mode': ModelMode.COMPLETION.value,
},
{
'id': 'abab5-chat',
'name': 'abab5-chat',
'mode': ModelMode.COMPLETION.value,
}
]
elif model_type == ModelType.EMBEDDINGS:
@@ -45,6 +47,9 @@ class MinimaxProvider(BaseModelProvider):
else:
return []
def _get_text_generation_model_mode(self, model_name) -> str:
return ModelMode.COMPLETION.value
def get_model_class(self, model_type: ModelType) -> Type[BaseProviderModel]:
"""
Returns the model class.

16
api/core/model_providers/providers/openai_provider.py
View File

@@ -13,8 +13,8 @@ from core.model_providers.models.entity.provider import ModelFeature
from core.model_providers.models.speech2text.openai_whisper import OpenAIWhisper
from core.model_providers.models.base import BaseProviderModel
from core.model_providers.models.embedding.openai_embedding import OpenAIEmbedding
from core.model_providers.models.entity.model_params import ModelKwargsRules, KwargRule, ModelType
from core.model_providers.models.llm.openai_model import OpenAIModel
from core.model_providers.models.entity.model_params import ModelKwargsRules, KwargRule, ModelType, ModelMode
from core.model_providers.models.llm.openai_model import OpenAIModel, COMPLETION_MODELS
from core.model_providers.models.moderation.openai_moderation import OpenAIModeration
from core.model_providers.providers.base import BaseModelProvider, CredentialsValidateFailedError
from core.model_providers.providers.hosted import hosted_model_providers
@@ -36,6 +36,7 @@ class OpenAIProvider(BaseModelProvider):
{
'id': 'gpt-3.5-turbo',
'name': 'gpt-3.5-turbo',
'mode': ModelMode.CHAT.value,
'features': [
ModelFeature.AGENT_THOUGHT.value
]
@@ -43,10 +44,12 @@ class OpenAIProvider(BaseModelProvider):
{
'id': 'gpt-3.5-turbo-instruct',
'name': 'GPT-3.5-Turbo-Instruct',
'mode': ModelMode.COMPLETION.value,
},
{
'id': 'gpt-3.5-turbo-16k',
'name': 'gpt-3.5-turbo-16k',
'mode': ModelMode.CHAT.value,
'features': [
ModelFeature.AGENT_THOUGHT.value
]
@@ -54,6 +57,7 @@ class OpenAIProvider(BaseModelProvider):
{
'id': 'gpt-4',
'name': 'gpt-4',
'mode': ModelMode.CHAT.value,
'features': [
ModelFeature.AGENT_THOUGHT.value
]
@@ -61,6 +65,7 @@ class OpenAIProvider(BaseModelProvider):
{
'id': 'gpt-4-32k',
'name': 'gpt-4-32k',
'mode': ModelMode.CHAT.value,
'features': [
ModelFeature.AGENT_THOUGHT.value
]
@@ -68,6 +73,7 @@ class OpenAIProvider(BaseModelProvider):
{
'id': 'text-davinci-003',
'name': 'text-davinci-003',
'mode': ModelMode.COMPLETION.value,
}
]
@@ -100,6 +106,12 @@ class OpenAIProvider(BaseModelProvider):
else:
return []
def _get_text_generation_model_mode(self, model_name) -> str:
if model_name in COMPLETION_MODELS:
return ModelMode.COMPLETION.value
else:
return ModelMode.CHAT.value
def get_model_class(self, model_type: ModelType) -> Type[BaseProviderModel]:
"""
Returns the model class.

5
api/core/model_providers/providers/openllm_provider.py
View File

@@ -3,7 +3,7 @@ from typing import Type
from core.helper import encrypter
from core.model_providers.models.embedding.openllm_embedding import OpenLLMEmbedding
from core.model_providers.models.entity.model_params import KwargRule, ModelKwargsRules, ModelType
from core.model_providers.models.entity.model_params import KwargRule, ModelKwargsRules, ModelType, ModelMode
from core.model_providers.models.llm.openllm_model import OpenLLMModel
from core.model_providers.providers.base import BaseModelProvider, CredentialsValidateFailedError
@@ -24,6 +24,9 @@ class OpenLLMProvider(BaseModelProvider):
def _get_fixed_model_list(self, model_type: ModelType) -> list[dict]:
return []
def _get_text_generation_model_mode(self, model_name) -> str:
return ModelMode.COMPLETION.value
def get_model_class(self, model_type: ModelType) -> Type[BaseProviderModel]:
"""
Returns the model class.

6
api/core/model_providers/providers/replicate_provider.py
View File

@@ -6,7 +6,8 @@ import replicate
from replicate.exceptions import ReplicateError
from core.helper import encrypter
from core.model_providers.models.entity.model_params import KwargRule, KwargRuleType, ModelKwargsRules, ModelType
from core.model_providers.models.entity.model_params import KwargRule, KwargRuleType, ModelKwargsRules, ModelType, \
ModelMode
from core.model_providers.models.llm.replicate_model import ReplicateModel
from core.model_providers.providers.base import BaseModelProvider, CredentialsValidateFailedError
@@ -26,6 +27,9 @@ class ReplicateProvider(BaseModelProvider):
def _get_fixed_model_list(self, model_type: ModelType) -> list[dict]:
return []
def _get_text_generation_model_mode(self, model_name) -> str:
return ModelMode.CHAT.value if model_name.endswith('-chat') else ModelMode.COMPLETION.value
def get_model_class(self, model_type: ModelType) -> Type[BaseProviderModel]:
"""
Returns the model class.

7
api/core/model_providers/providers/spark_provider.py
View File

@@ -7,7 +7,7 @@ from langchain.schema import HumanMessage
from core.helper import encrypter
from core.model_providers.models.base import BaseProviderModel
from core.model_providers.models.entity.model_params import ModelKwargsRules, KwargRule, ModelType
from core.model_providers.models.entity.model_params import ModelKwargsRules, KwargRule, ModelType, ModelMode
from core.model_providers.models.llm.spark_model import SparkModel
from core.model_providers.providers.base import BaseModelProvider, CredentialsValidateFailedError
from core.third_party.langchain.llms.spark import ChatSpark
@@ -30,15 +30,20 @@ class SparkProvider(BaseModelProvider):
{
'id': 'spark',
'name': 'Spark V1.5',
'mode': ModelMode.CHAT.value,
},
{
'id': 'spark-v2',
'name': 'Spark V2.0',
'mode': ModelMode.CHAT.value,
}
]
else:
return []
def _get_text_generation_model_mode(self, model_name) -> str:
return ModelMode.CHAT.value
def get_model_class(self, model_type: ModelType) -> Type[BaseProviderModel]:
"""
Returns the model class.

7
api/core/model_providers/providers/tongyi_provider.py
View File

@@ -4,7 +4,7 @@ from typing import Type
from core.helper import encrypter
from core.model_providers.models.base import BaseProviderModel
from core.model_providers.models.entity.model_params import ModelKwargsRules, KwargRule, ModelType
from core.model_providers.models.entity.model_params import ModelKwargsRules, KwargRule, ModelType, ModelMode
from core.model_providers.models.llm.tongyi_model import TongyiModel
from core.model_providers.providers.base import BaseModelProvider, CredentialsValidateFailedError
from core.third_party.langchain.llms.tongyi_llm import EnhanceTongyi
@@ -26,15 +26,20 @@ class TongyiProvider(BaseModelProvider):
{
'id': 'qwen-turbo',
'name': 'qwen-turbo',
'mode': ModelMode.COMPLETION.value,
},
{
'id': 'qwen-plus',
'name': 'qwen-plus',
'mode': ModelMode.COMPLETION.value,
}
]
else:
return []
def _get_text_generation_model_mode(self, model_name) -> str:
return ModelMode.COMPLETION.value
def get_model_class(self, model_type: ModelType) -> Type[BaseProviderModel]:
"""
Returns the model class.

8
api/core/model_providers/providers/wenxin_provider.py
View File

@@ -4,7 +4,7 @@ from typing import Type
from core.helper import encrypter
from core.model_providers.models.base import BaseProviderModel
from core.model_providers.models.entity.model_params import ModelKwargsRules, KwargRule, ModelType
from core.model_providers.models.entity.model_params import ModelKwargsRules, KwargRule, ModelType, ModelMode
from core.model_providers.models.llm.wenxin_model import WenxinModel
from core.model_providers.providers.base import BaseModelProvider, CredentialsValidateFailedError
from core.third_party.langchain.llms.wenxin import Wenxin
@@ -26,19 +26,25 @@ class WenxinProvider(BaseModelProvider):
{
'id': 'ernie-bot',
'name': 'ERNIE-Bot',
'mode': ModelMode.COMPLETION.value,
},
{
'id': 'ernie-bot-turbo',
'name': 'ERNIE-Bot-turbo',
'mode': ModelMode.COMPLETION.value,
},
{
'id': 'bloomz-7b',
'name': 'BLOOMZ-7B',
'mode': ModelMode.COMPLETION.value,
}
]
else:
return []
def _get_text_generation_model_mode(self, model_name) -> str:
return ModelMode.COMPLETION.value
def get_model_class(self, model_type: ModelType) -> Type[BaseProviderModel]:
"""
Returns the model class.

5
api/core/model_providers/providers/xinference_provider.py
View File

@@ -6,7 +6,7 @@ from langchain.embeddings import XinferenceEmbeddings
from core.helper import encrypter
from core.model_providers.models.embedding.xinference_embedding import XinferenceEmbedding
from core.model_providers.models.entity.model_params import KwargRule, ModelKwargsRules, ModelType
from core.model_providers.models.entity.model_params import KwargRule, ModelKwargsRules, ModelType, ModelMode
from core.model_providers.models.llm.xinference_model import XinferenceModel
from core.model_providers.providers.base import BaseModelProvider, CredentialsValidateFailedError
@@ -26,6 +26,9 @@ class XinferenceProvider(BaseModelProvider):
def _get_fixed_model_list(self, model_type: ModelType) -> list[dict]:
return []
def _get_text_generation_model_mode(self, model_name) -> str:
return ModelMode.COMPLETION.value
def get_model_class(self, model_type: ModelType) -> Type[BaseProviderModel]:
"""
Returns the model class.

9
api/core/model_providers/providers/zhipuai_provider.py
View File

@@ -7,7 +7,7 @@ from langchain.schema import HumanMessage
from core.helper import encrypter
from core.model_providers.models.base import BaseProviderModel
from core.model_providers.models.embedding.zhipuai_embedding import ZhipuAIEmbedding
from core.model_providers.models.entity.model_params import ModelKwargsRules, KwargRule, ModelType
from core.model_providers.models.entity.model_params import ModelKwargsRules, KwargRule, ModelType, ModelMode
from core.model_providers.models.llm.zhipuai_model import ZhipuAIModel
from core.model_providers.providers.base import BaseModelProvider, CredentialsValidateFailedError
from core.third_party.langchain.llms.zhipuai_llm import ZhipuAIChatLLM
@@ -29,18 +29,22 @@ class ZhipuAIProvider(BaseModelProvider):
{
'id': 'chatglm_pro',
'name': 'chatglm_pro',
'mode': ModelMode.CHAT.value,
},
{
'id': 'chatglm_std',
'name': 'chatglm_std',
'mode': ModelMode.CHAT.value,
},
{
'id': 'chatglm_lite',
'name': 'chatglm_lite',
'mode': ModelMode.CHAT.value,
},
{
'id': 'chatglm_lite_32k',
'name': 'chatglm_lite_32k',
'mode': ModelMode.CHAT.value,
}
]
elif model_type == ModelType.EMBEDDINGS:
@@ -53,6 +57,9 @@ class ZhipuAIProvider(BaseModelProvider):
else:
return []
def _get_text_generation_model_mode(self, model_name) -> str:
return ModelMode.CHAT.value
def get_model_class(self, model_type: ModelType) -> Type[BaseProviderModel]:
"""
Returns the model class.