AI/dify
1
0
Fork 0
mirror of https://git.mirrors.martin98.com/https://github.com/langgenius/dify.git synced 2025-08-12 20:59:00 +08:00
Code Issues Packages Projects Releases Wiki Activity

feat: add text-embedding functon and LLM models to Siliconflow (#7090)

Browse Source
This commit is contained in:
shAlfred 2024-08-08 17:08:28 +08:00 committed by GitHub
parent 12095f8cd6
commit a7162240e6
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
24 changed files with 466 additions and 22 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

28
api/core/model_runtime/model_providers/siliconflow/llm/_position.yaml
View File

@ -1,8 +1,20 @@
- deepseek-v2-chat
- qwen2-72b-instruct
- qwen2-57b-a14b-instruct
- qwen2-7b-instruct
- yi-1.5-34b-chat
- yi-1.5-9b-chat
- yi-1.5-6b-chat
- glm4-9B-chat
- Qwen/Qwen2-72B-Instruct
- Qwen/Qwen2-57B-A14B-Instruct
- Qwen/Qwen2-7B-Instruct
- Qwen/Qwen2-1.5B-Instruct
- 01-ai/Yi-1.5-34B-Chat
- 01-ai/Yi-1.5-9B-Chat-16K
- 01-ai/Yi-1.5-6B-Chat
- THUDM/glm-4-9b-chat
- deepseek-ai/DeepSeek-V2-Chat
- deepseek-ai/DeepSeek-Coder-V2-Instruct
- internlm/internlm2_5-7b-chat
- google/gemma-2-27b-it
- google/gemma-2-9b-it
- meta-llama/Meta-Llama-3-70B-Instruct
- meta-llama/Meta-Llama-3-8B-Instruct
- meta-llama/Meta-Llama-3.1-405B-Instruct
- meta-llama/Meta-Llama-3.1-70B-Instruct
- meta-llama/Meta-Llama-3.1-8B-Instruct
- mistralai/Mixtral-8x7B-Instruct-v0.1
- mistralai/Mistral-7B-Instruct-v0.2

2
api/core/model_runtime/model_providers/siliconflow/llm/deepseek-v2-chat.yaml
View File

@ -1,4 +1,4 @@
model: deepseek-ai/deepseek-v2-chat
model: deepseek-ai/DeepSeek-V2-Chat
label:
en_US: deepseek-ai/DeepSeek-V2-Chat
model_type: llm

30
api/core/model_runtime/model_providers/siliconflow/llm/gemma-2-27b-it.yaml Normal file
View File

@ -0,0 +1,30 @@
model: google/gemma-2-27b-it
label:
en_US: google/gemma-2-27b-it
model_type: llm
features:
- agent-thought
model_properties:
mode: chat
context_size: 8196
parameter_rules:
- name: temperature
use_template: temperature
- name: max_tokens
use_template: max_tokens
type: int
default: 512
min: 1
max: 4096
help:
zh_Hans: 指定生成结果长度的上限。如果生成结果截断，可以调大该参数。
en_US: Specifies the upper limit on the length of generated results. If the generated results are truncated, you can increase this parameter.
- name: top_p
use_template: top_p
- name: frequency_penalty
use_template: frequency_penalty
pricing:
input: '1.26'
output: '1.26'
unit: '0.000001'
currency: RMB

30
api/core/model_runtime/model_providers/siliconflow/llm/gemma-2-9b-it.yaml Normal file
View File

@ -0,0 +1,30 @@
model: google/gemma-2-9b-it
label:
en_US: google/gemma-2-9b-it
model_type: llm
features:
- agent-thought
model_properties:
mode: chat
context_size: 8196
parameter_rules:
- name: temperature
use_template: temperature
- name: max_tokens
use_template: max_tokens
type: int
default: 512
min: 1
max: 4096
help:
zh_Hans: 指定生成结果长度的上限。如果生成结果截断，可以调大该参数。
en_US: Specifies the upper limit on the length of generated results. If the generated results are truncated, you can increase this parameter.
- name: top_p
use_template: top_p
- name: frequency_penalty
use_template: frequency_penalty
pricing:
input: '0'
output: '0'
unit: '0.000001'
currency: RMB

6
api/core/model_runtime/model_providers/siliconflow/llm/glm4-9b-chat.yaml
View File

@ -1,4 +1,4 @@
model: zhipuai/glm4-9B-chat
model: THUDM/glm-4-9b-chat
label:
en_US: THUDM/glm-4-9b-chat
model_type: llm
@ -24,7 +24,7 @@ parameter_rules:
- name: frequency_penalty
use_template: frequency_penalty
pricing:
input: '0.6'
output: '0.6'
input: '0'
output: '0'
unit: '0.000001'
currency: RMB

30
api/core/model_runtime/model_providers/siliconflow/llm/internlm2_5-7b-chat.yaml Normal file
View File

@ -0,0 +1,30 @@
model: internlm/internlm2_5-7b-chat
label:
en_US: internlm/internlm2_5-7b-chat
model_type: llm
features:
- agent-thought
model_properties:
mode: chat
context_size: 32768
parameter_rules:
- name: temperature
use_template: temperature
- name: max_tokens
use_template: max_tokens
type: int
default: 512
min: 1
max: 4096
help:
zh_Hans: 指定生成结果长度的上限。如果生成结果截断，可以调大该参数。
en_US: Specifies the upper limit on the length of generated results. If the generated results are truncated, you can increase this parameter.
- name: top_p
use_template: top_p
- name: frequency_penalty
use_template: frequency_penalty
pricing:
input: '0'
output: '0'
unit: '0.000001'
currency: RMB

30
api/core/model_runtime/model_providers/siliconflow/llm/meta-mlama-3-70b-instruct.yaml Normal file
View File

@ -0,0 +1,30 @@
model: meta-llama/Meta-Llama-3-70B-Instruct
label:
en_US: meta-llama/Meta-Llama-3-70B-Instruct
model_type: llm
features:
- agent-thought
model_properties:
mode: chat
context_size: 32768
parameter_rules:
- name: temperature
use_template: temperature
- name: max_tokens
use_template: max_tokens
type: int
default: 512
min: 1
max: 4096
help:
zh_Hans: 指定生成结果长度的上限。如果生成结果截断，可以调大该参数。
en_US: Specifies the upper limit on the length of generated results. If the generated results are truncated, you can increase this parameter.
- name: top_p
use_template: top_p
- name: frequency_penalty
use_template: frequency_penalty
pricing:
input: '4.13'
output: '4.13'
unit: '0.000001'
currency: RMB

30
api/core/model_runtime/model_providers/siliconflow/llm/meta-mlama-3-8b-instruct.yaml Normal file
View File

@ -0,0 +1,30 @@
model: meta-llama/Meta-Llama-3-8B-Instruct
label:
en_US: meta-llama/Meta-Llama-3-8B-Instruct
model_type: llm
features:
- agent-thought
model_properties:
mode: chat
context_size: 8192
parameter_rules:
- name: temperature
use_template: temperature
- name: max_tokens
use_template: max_tokens
type: int
default: 512
min: 1
max: 4096
help:
zh_Hans: 指定生成结果长度的上限。如果生成结果截断，可以调大该参数。
en_US: Specifies the upper limit on the length of generated results. If the generated results are truncated, you can increase this parameter.
- name: top_p
use_template: top_p
- name: frequency_penalty
use_template: frequency_penalty
pricing:
input: '0'
output: '0'
unit: '0.000001'
currency: RMB

30
api/core/model_runtime/model_providers/siliconflow/llm/meta-mlama-3.1-405b-instruct.yaml Normal file
View File

@ -0,0 +1,30 @@
model: meta-llama/Meta-Llama-3.1-405B-Instruct
label:
en_US: meta-llama/Meta-Llama-3.1-405B-Instruct
model_type: llm
features:
- agent-thought
model_properties:
mode: chat
context_size: 32768
parameter_rules:
- name: temperature
use_template: temperature
- name: max_tokens
use_template: max_tokens
type: int
default: 512
min: 1
max: 4096
help:
zh_Hans: 指定生成结果长度的上限。如果生成结果截断，可以调大该参数。
en_US: Specifies the upper limit on the length of generated results. If the generated results are truncated, you can increase this parameter.
- name: top_p
use_template: top_p
- name: frequency_penalty
use_template: frequency_penalty
pricing:
input: '21'
output: '21'
unit: '0.000001'
currency: RMB

30
api/core/model_runtime/model_providers/siliconflow/llm/meta-mlama-3.1-70b-instruct.yaml Normal file
View File

@ -0,0 +1,30 @@
model: meta-llama/Meta-Llama-3.1-70B-Instruct
label:
en_US: meta-llama/Meta-Llama-3.1-70B-Instruct
model_type: llm
features:
- agent-thought
model_properties:
mode: chat
context_size: 32768
parameter_rules:
- name: temperature
use_template: temperature
- name: max_tokens
use_template: max_tokens
type: int
default: 512
min: 1
max: 4096
help:
zh_Hans: 指定生成结果长度的上限。如果生成结果截断，可以调大该参数。
en_US: Specifies the upper limit on the length of generated results. If the generated results are truncated, you can increase this parameter.
- name: top_p
use_template: top_p
- name: frequency_penalty
use_template: frequency_penalty
pricing:
input: '4.13'
output: '4.13'
unit: '0.000001'
currency: RMB

30
api/core/model_runtime/model_providers/siliconflow/llm/meta-mlama-3.1-8b-instruct.yaml Normal file
View File

@ -0,0 +1,30 @@
model: meta-llama/Meta-Llama-3.1-8B-Instruct
label:
en_US: meta-llama/Meta-Llama-3.1-8B-Instruct
model_type: llm
features:
- agent-thought
model_properties:
mode: chat
context_size: 8192
parameter_rules:
- name: temperature
use_template: temperature
- name: max_tokens
use_template: max_tokens
type: int
default: 512
min: 1
max: 4096
help:
zh_Hans: 指定生成结果长度的上限。如果生成结果截断，可以调大该参数。
en_US: Specifies the upper limit on the length of generated results. If the generated results are truncated, you can increase this parameter.
- name: top_p
use_template: top_p
- name: frequency_penalty
use_template: frequency_penalty
pricing:
input: '0'
output: '0'
unit: '0.000001'
currency: RMB

30
api/core/model_runtime/model_providers/siliconflow/llm/mistral-7b-instruct-v0.2.yaml Normal file
View File

@ -0,0 +1,30 @@
model: mistralai/Mistral-7B-Instruct-v0.2
label:
en_US: mistralai/Mistral-7B-Instruct-v0.2
model_type: llm
features:
- agent-thought
model_properties:
mode: chat
context_size: 32768
parameter_rules:
- name: temperature
use_template: temperature
- name: max_tokens
use_template: max_tokens
type: int
default: 512
min: 1
max: 4096
help:
zh_Hans: 指定生成结果长度的上限。如果生成结果截断，可以调大该参数。
en_US: Specifies the upper limit on the length of generated results. If the generated results are truncated, you can increase this parameter.
- name: top_p
use_template: top_p
- name: frequency_penalty
use_template: frequency_penalty
pricing:
input: '0'
output: '0'
unit: '0.000001'
currency: RMB

30
api/core/model_runtime/model_providers/siliconflow/llm/mistral-8x7b-instruct-v0.1.yaml Normal file
View File

@ -0,0 +1,30 @@
model: mistralai/Mixtral-8x7B-Instruct-v0.1
label:
en_US: mistralai/Mixtral-8x7B-Instruct-v0.1
model_type: llm
features:
- agent-thought
model_properties:
mode: chat
context_size: 32768
parameter_rules:
- name: temperature
use_template: temperature
- name: max_tokens
use_template: max_tokens
type: int
default: 512
min: 1
max: 4096
help:
zh_Hans: 指定生成结果长度的上限。如果生成结果截断，可以调大该参数。
en_US: Specifies the upper limit on the length of generated results. If the generated results are truncated, you can increase this parameter.
- name: top_p
use_template: top_p
- name: frequency_penalty
use_template: frequency_penalty
pricing:
input: '1.26'
output: '1.26'
unit: '0.000001'
currency: RMB

30
api/core/model_runtime/model_providers/siliconflow/llm/qwen2-1.5b-instruct.yaml Normal file
View File

@ -0,0 +1,30 @@
model: Qwen/Qwen2-1.5B-Instruct
label:
en_US: Qwen/Qwen2-1.5B-Instruct
model_type: llm
features:
- agent-thought
model_properties:
mode: chat
context_size: 32768
parameter_rules:
- name: temperature
use_template: temperature
- name: max_tokens
use_template: max_tokens
type: int
default: 512
min: 1
max: 4096
help:
zh_Hans: 指定生成结果长度的上限。如果生成结果截断，可以调大该参数。
en_US: Specifies the upper limit on the length of generated results. If the generated results are truncated, you can increase this parameter.
- name: top_p
use_template: top_p
- name: frequency_penalty
use_template: frequency_penalty
pricing:
input: '0'
output: '0'
unit: '0.000001'
currency: RMB

2
api/core/model_runtime/model_providers/siliconflow/llm/qwen2-57b-a14b-instruct.yaml
View File

@ -1,4 +1,4 @@
model: alibaba/Qwen2-57B-A14B-Instruct
model: Qwen/Qwen2-57B-A14B-Instruct
label:
en_US: Qwen/Qwen2-57B-A14B-Instruct
model_type: llm

2
api/core/model_runtime/model_providers/siliconflow/llm/qwen2-72b-instruct.yaml
View File

@ -1,4 +1,4 @@
model: alibaba/Qwen2-72B-Instruct
model: Qwen/Qwen2-72B-Instruct
label:
en_US: Qwen/Qwen2-72B-Instruct
model_type: llm

6
api/core/model_runtime/model_providers/siliconflow/llm/qwen2-7b-instruct.yaml
View File

@ -1,4 +1,4 @@
model: alibaba/Qwen2-7B-Instruct
model: Qwen/Qwen2-7B-Instruct
label:
en_US: Qwen/Qwen2-7B-Instruct
model_type: llm
@ -24,7 +24,7 @@ parameter_rules:
- name: frequency_penalty
use_template: frequency_penalty
pricing:
input: '0.35'
output: '0.35'
input: '0'
output: '0'
unit: '0.000001'
currency: RMB

4
api/core/model_runtime/model_providers/siliconflow/llm/yi-1.5-6b-chat.yaml
View File

@ -24,7 +24,7 @@ parameter_rules:
- name: frequency_penalty
use_template: frequency_penalty
pricing:
input: '0.35'
output: '0.35'
input: '0'
output: '0'
unit: '0.000001'
currency: RMB

6
api/core/model_runtime/model_providers/siliconflow/llm/yi-1.5-9b-chat.yaml
View File

@ -1,4 +1,4 @@
model: 01-ai/Yi-1.5-9B-Chat
model: 01-ai/Yi-1.5-9B-Chat-16K
label:
en_US: 01-ai/Yi-1.5-9B-Chat-16K
model_type: llm
@ -24,7 +24,7 @@ parameter_rules:
- name: frequency_penalty
use_template: frequency_penalty
pricing:
input: '0.42'
output: '0.42'
input: '0'
output: '0'
unit: '0.000001'
currency: RMB

1
api/core/model_runtime/model_providers/siliconflow/siliconflow.yaml
View File

@ -15,6 +15,7 @@ help:
en_US: https://cloud.siliconflow.cn/keys
supported_model_types:
- llm
- text-embedding
configurate_methods:
- predefined-model
provider_credential_schema:

5
api/core/model_runtime/model_providers/siliconflow/text_embedding/bge-large-en-v1.5.yaml Normal file
View File

@ -0,0 +1,5 @@
model: BAAI/bge-large-en-v1.5
model_type: text-embedding
model_properties:
context_size: 512
max_chunks: 1

5
api/core/model_runtime/model_providers/siliconflow/text_embedding/bge-large-zh-v1.5.yaml Normal file
View File

@ -0,0 +1,5 @@
model: BAAI/bge-large-zh-v1.5
model_type: text-embedding
model_properties:
context_size: 512
max_chunks: 1

29
api/core/model_runtime/model_providers/siliconflow/text_embedding/text_embedding.py Normal file
View File

@ -0,0 +1,29 @@
from typing import Optional
from core.model_runtime.entities.text_embedding_entities import TextEmbeddingResult
from core.model_runtime.model_providers.openai_api_compatible.text_embedding.text_embedding import (
OAICompatEmbeddingModel,
)
class SiliconflowTextEmbeddingModel(OAICompatEmbeddingModel):
"""
Model class for Siliconflow text embedding model.
"""
def validate_credentials(self, model: str, credentials: dict) -> None:
self._add_custom_parameters(credentials)
super().validate_credentials(model, credentials)
def _invoke(self, model: str, credentials: dict,
texts: list[str], user: Optional[str] = None) \
-> TextEmbeddingResult:
self._add_custom_parameters(credentials)
return super()._invoke(model, credentials, texts, user)
def get_num_tokens(self, model: str, credentials: dict, texts: list[str]) -> int:
self._add_custom_parameters(credentials)
return super().get_num_tokens(model, credentials, texts)
@classmethod
def _add_custom_parameters(cls, credentials: dict) -> None:
credentials['endpoint_url'] = 'https://api.siliconflow.cn/v1'

62
api/tests/integration_tests/model_runtime/siliconflow/test_text_embedding.py Normal file
View File

@ -0,0 +1,62 @@
import os
import pytest
from core.model_runtime.entities.text_embedding_entities import TextEmbeddingResult
from core.model_runtime.errors.validate import CredentialsValidateFailedError
from core.model_runtime.model_providers.siliconflow.text_embedding.text_embedding import (
SiliconflowTextEmbeddingModel,
)
def test_validate_credentials():
model = SiliconflowTextEmbeddingModel()
with pytest.raises(CredentialsValidateFailedError):
model.validate_credentials(
model="BAAI/bge-large-zh-v1.5",
credentials={
"api_key": "invalid_key"
},
)
model.validate_credentials(
model="BAAI/bge-large-zh-v1.5",
credentials={
"api_key": os.environ.get("API_KEY"),
},
)
def test_invoke_model():
model = SiliconflowTextEmbeddingModel()
result = model.invoke(
model="BAAI/bge-large-zh-v1.5",
credentials={
"api_key": os.environ.get("API_KEY"),
},
texts=[
"hello",
"world",
],
user="abc-123",
)
assert isinstance(result, TextEmbeddingResult)
assert len(result.embeddings) == 2
assert result.usage.total_tokens == 6
def test_get_num_tokens():
model = SiliconflowTextEmbeddingModel()
num_tokens = model.get_num_tokens(
model="BAAI/bge-large-zh-v1.5",
credentials={
"api_key": os.environ.get("API_KEY"),
},
texts=["hello", "world"],
)
assert num_tokens == 2