Skip to content

ff #2239

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Draft
QIN2DIM wants to merge 1 commit into
base: main
Choose a base branch
from
+131 −88
Draft

ff #2239

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
28 changes: 14 additions & 14 deletions models/siliconflow/models/llm/deepseek-v3.2.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -6,12 +6,25 @@ features:
- agent-thought
- tool-call
- stream-tool-call
- multi-tool-call
- document
- structured-output
model_properties:
mode: chat
context_size: 163840
parameter_rules:
- name: temperature
use_template: temperature
type: float
required: true
default: 1
min: 0
max: 2
label:
en_US: Temperature
zh_Hans: 温度
help:
en_US: For Gemini 3, best results at default 1.0. Lower values may impact reasoning.
zh_Hans: 对于 Gemini 3，使用默认值 1.0 可获得最佳效果。数值过低可能会影响推理能力。
Comment on lines +26 to +27
Copy link
Contributor

@gemini-code-assist gemini-code-assist bot Dec 12, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

medium

The help text for temperature appears to be incorrect. It mentions "Gemini 3", but this configuration is for a DeepSeek model. This was likely a copy-paste error and should be corrected to refer to the correct model to avoid confusing users.

      en_US: For DeepSeek V3.2, best results at default 1.0. Lower values may impact reasoning.
      zh_Hans: 对于 DeepSeek V3.2，使用默认值 1.0 可获得最佳效果。数值过低可能会影响推理能力.

- name: max_tokens
use_template: max_tokens
type: int
Expand All @@ -21,19 +34,6 @@ parameter_rules:
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: top_k
label:
zh_Hans: 取样数量
en_US: Top k
type: int
help:
zh_Hans: 仅从每个后续标记的前 K 个选项中采样。
en_US: Only sample from the top K options for each subsequent token.
required: false
- name: frequency_penalty
use_template: frequency_penalty
- name: response_format
label:
zh_Hans: 回复格式
Expand Down
191 changes: 117 additions & 74 deletions models/siliconflow/models/llm/llm.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -10,11 +10,50 @@
ModelType,
ParameterRule,
ParameterType,
DefaultParameterName,
)
from dify_plugin.entities.model.llm import LLMMode, LLMResult
from dify_plugin.entities.model.message import PromptMessage, PromptMessageTool


class RecommendModelSchema:
"""
This is not a post-patch override.
Parameters manually set by the user have a higher priority than this.
"""

@staticmethod
def set_deepseek_v32(entity: AIModelEntity):
pending_features = [
ModelFeature.TOOL_CALL,
ModelFeature.MULTI_TOOL_CALL,
ModelFeature.STREAM_TOOL_CALL,
ModelFeature.AGENT_THOUGHT,
]
Comment on lines +27 to +32
Copy link
Contributor

@gemini-code-assist gemini-code-assist bot Dec 12, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

medium

This list of features is defined inside the method. For better organization and to make it clear that it's a static configuration, consider defining it as a private class-level constant (e.g., _DEEPSEEK_V32_FEATURES). This improves readability and maintainability.


# Add missing features
for feature in pending_features:
if feature not in entity.features:
entity.features.append(feature)

# Set JSON schema to default
entity.parameter_rules.append(
ParameterRule(
name=DefaultParameterName.JSON_SCHEMA.value,
use_template=DefaultParameterName.JSON_SCHEMA.value,
) # type: ignore
)

# Set temperature to 1.0
for rule in entity.parameter_rules:
if rule.name == DefaultParameterName.TEMPERATURE.value:
rule.default = 1.0

@staticmethod
def patch_siliconflow_default(entity: AIModelEntity):
pass


class SiliconflowLargeLanguageModel(OAICompatLargeLanguageModel):
def _invoke(
self,
Expand All @@ -41,86 +80,90 @@ def validate_credentials(self, model: str, credentials: dict) -> None:
def _add_custom_parameters(cls, credentials: dict) -> None:
credentials["mode"] = "chat"
credentials["endpoint_url"] = "https://api.siliconflow.cn/v1"

def _add_function_call(self, model: str, credentials: dict) -> None:
model_schema = self.get_model_schema(model, credentials)
if model_schema and {ModelFeature.TOOL_CALL, ModelFeature.MULTI_TOOL_CALL}.intersection(
model_schema.features or []
):
credentials["function_calling_type"] = "tool_call"

def get_customizable_model_schema(
self, model: str, credentials: dict
) -> Optional[AIModelEntity]:
return AIModelEntity(
model=model,
label=I18nObject(en_US=model, zh_Hans=model),
model_type=ModelType.LLM,
features=(
[
ModelFeature.TOOL_CALL,
ModelFeature.MULTI_TOOL_CALL,
ModelFeature.STREAM_TOOL_CALL,
]
if credentials.get("function_calling_type") == "tool_call"
else []
),
fetch_from=FetchFrom.CUSTOMIZABLE_MODEL,
model_properties={
ModelPropertyKey.CONTEXT_SIZE: int(
credentials.get("context_size", 8000)
),
ModelPropertyKey.MODE: LLMMode.CHAT.value,
},
parameter_rules=[
ParameterRule(
name="temperature",
use_template="temperature",
label=I18nObject(en_US="Temperature", zh_Hans="温度"),
type=ParameterType.FLOAT,
),
ParameterRule(
name="max_tokens",
use_template="max_tokens",
default=4096,
min=1,
max=int(credentials.get("max_tokens", 16384)),
label=I18nObject(en_US="Max Tokens", zh_Hans="最大标记"),
type=ParameterType.INT,
),
ParameterRule(
name="top_p",
use_template="top_p",
label=I18nObject(en_US="Top P", zh_Hans="Top P"),
type=ParameterType.FLOAT,
),
ParameterRule(
name="top_k",
use_template="top_k",
label=I18nObject(en_US="Top K", zh_Hans="Top K"),
type=ParameterType.FLOAT,
),
ParameterRule(
name="frequency_penalty",
use_template="frequency_penalty",
label=I18nObject(en_US="Frequency Penalty", zh_Hans="重复惩罚"),
type=ParameterType.FLOAT,
),
ParameterRule(
name="enable_thinking",
use_template="enable_thinking",
default=True,
label=I18nObject(en_US="Thinking mode", zh_Hans="启用思考模式"),
type=ParameterType.BOOLEAN,
),
ParameterRule(
name="thinking_budget",
use_template="thinking_budget",
default=512,
min=1,
max=int(credentials.get("thinking_budget", 8192)),
label=I18nObject(en_US="Thinking budget", zh_Hans="思考长度限制"),
type=ParameterType.INT,
),
],
)
entity = super().get_customizable_model_schema(model, credentials)

print(entity)
Copy link
Contributor

@gemini-code-assist gemini-code-assist bot Dec 12, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

high

This print() statement appears to be for debugging purposes. It should be removed before merging to avoid polluting logs in a production environment and potentially exposing sensitive information.

if model in ["deepseek-ai/DeepSeek-V3.2", "Pro/deepseek-ai/DeepSeek-V3.2"]:
Copy link
Contributor

@gemini-code-assist gemini-code-assist bot Dec 12, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

medium

For better maintainability and performance, consider defining these model names as a module-level set constant. Using a set is more efficient for membership testing than a list.

Suggested change
if model in ["deepseek-ai/DeepSeek-V3.2", "Pro/deepseek-ai/DeepSeek-V3.2"]:
if model in {"deepseek-ai/DeepSeek-V3.2", "Pro/deepseek-ai/DeepSeek-V3.2"}:

RecommendModelSchema.set_deepseek_v32(entity)
return entity
# return AIModelEntity(
# model=model,
# label=I18nObject(en_US=model, zh_Hans=model),
# model_type=ModelType.LLM,
# features=(
# [
# ModelFeature.TOOL_CALL,
# ModelFeature.MULTI_TOOL_CALL,
# ModelFeature.STREAM_TOOL_CALL,
# ]
# if credentials.get("function_calling_type") == "tool_call"
# else []
# ),
# fetch_from=FetchFrom.CUSTOMIZABLE_MODEL,
# model_properties={
# ModelPropertyKey.CONTEXT_SIZE: int(credentials.get("context_size", 8000)),
# ModelPropertyKey.MODE: LLMMode.CHAT.value,
# },
# parameter_rules=[
# ParameterRule(
# name="temperature",
# use_template="temperature",
# label=I18nObject(en_US="Temperature", zh_Hans="温度"),
# type=ParameterType.FLOAT,
# ),
# ParameterRule(
# name="max_tokens",
# use_template="max_tokens",
# default=4096,
# min=1,
# max=int(credentials.get("max_tokens", 16384)),
# label=I18nObject(en_US="Max Tokens", zh_Hans="最大标记"),
# type=ParameterType.INT,
# ),
# ParameterRule(
# name="top_p",
# use_template="top_p",
# label=I18nObject(en_US="Top P", zh_Hans="Top P"),
# type=ParameterType.FLOAT,
# ),
# ParameterRule(
# name="top_k",
# use_template="top_k",
# label=I18nObject(en_US="Top K", zh_Hans="Top K"),
# type=ParameterType.FLOAT,
# ),
# ParameterRule(
# name="frequency_penalty",
# use_template="frequency_penalty",
# label=I18nObject(en_US="Frequency Penalty", zh_Hans="重复惩罚"),
# type=ParameterType.FLOAT,
# ),
# ParameterRule(
# name="enable_thinking",
# use_template="enable_thinking",
# default=True,
# label=I18nObject(en_US="Thinking mode", zh_Hans="启用思考模式"),
# type=ParameterType.BOOLEAN,
# ),
# ParameterRule(
# name="thinking_budget",
# use_template="thinking_budget",
# default=512,
# min=1,
# max=int(credentials.get("thinking_budget", 8192)),
# label=I18nObject(en_US="Thinking budget", zh_Hans="思考长度限制"),
# type=ParameterType.INT,
# ),
# ],
# )
Loading