feat(aihubmix): add qwen3.7-plus/step-3.7-flash and fix hy3 reasoning control (#3427)

* fix(aihubmix): use reasoning_effort for hy3 thinking control

aihubmix gateway rejects boolean `thinking` for Hunyuan (hy3) with 400: "cannot unmarshal bool into ... model.Thinking". The `thinking` field must be an object {type:enabled|disabled}, which a Dify yaml parameter rule cannot emit.

Switch hy3.yaml from `thinking`(boolean) + `thinking_budget`(int) to a single `reasoning_effort` string parameter (none/low/medium/high, default none). Verified on aihubmix: all four values return 200; none disables thinking (reasoning_tokens=0), low/medium/high enable it. hy3 + reasoning_effort:high + json_schema end-to-end returns 200 with schema-conformant JSON.

Per Tencent Hunyuan docs, hy3 reasoning_effort supports low/medium/high (default low); none is an aihubmix extension to turn thinking off. llm.py is unchanged.

* feat(aihubmix): add qwen3.7-plus and step-3.7-flash models

* chore(aihubmix): bump version to 0.0.38

Author: orangeclk

models/aihubmix/manifest.yaml

@@ -1,4 +1,4 @@
-version: 0.0.37
+version: 0.0.38
 type: plugin
 author: langgenius
 name: aihubmix

models/aihubmix/models/llm/_position.yaml

@@ -12,6 +12,7 @@
 - doubao-seed-2-1-turbo
 - grok-build-0.1
 - qwen3.7-max
+- qwen3.7-plus
 - gemini-3.5-flash
 - grok-4.3
 - gpt-5.5
@@ -102,6 +103,7 @@
 - coding-minimax-m2.7
 - coding-minimax-m2.7-highspeed
 - o3
+- step-3.7-flash
 - step-3.5-flash
 - glm-4.6
 - kimi-for-coding-free

models/aihubmix/models/llm/hy3.yaml

@@ -11,34 +11,21 @@ model_properties:
   mode: chat
   context_size: 256000
 parameter_rules:
-  - name: thinking
+  - name: reasoning_effort
     label:
-      zh_Hans: 推理模式
-      en_US: Thinking Mode
-    type: boolean
-    default: false
+      zh_Hans: 推理深度
+      en_US: Reasoning Effort
+    type: string
     required: false
+    default: none
+    options:
+      - none
+      - low
+      - medium
+      - high
     help:
-      zh_Hans: 控制模型的推理能力。开启后使用深度推理模式。
-      en_US: Controls the model's thinking capability. Enables deep reasoning when turned on.
-  - name: thinking_budget
-    label:
-      zh_Hans: 推理预算
-      en_US: Thinking Budget
-    type: int
-    default: 1024
-    min: 0
-    max: 32000
-    required: false
-  - name: temperature
-    use_template: temperature
-  - name: top_p
-    use_template: top_p
-  - name: max_tokens
-    use_template: max_tokens
-    default: 8192
-    min: 1
-    max: 128000
+      zh_Hans: 控制推理深度。none 关闭思考；low 轻量推理（速度快，适合简单任务）、medium 平衡（适合日常适中复杂任务）、high 深度推理（延迟与成本最高，适合高难度数学/编程/复杂逻辑）。
+      en_US: Controls reasoning depth. none disables thinking; low = lightweight (fast, simple tasks), medium = balanced (everyday tasks), high = deep reasoning (highest latency/cost, for hard math/coding/logic).
   - name: response_format
     label:
       zh_Hans: 回复格式
@@ -52,7 +39,7 @@ parameter_rules:
   - name: json_schema
     use_template: json_schema
 pricing:
-  input: '0.1562'
-  output: '0.6248'
-  unit: '0.000001'
+  input: "0.1562"
+  output: "0.6248"
+  unit: "0.000001"
   currency: USD

models/aihubmix/models/llm/qwen3.7-plus.yaml

@@ -0,0 +1,42 @@
+model: qwen3.7-plus
+label:
+  en_US: Qwen3.7 Plus
+model_type: llm
+features:
+  - agent-thought
+  - tool-call
+  - stream-tool-call
+  - multi-tool-call
+  - vision
+  - video
+model_properties:
+  mode: chat
+  context_size: 1014784
+parameter_rules:
+  - name: enable_thinking
+    required: false
+    type: boolean
+    default: false
+    label:
+      zh_Hans: 深度思考
+      en_US: Deep Thinking
+    help:
+      zh_Hans: 是否开启深度思考模式，开启后模型会先进行推理再给出回答。
+      en_US: Whether to enable deep thinking mode. When enabled, the model will reason before answering.
+  - name: response_format
+    label:
+      zh_Hans: 回复格式
+      en_US: Response Format
+    type: string
+    required: false
+    options:
+      - text
+      - json_object
+      - json_schema
+  - name: json_schema
+    use_template: json_schema
+pricing:
+  input: '2'
+  output: '8'
+  unit: '0.000001'
+  currency: RMB

models/aihubmix/models/llm/step-3.7-flash.yaml

@@ -0,0 +1,46 @@
+model: step-3.7-flash
+label:
+  en_US: Step 3.7 Flash
+model_type: llm
+features:
+  - agent-thought
+  - tool-call
+  - stream-tool-call
+  - multi-tool-call
+  - vision
+  - video
+model_properties:
+  mode: chat
+  context_size: 262144
+parameter_rules:
+  - name: reasoning_effort
+    label:
+      zh_Hans: 推理强度
+      en_US: Reasoning Effort
+    type: string
+    required: false
+    default: medium
+    options:
+      - low
+      - medium
+      - high
+    help:
+      zh_Hans: "根据任务复杂度选择合适档位。low 适合简单问答、摘要、改写、信息抽取；medium 默认推荐，适合一般推理和多步骤任务；high 适合复杂推理、数学、规划、代码分析。"
+      en_US: "Select effort by task complexity. low: simple Q&A, summary, rewrite, extraction; medium: default for general reasoning and multi-step tasks; high: complex reasoning, math, planning, code analysis."
+  - name: response_format
+    label:
+      zh_Hans: 回复格式
+      en_US: Response Format
+    type: string
+    required: false
+    options:
+      - text
+      - json_object
+      - json_schema
+  - name: json_schema
+    use_template: json_schema
+pricing:
+  input: '1.35'
+  output: '8.1'
+  unit: '0.000001'
+  currency: RMB