Merge pull request #257 from NikkeTryHard/fix/thinking-budget-stability

CJackHwang · web-flow · commit 32dd637161fb · 2025-11-19T23:16:18.000+08:00
Fix thinking budget across models; gate levels to Gemini 3 Pro
diff --git a/browser_utils/page_controller.py b/browser_utils/page_controller.py
diff --git a/browser_utils/thinking_normalizer.py b/browser_utils/thinking_normalizer.py
@@ -123,22 +123,11 @@ def _parse_budget_value(reasoning_effort: Any) -> Optional[int]:
     if isinstance(reasoning_effort, int) and reasoning_effort > 0:
         return reasoning_effort
 
-    # 如果是字符串，尝试匹配预设值或解析为数字
+    # 如果是字符串，尝试解析为数字
     if isinstance(reasoning_effort, str):
         effort_str = reasoning_effort.strip().lower()
 
-        # 预设值映射
-        effort_map = {
-            "low": 1000,
-            "medium": 8000,
-            "high": 24000,
-        }
-
-        # 先尝试预设值
-        if effort_str in effort_map:
-            return effort_map[effort_str]
-
-        # 再尝试解析为数字
+        # 解析为数字
         try:
             value = int(effort_str)
             if value > 0:
diff --git a/config/selectors.py b/config/selectors.py
@@ -4,53 +4,74 @@
 """
 
 # --- 输入相关选择器 ---
-PROMPT_TEXTAREA_SELECTOR = 'ms-prompt-input-wrapper ms-autosize-textarea textarea'
+PROMPT_TEXTAREA_SELECTOR = "ms-prompt-input-wrapper ms-autosize-textarea textarea"
 INPUT_SELECTOR = PROMPT_TEXTAREA_SELECTOR
 INPUT_SELECTOR2 = PROMPT_TEXTAREA_SELECTOR
 
 # --- 按钮选择器 ---
 # 发送按钮：优先匹配 aria-label="Run" 的按钮；如页面结构变更，可退化到容器内的提交按钮。
 SUBMIT_BUTTON_SELECTOR = 'button[aria-label="Run"].run-button, ms-run-button button[type="submit"].run-button'
-CLEAR_CHAT_BUTTON_SELECTOR = 'button[data-test-clear="outside"][aria-label="New chat"]'
-CLEAR_CHAT_CONFIRM_BUTTON_SELECTOR = 'button.ms-button-primary:has-text("Discard and continue")'
+CLEAR_CHAT_BUTTON_SELECTOR = 'button[data-test-clear="outside"][aria-label="New chat"], button[aria-label="New chat"]'
+CLEAR_CHAT_CONFIRM_BUTTON_SELECTOR = (
+    'button.ms-button-primary:has-text("Discard and continue")'
+)
 UPLOAD_BUTTON_SELECTOR = 'button[aria-label^="Insert assets"]'
 
 # --- 响应相关选择器 ---
-RESPONSE_CONTAINER_SELECTOR = 'ms-chat-turn .chat-turn-container.model'
-RESPONSE_TEXT_SELECTOR = 'ms-cmark-node.cmark-node'
+RESPONSE_CONTAINER_SELECTOR = "ms-chat-turn .chat-turn-container.model"
+RESPONSE_TEXT_SELECTOR = "ms-cmark-node.cmark-node"
 
 # --- 加载和状态选择器 ---
 LOADING_SPINNER_SELECTOR = 'button[aria-label="Run"].run-button svg .stoppable-spinner'
-OVERLAY_SELECTOR = '.mat-mdc-dialog-inner-container'
+OVERLAY_SELECTOR = ".mat-mdc-dialog-inner-container"
 
 # --- 错误提示选择器 ---
-ERROR_TOAST_SELECTOR = 'div.toast.warning, div.toast.error'
+ERROR_TOAST_SELECTOR = "div.toast.warning, div.toast.error"
 
 # --- 编辑相关选择器 ---
-EDIT_MESSAGE_BUTTON_SELECTOR = 'ms-chat-turn:last-child .actions-container button.toggle-edit-button'
-MESSAGE_TEXTAREA_SELECTOR = 'ms-chat-turn:last-child ms-text-chunk ms-autosize-textarea'
+EDIT_MESSAGE_BUTTON_SELECTOR = (
+    "ms-chat-turn:last-child .actions-container button.toggle-edit-button"
+)
+MESSAGE_TEXTAREA_SELECTOR = "ms-chat-turn:last-child ms-text-chunk ms-autosize-textarea"
 FINISH_EDIT_BUTTON_SELECTOR = 'ms-chat-turn:last-child .actions-container button.toggle-edit-button[aria-label="Stop editing"]'
 
 # --- 菜单和复制相关选择器 ---
-MORE_OPTIONS_BUTTON_SELECTOR = 'div.actions-container div ms-chat-turn-options div > button'
-COPY_MARKDOWN_BUTTON_SELECTOR = 'button.mat-mdc-menu-item:nth-child(4)'
+MORE_OPTIONS_BUTTON_SELECTOR = (
+    "div.actions-container div ms-chat-turn-options div > button"
+)
+COPY_MARKDOWN_BUTTON_SELECTOR = "button.mat-mdc-menu-item:nth-child(4)"
 COPY_MARKDOWN_BUTTON_SELECTOR_ALT = 'div[role="menu"] button:has-text("Copy Markdown")'
 
 # --- 设置相关选择器 ---
 MAX_OUTPUT_TOKENS_SELECTOR = 'input[aria-label="Maximum output tokens"]'
 STOP_SEQUENCE_INPUT_SELECTOR = 'input[aria-label="Add stop token"]'
-MAT_CHIP_REMOVE_BUTTON_SELECTOR = 'mat-chip-set mat-chip-row button[aria-label*="Remove"]'
+MAT_CHIP_REMOVE_BUTTON_SELECTOR = (
+    'mat-chip-set mat-chip-row button[aria-label*="Remove"]'
+)
 TOP_P_INPUT_SELECTOR = 'ms-slider input[type="number"][max="1"]'
 TEMPERATURE_INPUT_SELECTOR = 'ms-slider input[type="number"][max="2"]'
 USE_URL_CONTEXT_SELECTOR = 'button[aria-label="Browse the url context"]'
 
 # --- 思考模式相关选择器 ---
 # 主思考开关：控制是否启用思考模式（总开关）
-ENABLE_THINKING_MODE_TOGGLE_SELECTOR = '[data-test-toggle="enable-thinking"] button'
+ENABLE_THINKING_MODE_TOGGLE_SELECTOR = (
+    'mat-slide-toggle[data-test-toggle="enable-thinking"] button[role="switch"].mdc-switch, '
+    '[data-test-toggle="enable-thinking"] button[role="switch"].mdc-switch'
+)
 # 手动预算开关：控制是否手动限制思考预算
-SET_THINKING_BUDGET_TOGGLE_SELECTOR = '[data-test-toggle="manual-budget"] button'
+SET_THINKING_BUDGET_TOGGLE_SELECTOR = (
+    'mat-slide-toggle[data-test-toggle="manual-budget"] button[role="switch"].mdc-switch, '
+    '[data-test-toggle="manual-budget"] button[role="switch"].mdc-switch'
+)
 # 思考预算输入框
 THINKING_BUDGET_INPUT_SELECTOR = '[data-test-slider] input[type="number"]'
 
+# 思考等级下拉
+THINKING_LEVEL_SELECT_SELECTOR = '[role="combobox"][aria-label="Thinking Level"], mat-select[aria-label="Thinking Level"], [role="combobox"][aria-label="Thinking level"], mat-select[aria-label="Thinking level"]'
+THINKING_LEVEL_OPTION_LOW_SELECTOR = '[role="listbox"][aria-label="Thinking Level"] [role="option"]:has-text("Low"), [role="listbox"][aria-label="Thinking level"] [role="option"]:has-text("Low")'
+THINKING_LEVEL_OPTION_HIGH_SELECTOR = '[role="listbox"][aria-label="Thinking Level"] [role="option"]:has-text("High"), [role="listbox"][aria-label="Thinking level"] [role="option"]:has-text("High")'
+
 # --- Google Search Grounding ---
-GROUNDING_WITH_GOOGLE_SEARCH_TOGGLE_SELECTOR = 'div[data-test-id="searchAsAToolTooltip"] mat-slide-toggle button'
+GROUNDING_WITH_GOOGLE_SEARCH_TOGGLE_SELECTOR = (
+    'div[data-test-id="searchAsAToolTooltip"] mat-slide-toggle button'
+)
diff --git a/docs/api-usage.md b/docs/api-usage.md
@@ -177,9 +177,9 @@ response = requests.post(
 使用预设值:
 ```json
 {
-  "reasoning_effort": "low"      // 1000 tokens
-  // 或 "medium"   (8000 tokens)
-  // 或 "high"     (24000 tokens)
+  "reasoning_effort": 1000       // 使用具体数值限制预算
+  // 或 8000
+  // 或 32768
 }
 ```
 
diff --git a/index.html b/index.html
@@ -240,6 +240,45 @@ <h3>生成参数</h3>
                             </div>
                         </div>
 
+                        <div class="settings-group" id="thinkingModeGroup">
+                            <label for="enableThinkingToggle">思考模式 (Thinking Mode):</label>
+                            <input type="checkbox" id="enableThinkingToggle">
+                            <div class="settings-description">
+                                关闭则不使用思考模式；开启时可选择“思考等级”或设置“思考预算”。
+                            </div>
+                        </div>
+
+                        <div class="settings-group" id="thinkingLevelGroup">
+                            <label for="thinkingLevelSelector">思考等级 (Thinking Level):</label>
+                            <select id="thinkingLevelSelector" class="settings-input">
+                                <option value="">未指定</option>
+                                <option value="low">Low</option>
+                                <option value="high">High</option>
+                            </select>
+                            <div class="settings-description">
+                                对于使用“思考等级”的模型（如 Gemini 3 Pro Preview），这里的选择将被优先使用。
+                            </div>
+                        </div>
+
+                        <div class="settings-group" id="manualBudgetGroup">
+                            <label for="enableManualBudgetToggle">限制思考预算 (Manual Budget):</label>
+                            <input type="checkbox" id="enableManualBudgetToggle">
+                            <div class="settings-description">
+                                打开以手动限制思考预算；关闭则不限制预算。
+                            </div>
+                        </div>
+
+                        <div class="settings-group" id="thinkingBudgetGroup">
+                            <label for="thinkingBudgetValue">思考预算 (Thinking Budget):</label>
+                            <div id="thinkingBudgetControlsContainer" class="settings-slider-container">
+                                <input type="range" id="thinkingBudgetSlider" class="settings-slider" min="128" max="32768" step="1" value="8192">
+                                <input type="number" id="thinkingBudgetValue" class="settings-number" min="128" max="32768" step="1" value="8192">
+                            </div>
+                            <div class="settings-description">
+                                对于使用预算的模型，设定具体的思考 Token 数。若同时指定“思考等级”，将优先使用等级。
+                            </div>
+                        </div>
+
                         <div class="settings-group">
                             <label for="stopSequences">停止序列 (Stop Sequences):</label>
                             <input type="text" id="stopSequences" class="settings-input" placeholder="用逗号分隔多个停止序列">
@@ -250,6 +289,15 @@ <h3>生成参数</h3>
                         </div>
                     </div>
 
+                    <div class="info-card">
+                        <h3>工具</h3>
+                        <div class="settings-group">
+                            <label for="enableGoogleSearchToggle">Grounding with Google Search</label>
+                            <input type="checkbox" id="enableGoogleSearchToggle">
+                            <div class="settings-description">启用搜索作为工具以改进事实性。</div>
+                        </div>
+                    </div>
+
                     <div class="info-card">
                         <h3>设置保存状态</h3>
                         <div id="settings-status" class="settings-status">
@@ -285,4 +333,4 @@ <h3>设置保存状态</h3>
     <script src="webui.js" defer></script>
 </body>
 
-</html>
+</html>
diff --git a/stream/interceptors.py b/stream/interceptors.py
@@ -22,6 +22,8 @@ def setup_logging():
                 logging.StreamHandler()
             ]
         )
+        logging.getLogger('asyncio').setLevel(logging.ERROR)
+        logging.getLogger('websockets').setLevel(logging.ERROR)
     
     @staticmethod
     def should_intercept(host, path):
diff --git a/stream/main.py b/stream/main.py
@@ -32,6 +32,8 @@ async def main():
             logging.StreamHandler()
         ]
     )
+    logging.getLogger('asyncio').setLevel(logging.ERROR)
+    logging.getLogger('websockets').setLevel(logging.ERROR)
     
     logger = logging.getLogger('main')
 
diff --git a/webui.js b/webui.js

Original file line number	Diff line number	Diff line change
`@@ -177,9 +177,9 @@ response = requests.post(`
`177`	`177`	`使用预设值:`
`178`	`178`	```json
`179`	`179`	`{`
`180`		`- "reasoning_effort": "low" // 1000 tokens`
`181`		`- // 或 "medium" (8000 tokens)`
`182`		`- // 或 "high" (24000 tokens)`
	`180`	`+ "reasoning_effort": 1000 // 使用具体数值限制预算`
	`181`	`+ // 或 8000`
	`182`	`+ // 或 32768`
`183`	`183`	`}`
`184`	`184`	```
`185`	`185`
Original file line number	Diff line number	Diff line change
`@@ -22,6 +22,8 @@ def setup_logging():`
`22`	`22`	`logging.StreamHandler()`
`23`	`23`	`]`
`24`	`24`	`)`
	`25`	`+ logging.getLogger('asyncio').setLevel(logging.ERROR)`
	`26`	`+ logging.getLogger('websockets').setLevel(logging.ERROR)`
`25`	`27`
`26`	`28`	`@staticmethod`
`27`	`29`	`def should_intercept(host, path):`
Original file line number	Diff line number	Diff line change
`@@ -32,6 +32,8 @@ async def main():`
`32`	`32`	`logging.StreamHandler()`
`33`	`33`	`]`
`34`	`34`	`)`
	`35`	`+ logging.getLogger('asyncio').setLevel(logging.ERROR)`
	`36`	`+ logging.getLogger('websockets').setLevel(logging.ERROR)`
`35`	`37`
`36`	`38`	`logger = logging.getLogger('main')`
`37`	`39`