A browser extension that lets you summarize any webpage and ask questions using AI.
Add thinking mode toggle for Ollama models
+73
-14
+19
options/options.css
+19
options/options.css
···
52
52
margin-bottom: 6px;
53
53
}
54
54
55
+
/* Checkbox label styling */
56
+
label:has(input[type="checkbox"]) {
57
+
display: flex;
58
+
align-items: center;
59
+
gap: 8px;
60
+
cursor: pointer;
61
+
text-transform: none;
62
+
font-weight: 500;
63
+
font-size: 13.5px;
64
+
color: #1a1a1a;
65
+
}
66
+
67
+
input[type="checkbox"] {
68
+
width: 16px;
69
+
height: 16px;
70
+
cursor: pointer;
71
+
accent-color: #1a1a1a;
72
+
}
73
+
55
74
input[type="text"],
56
75
input[type="url"],
57
76
input[type="password"],
+12
options/options.html
+12
options/options.html
···
72
72
</p>
73
73
</div>
74
74
75
+
<div class="form-group" id="thinking-mode-group">
76
+
<label for="disable-thinking">
77
+
<input type="checkbox" id="disable-thinking" />
78
+
Disable thinking mode
79
+
</label>
80
+
<p class="help">
81
+
Disable the model's thinking process for faster responses.
82
+
Only applies to thinking-capable models (DeepSeek R1, Qwen 3, etc.).
83
+
<strong>Only works with Ollama native API.</strong>
84
+
</p>
85
+
</div>
86
+
75
87
<div class="form-group">
76
88
<label>Keyboard Shortcut</label>
77
89
<div class="shortcut-display">
+15
-1
options/options.js
+15
-1
options/options.js
···
5
5
const apiBaseUrlInput = document.getElementById("api-base-url");
6
6
const modelInput = document.getElementById("model");
7
7
const apiKeyInput = document.getElementById("api-key");
8
+
const disableThinkingInput = document.getElementById("disable-thinking");
9
+
const thinkingModeGroup = document.getElementById("thinking-mode-group");
8
10
const statusDiv = document.getElementById("status");
9
11
const testBtn = document.getElementById("test-connection");
10
12
const resetBtn = document.getElementById("reset-defaults");
···
14
16
apiBaseUrl: "http://localhost:11434",
15
17
model: "gpt-oss:20b-cloud",
16
18
apiKey: "",
19
+
disableThinking: false,
17
20
};
18
21
19
22
// Load settings on page load
20
23
document.addEventListener("DOMContentLoaded", loadSettings);
21
24
22
-
// Update URL placeholder when mode changes
25
+
// Update URL placeholder and thinking mode visibility when mode changes
23
26
apiModeInput.addEventListener("change", () => {
24
27
if (apiModeInput.value === "ollama") {
25
28
apiBaseUrlInput.placeholder = "http://localhost:11434";
29
+
thinkingModeGroup.style.display = "block";
26
30
} else {
27
31
apiBaseUrlInput.placeholder = "http://localhost:11434/v1";
32
+
thinkingModeGroup.style.display = "none";
28
33
}
29
34
});
30
35
···
37
42
apiBaseUrl: apiBaseUrlInput.value.trim() || defaultSettings.apiBaseUrl,
38
43
model: modelInput.value.trim() || defaultSettings.model,
39
44
apiKey: apiKeyInput.value.trim(),
45
+
disableThinking: disableThinkingInput.checked,
40
46
};
41
47
42
48
try {
···
54
60
apiBaseUrl: apiBaseUrlInput.value.trim() || defaultSettings.apiBaseUrl,
55
61
model: modelInput.value.trim() || defaultSettings.model,
56
62
apiKey: apiKeyInput.value.trim(),
63
+
disableThinking: disableThinkingInput.checked,
57
64
};
58
65
59
66
showStatus("🔄 Testing connection...", "loading");
···
71
78
{ role: "user", content: 'Say "Connection successful!"' },
72
79
],
73
80
apiMode: settings.apiMode,
81
+
disableThinking: settings.disableThinking,
74
82
},
75
83
});
76
84
···
112
120
apiBaseUrlInput.value = defaultSettings.apiBaseUrl;
113
121
modelInput.value = defaultSettings.model;
114
122
apiKeyInput.value = defaultSettings.apiKey;
123
+
disableThinkingInput.checked = defaultSettings.disableThinking;
124
+
// Show/hide thinking mode group based on API mode
125
+
thinkingModeGroup.style.display = defaultSettings.apiMode === "ollama" ? "block" : "none";
115
126
showStatus("Settings reset to defaults. Click Save to apply.", "success");
116
127
});
117
128
···
148
159
apiBaseUrlInput.value = settings.apiBaseUrl;
149
160
modelInput.value = settings.model;
150
161
apiKeyInput.value = settings.apiKey;
162
+
disableThinkingInput.checked = settings.disableThinking;
163
+
// Show/hide thinking mode group based on API mode
164
+
thinkingModeGroup.style.display = settings.apiMode === "ollama" ? "block" : "none";
151
165
} catch (error) {
152
166
showStatus("Error loading settings: " + error.message, "error");
153
167
}
+4
popup/popup.js
+4
popup/popup.js
···
507
507
apiBaseUrl: "http://localhost:11434",
508
508
model: "gpt-oss:20b-cloud",
509
509
apiKey: "",
510
+
disableThinking: false,
510
511
});
511
512
512
513
// Increased context limit for LLM (was 6000, now 12000)
···
532
533
apiKey: settings.apiKey,
533
534
messages: apiMessages,
534
535
apiMode: settings.apiMode,
536
+
disableThinking: settings.disableThinking,
535
537
},
536
538
});
537
539
···
610
612
apiBaseUrl: "http://localhost:11434",
611
613
model: "gpt-oss:20b-cloud",
612
614
apiKey: "",
615
+
disableThinking: false,
613
616
});
614
617
615
618
// Increased context limit for LLM (was 8000, now 15000)
···
637
640
apiKey: settings.apiKey,
638
641
messages: apiMessages,
639
642
apiMode: settings.apiMode,
643
+
disableThinking: settings.disableThinking,
640
644
},
641
645
});
642
646
+23
-13
scripts/background.js
+23
-13
scripts/background.js
···
20
20
apiBaseUrl: "http://localhost:11434",
21
21
model: "gpt-oss:20b-cloud",
22
22
apiKey: "",
23
+
disableThinking: false,
23
24
});
24
25
}
25
26
});
···
137
138
}
138
139
139
140
async function handleChatRequest(data) {
140
-
const { apiBaseUrl, model, apiKey, messages, apiMode } = data;
141
+
const { apiBaseUrl, model, apiKey, messages, apiMode, disableThinking } = data;
141
142
142
143
let useNativeOllama = apiMode === "ollama";
143
144
144
145
if (useNativeOllama) {
145
-
return await callOllamaNative(apiBaseUrl, model, messages);
146
+
return await callOllamaNative(apiBaseUrl, model, messages, disableThinking);
146
147
} else {
147
148
return await callOpenAICompatible(apiBaseUrl, model, apiKey, messages);
148
149
}
149
150
}
150
151
151
-
async function callOllamaNative(baseUrl, model, messages) {
152
+
async function callOllamaNative(baseUrl, model, messages, disableThinking) {
152
153
// Merge all system messages into one so none are dropped
153
154
const systemMsgs = messages.filter((m) => m.role === "system");
154
155
const systemContent = systemMsgs.map((m) => m.content).join("\n\n");
···
171
172
172
173
const url = baseUrl.replace(/\/$/, "") + "/api/generate";
173
174
175
+
// Build request body
176
+
const requestBody = {
177
+
model: model,
178
+
prompt: prompt,
179
+
system: systemContent,
180
+
stream: false,
181
+
options: {
182
+
temperature: 0.7,
183
+
num_predict: 2048,
184
+
},
185
+
};
186
+
187
+
// Only include think: false when user explicitly disables thinking
188
+
// (thinking is enabled by default in Ollama, and not all models support it)
189
+
if (disableThinking === true) {
190
+
requestBody.think = false;
191
+
}
192
+
174
193
const response = await fetch(url, {
175
194
method: "POST",
176
195
headers: {
177
196
"Content-Type": "application/json",
178
197
},
179
-
body: JSON.stringify({
180
-
model: model,
181
-
prompt: prompt,
182
-
system: systemContent,
183
-
stream: false,
184
-
options: {
185
-
temperature: 0.7,
186
-
num_predict: 2048,
187
-
},
188
-
}),
198
+
body: JSON.stringify(requestBody),
189
199
});
190
200
191
201
if (!response.ok) {