Thinking levels: Off / Low / Medium / High (edge maps the level to the trained effort line)

README.md

@@ -16,7 +16,7 @@ Static chat client for the hosted OpenAI-compatible endpoint at `https://api.les
 - Model: `gemma-4-31b-humanlike-chat`
 - No API key required
 - Optional system prompt for a persona, setting, or relationship, kept in the browser between reloads
-- Thinking: off (default) or on
+- Thinking: off (default), low, medium, or high
 - One conversation per page load; refreshing starts a new session
 - Requests send only `model`, `messages`, `stream`, `max_tokens`, and `chat_template_kwargs`, so the endpoint defaults match the deployment gate
 

index.html

@@ -152,7 +152,9 @@
           <label for="reasoning">Thinking</label>
           <select id="reasoning">
             <option value="off">Off (fastest, the default)</option>
-            <option value="on">On</option>
+            <option value="low">Low</option>
+            <option value="medium">Medium</option>
+            <option value="high">High</option>
           </select>
         </div>
       </details>
@@ -172,7 +174,7 @@
     const API = "https://api.lessthanthreeai.com/v1";
     const MODEL = "gemma-4-31b-humanlike-chat";
     const INTRO = "Say something. The first reply can take a few minutes while the inference worker starts.";
-    const MAX_TOKENS = { off: 512, on: 4096 };
+    const MAX_TOKENS = { off: 512, low: 4096, medium: 4096, high: 8192 };
 
     const chat = document.querySelector("#chat");
     const form = document.querySelector("#form");
@@ -186,9 +188,9 @@
     const messages = [];
 
     system.value = localStorage.getItem("systemPrompt") || "";
-    reasoning.value = localStorage.getItem("thinking") === "on" ? "on" : "off";
+    reasoning.value = localStorage.getItem("thinkingLevel") in MAX_TOKENS ? localStorage.getItem("thinkingLevel") : "off";
     system.addEventListener("input", () => localStorage.setItem("systemPrompt", system.value));
-    reasoning.addEventListener("change", () => localStorage.setItem("thinking", reasoning.value));
+    reasoning.addEventListener("change", () => localStorage.setItem("thinkingLevel", reasoning.value));
 
     function addMessage(role, text) {
       const element = document.createElement("div");
@@ -266,7 +268,9 @@
         ? [{ role: "system", content: systemPrompt }, ...messages]
         : messages;
 
-      const chatTemplateKwargs = { enable_thinking: effort === "on", preserve_thinking: false };
+      const chatTemplateKwargs = effort === "off"
+        ? { enable_thinking: false, preserve_thinking: false }
+        : { enable_thinking: true, reasoning_effort: effort, preserve_thinking: false };
 
       try {
         const response = await fetch(`${API}/chat/completions`, {