refactor: move thinking toggle to runtime settings for dynamic control (#5800)

* refactor: move thinking toggle to runtime settings for per-message control

Replaces the static `reasoning_budget` config with a dynamic `enable_thinking` flag under `chat_template_kwargs`, allowing models like Jan-nano and Qwen3 to enable/disable thinking behavior at runtime, even mid-conversation.
Requires UI update

* remove engine argument

core/src/browser/extensions/engines/AIEngine.ts

@@ -98,6 +98,11 @@ export interface chatCompletionRequest {
   samplers?: string[] | null
   timings_per_token?: boolean | null
   post_sampling_probs?: boolean | null
+  chat_template_kwargs?: chat_template_kdict | null
+}
+
+export interface chat_template_kdict {
+  enable_thinking: false
 }
 
 export interface chatCompletionChunkChoiceDelta {

extensions/llamacpp-extension/settings.json

@@ -492,18 +492,5 @@
       "placeholder": "path/to/schema.json",
       "type": "text"
     }
-  },
-  {
-    "key": "reasoning_budget",
-    "title": "controls the amount of thinking allowed; currently only one of: -1 for unrestricted thinking budget, or 0 to disable thinking (default: -1)",
-    "description": "Mirostat target entropy (tau).",
-    "controllerType": "input",
-    "controllerProps": {
-      "value": 0,
-      "options": [
-        { "value": -1, "name": "unrestricted thinking budget" },
-        { "value": 0, "name": "disable thinking" }
-      ]
-    }
   }
 ]

extensions/llamacpp-extension/src/index.ts

@@ -25,7 +25,6 @@ import {
   downloadBackend,
   isBackendInstalled,
   getBackendExePath,
-  getBackendDir,
 } from './backend'
 import { invoke } from '@tauri-apps/api/core'
 
@@ -56,7 +55,6 @@ type LlamacppConfig = {
   rope_scale: number
   rope_freq_base: number
   rope_freq_scale: number
-  reasoning_budget: number
   ctx_shift: boolean
 }
 
@@ -855,7 +853,6 @@ export default class llamacpp_extension extends AIEngine {
       args.push('--rope-scale', String(cfg.rope_scale))
       args.push('--rope-freq-base', String(cfg.rope_freq_base))
       args.push('--rope-freq-scale', String(cfg.rope_freq_scale))
-      args.push('--reasoning-budget', String(cfg.reasoning_budget))
     }
 
     console.log('Calling Tauri command llama_load with args:', args)