feat: use managed model capabilities for reasoning and image input

electron/api/routes/coding-conversations.ts

@@ -1,4 +1,5 @@
 import type { IncomingMessage, ServerResponse } from 'node:http';
+import { parseManagedReasoningChoice } from '../../../shared/managed-model-capabilities';
 import type { ConversationInteractionResponse } from '../../coding-runtime/contracts';
 import { CodingConversationServiceError } from '../../coding-runtime/conversation-service';
 import { normalizeProductModelRef } from '../../coding-projects/project-config';
@@ -182,12 +183,15 @@ export async function handleCodingConversationRoutes(
       return true;
     }
     if (action === 'thinking' && req.method === 'POST') {
-      const body = await parseJsonBody<{ thinkingLevel?: unknown }>(req);
+      const body = await parseJsonBody<{ thinkingLevel?: unknown; reasoningChoice?: unknown }>(req);
+      const reasoningChoice = body.reasoningChoice === undefined ? undefined : parseManagedReasoningChoice(body.reasoningChoice);
+      if (reasoningChoice === null) invalidRequest('Reasoning choice is invalid');
       if (!THINKING_LEVELS.has(String(body.thinkingLevel))) invalidRequest('Thinking level is invalid');
       sendJson(res, 200, {
         model: await service.setThinking(
           conversationId,
           body.thinkingLevel as 'off' | 'minimal' | 'low' | 'medium' | 'high' | 'max',
+          reasoningChoice ?? undefined,
         ),
       });
       return true;