🤖 feat: disable auto truncation by default for OpenAI models (#1217)

Change the default value of `disableAutoTruncation` from `false` to
`true`, so the dev-only "No Trunc" checkbox is checked by default.

---
_Generated with `mux` • Model: `mux-gateway:anthropic/claude-opus-4-5` •
Thinking: `high`_

Author: ibetitsmike

src/browser/contexts/ProviderOptionsContext.tsx

@@ -1,5 +1,9 @@
-import React, { createContext, useContext } from "react";
-import { usePersistedState } from "@/browser/hooks/usePersistedState";
+import React, { createContext, useContext, useLayoutEffect } from "react";
+import {
+  readPersistedState,
+  updatePersistedState,
+  usePersistedState,
+} from "@/browser/hooks/usePersistedState";
 import type { MuxProviderOptions } from "@/common/types/providerOptions";
 
 interface ProviderOptionsContextType {
@@ -11,6 +15,10 @@ interface ProviderOptionsContextType {
 
 const ProviderOptionsContext = createContext<ProviderOptionsContextType | undefined>(undefined);
 
+const OPENAI_OPTIONS_KEY = "provider_options_openai";
+// One-time migration key: force disableAutoTruncation to true for existing users
+const OPENAI_TRUNCATION_MIGRATION_KEY = "provider_options_openai_truncation_migrated";
+
 export function ProviderOptionsProvider({ children }: { children: React.ReactNode }) {
   const [anthropicOptions, setAnthropicOptions] = usePersistedState<
     MuxProviderOptions["anthropic"]
@@ -19,12 +27,20 @@ export function ProviderOptionsProvider({ children }: { children: React.ReactNod
   });
 
   const [openaiOptions, setOpenAIOptions] = usePersistedState<MuxProviderOptions["openai"]>(
-    "provider_options_openai",
-    {
-      disableAutoTruncation: false,
-    }
+    OPENAI_OPTIONS_KEY,
+    { disableAutoTruncation: true }
   );
 
+  // One-time migration: force disableAutoTruncation to true for existing users
+  useLayoutEffect(() => {
+    const alreadyMigrated = readPersistedState<boolean>(OPENAI_TRUNCATION_MIGRATION_KEY, false);
+    if (alreadyMigrated) {
+      return;
+    }
+    updatePersistedState(OPENAI_OPTIONS_KEY, { disableAutoTruncation: true });
+    updatePersistedState(OPENAI_TRUNCATION_MIGRATION_KEY, true);
+  }, []);
+
   const [googleOptions, setGoogleOptions] = usePersistedState<MuxProviderOptions["google"]>(
     "provider_options_google",
     {}

src/browser/utils/messages/sendOptions.ts

@@ -26,7 +26,7 @@ function getProviderOptions(): MuxProviderOptions {
     { use1MContext: false }
   );
   const openai = readPersistedState<MuxProviderOptions["openai"]>("provider_options_openai", {
-    disableAutoTruncation: false,
+    disableAutoTruncation: true,
   });
   const google = readPersistedState<MuxProviderOptions["google"]>("provider_options_google", {});
 

tests/ipc/sendMessage.heavy.test.ts

@@ -41,13 +41,21 @@ describeIntegration("sendMessage heavy/load tests", () => {
         await withSharedWorkspace(provider, async ({ env, workspaceId, collector }) => {
           // Build up large conversation history to exceed context limit
           // This approach is model-agnostic - it keeps sending until we've built up enough history
+          // Use auto-truncation enabled (disableAutoTruncation: false) so history builds up successfully
           const largeMessage = "x".repeat(50_000);
           for (let i = 0; i < 10; i++) {
             await sendMessageWithModel(
               env,
               workspaceId,
               `Message ${i}: ${largeMessage}`,
-              modelString(provider, model)
+              modelString(provider, model),
+              {
+                providerOptions: {
+                  openai: {
+                    disableAutoTruncation: false,
+                  },
+                },
+              }
             );
             await collector.waitForEvent("stream-end", 30000);
             collector.clear();
@@ -93,8 +101,14 @@ describeIntegration("sendMessage heavy/load tests", () => {
             env,
             workspaceId,
             "This should succeed with auto-truncation",
-            modelString(provider, model)
-            // disableAutoTruncation defaults to false (auto-truncation enabled)
+            modelString(provider, model),
+            {
+              providerOptions: {
+                openai: {
+                  disableAutoTruncation: false,
+                },
+              },
+            }
           );
 
           expect(successResult.success).toBe(true);