fix: strip Anthropic params from 3P resume paths (#479)

* fix: strip Anthropic-specific params from 3P provider paths

Three silent failure modes affecting all third-party provider users:

1. Thinking blocks serialized as <thinking> text corrupt multi-turn
   context — strip them instead of converting to raw text tags.

2. Unknown models fall through to 200k context window default, so
   auto-compact never triggers — use conservative 8k for unknown
   3P models with a warning log.

3. Session resume with thinking blocks causes 400 or context corruption
   on 3P providers — strip thinking/redacted_thinking content blocks
   from deserialized messages when resuming against a non-Anthropic
   provider.

Addresses findings 2, 3, and 5 from #248.

* test: align resume stripping expectation with orphan-thinking filter

* test: isolate provider env in conversation recovery tests

* test: move provider-sensitive resume coverage behind module mocks

* test: trim extra blank lines in conversation recovery test

Keep the focused provider-resume test diff clean so the regression branch stays easy to review.

Co-Authored-By: Claude Opus 4.6 <noreply@openclaude.dev>

---------

Co-authored-by: Claude Opus 4.6 <noreply@openclaude.dev>

src/utils/context.ts

@@ -72,16 +72,23 @@ export function getContextWindowForModel(
     return 1_000_000
   }
 
-  // OpenAI-compatible provider — use known context windows for the model
-  if (
+  // OpenAI-compatible provider — use known context windows for the model.
+  // Unknown models get a conservative 8k default so auto-compact triggers
+  // before hitting a hard context_window_exceeded error (issue #248 finding 3).
+  const isOpenAIProvider =
     isEnvTruthy(process.env.CLAUDE_CODE_USE_OPENAI) ||
     isEnvTruthy(process.env.CLAUDE_CODE_USE_GEMINI) ||
     isEnvTruthy(process.env.CLAUDE_CODE_USE_GITHUB)
-  ) {
+  if (isOpenAIProvider) {
     const openaiWindow = getOpenAIContextWindow(model)
     if (openaiWindow !== undefined) {
       return openaiWindow
     }
+    console.error(
+      `[context] Warning: model "${model}" not in context window table — using conservative 8k default. ` +
+      'Add it to src/utils/model/openaiContextWindows.ts for accurate compaction.',
+    )
+    return 8_000
   }
 
   const cap = getModelCapability(model)

src/utils/conversationRecovery.hooks.test.ts

@@ -69,3 +69,93 @@ test('loadConversationForResume rejects oversized transcripts before resume hook
   )
   expect(hookSpy).not.toHaveBeenCalled()
 })
+
+test('deserializeMessagesWithInterruptDetection strips thinking blocks only for OpenAI-compatible providers', async () => {
+  const serializedMessages = [
+    user(id(10), 'hello'),
+    {
+      type: 'assistant',
+      uuid: id(11),
+      parentUuid: id(10),
+      timestamp: ts,
+      cwd: '/tmp',
+      sessionId,
+      version: 'test',
+      message: {
+        role: 'assistant',
+        content: [
+          { type: 'thinking', thinking: 'secret reasoning' },
+          { type: 'text', text: 'visible reply' },
+        ],
+      },
+    },
+    {
+      type: 'assistant',
+      uuid: id(12),
+      parentUuid: id(11),
+      timestamp: ts,
+      cwd: '/tmp',
+      sessionId,
+      version: 'test',
+      message: {
+        role: 'assistant',
+        content: [{ type: 'thinking', thinking: 'only hidden reasoning' }],
+      },
+    },
+    user(id(13), 'follow up'),
+  ]
+
+  mock.module('./model/providers.js', () => ({
+    getAPIProvider: () => 'openai',
+    isOpenAICompatibleProvider: (provider: string) =>
+      provider === 'openai' ||
+      provider === 'gemini' ||
+      provider === 'github' ||
+      provider === 'codex',
+  }))
+
+  const openaiModule = await import(`./conversationRecovery.ts?provider=openai-${Date.now()}`)
+  const thirdParty = openaiModule.deserializeMessagesWithInterruptDetection(serializedMessages as never[])
+  const thirdPartyAssistantMessages = thirdParty.messages.filter(
+    message => message.type === 'assistant',
+  )
+
+  expect(thirdPartyAssistantMessages).toHaveLength(2)
+  expect(thirdPartyAssistantMessages[0]?.message?.content).toEqual([
+    { type: 'text', text: 'visible reply' },
+  ])
+  expect(
+    JSON.stringify(thirdPartyAssistantMessages.map(message => message.message?.content)),
+  ).not.toContain('secret reasoning')
+  expect(
+    JSON.stringify(thirdPartyAssistantMessages.map(message => message.message?.content)),
+  ).not.toContain('only hidden reasoning')
+
+  mock.restore()
+  mock.module('./model/providers.js', () => ({
+    getAPIProvider: () => 'bedrock',
+    isOpenAICompatibleProvider: (provider: string) =>
+      provider === 'openai' ||
+      provider === 'gemini' ||
+      provider === 'github' ||
+      provider === 'codex',
+  }))
+
+  const bedrockModule = await import(`./conversationRecovery.ts?provider=bedrock-${Date.now()}`)
+  const anthropicCompatible = bedrockModule.deserializeMessagesWithInterruptDetection(serializedMessages as never[])
+  const anthropicAssistantMessages = anthropicCompatible.messages.filter(
+    message => message.type === 'assistant',
+  )
+
+  expect(anthropicAssistantMessages).toHaveLength(2)
+  expect(anthropicAssistantMessages[0]?.message?.content).toEqual([
+    { type: 'thinking', thinking: 'secret reasoning' },
+    { type: 'text', text: 'visible reply' },
+  ])
+  expect(
+    JSON.stringify(anthropicAssistantMessages.map(message => message.message?.content)),
+  ).toContain('secret reasoning')
+  expect(
+    JSON.stringify(anthropicAssistantMessages.map(message => message.message?.content)),
+  ).not.toContain('only hidden reasoning')
+})

src/utils/conversationRecovery.test.ts

@@ -13,6 +13,7 @@ const originalSimple = process.env.CLAUDE_CODE_SIMPLE
 const sessionId = '00000000-0000-4000-8000-000000001999'
 const ts = '2026-04-02T00:00:00.000Z'
 
+
 function id(n: number): string {
   return `00000000-0000-4000-8000-${String(n).padStart(12, '0')}`
 }
@@ -76,4 +77,3 @@ test('loadConversationForResume rejects oversized reconstructed transcripts', as
     'Reconstructed transcript is too large to resume safely',
   )
 })
-

src/utils/conversationRecovery.ts

@@ -24,6 +24,7 @@ import {
   type FileHistorySnapshot,
 } from './fileHistory.js'
 import { logError } from './log.js'
+import { getAPIProvider } from './model/providers.js'
 import {
   createAssistantMessage,
   createUserMessage,
@@ -177,6 +178,25 @@ export type DeserializeResult = {
   turnInterruptionState: TurnInterruptionState
 }
 
+/**
+ * Remove thinking/redacted_thinking content blocks from assistant messages.
+ * Messages that become empty after stripping are removed entirely.
+ */
+function stripThinkingBlocks(messages: NormalizedMessage[]): NormalizedMessage[] {
+  return messages.reduce<NormalizedMessage[]>((acc, msg) => {
+    if (msg.type !== 'assistant' || !Array.isArray(msg.message?.content)) {
+      acc.push(msg)
+      return acc
+    }
+    const filtered = msg.message.content.filter(
+      (block: { type?: string }) => block.type !== 'thinking' && block.type !== 'redacted_thinking',
+    )
+    if (filtered.length === 0) return acc
+    acc.push({ ...msg, message: { ...msg.message, content: filtered } })
+    return acc
+  }, [])
+}
+
 /**
  * Deserializes messages from a log file into the format expected by the REPL.
  * Filters unresolved tool uses, orphaned thinking messages, and appends a
@@ -227,10 +247,19 @@ export function deserializeMessagesWithInterruptDetection(
       filteredToolUses,
     ) as NormalizedMessage[]
 
+    // Strip thinking/redacted_thinking content blocks from assistant messages
+    // when resuming against a 3P provider. These Anthropic-specific blocks cause
+    // 400 errors or context corruption on OpenAI-compatible providers (issue #248 finding 5).
+    const provider = getAPIProvider()
+    const isThirdPartyProvider = provider !== 'firstParty' && provider !== 'bedrock' && provider !== 'vertex' && provider !== 'foundry'
+    const thinkingStripped = isThirdPartyProvider
+      ? stripThinkingBlocks(filteredThinking)
+      : filteredThinking
+
     // Filter out assistant messages with only whitespace text content.
     // This can happen when model outputs "\n\n" before thinking, user cancels mid-stream.
     const filteredMessages = filterWhitespaceOnlyAssistantMessages(
-      filteredThinking,
+      thinkingStripped,
     ) as NormalizedMessage[]
 
     const internalState = detectTurnInterruption(filteredMessages)