fix: make ai-agent test context dynamic for proper test isolation

Convert baseContext from module-level constant to createTestContext() function
to ensure each test gets a fresh context with the current mocking state.
This fixes the race condition in CI where nextGenerateResultOverride wasn't
being properly captured by the MockToolLoopAgent due to global state sharing.

Also fixes missing getVolumeName() method in opencode.test.ts IsolatedContainerVolume mock.

Signed-off-by: betterclever <[email protected]>
Amp-Thread-ID: https://ampcode.com/threads/T-019c1a9e-e1f3-755b-99e7-5a617c149f9e
Co-authored-by: Amp <[email protected]>

Author: betterclever

worker/src/components/ai/__tests__/ai-agent.test.ts

@@ -31,32 +31,35 @@ class MockToolLoopAgent {
   }
 }
 
-const baseContext: ExecutionContext = {
-  runId: 'test-run',
-  componentRef: 'core.ai.agent',
-  logger: {
-    debug: () => {},
-    info: () => {},
-    error: () => {},
-    warn: () => {},
-  },
-  emitProgress: () => {},
-  metadata: {
+function createTestContext(overrides?: Partial<ExecutionContext>): ExecutionContext {
+  return {
     runId: 'test-run',
     componentRef: 'core.ai.agent',
-    aiSdkOverrides: {
-      ToolLoopAgent: MockToolLoopAgent,
-      stepCountIs: stepCountIsMock,
-      createOpenAI: createOpenAIMock,
-      createGoogleGenerativeAI: createGoogleGenerativeAIMock,
-      createMCPClient: createMCPClientMock,
+    logger: {
+      debug: () => {},
+      info: () => {},
+      error: () => {},
+      warn: () => {},
+    },
+    emitProgress: () => {},
+    metadata: {
+      runId: 'test-run',
+      componentRef: 'core.ai.agent',
+      aiSdkOverrides: {
+        ToolLoopAgent: MockToolLoopAgent,
+        stepCountIs: stepCountIsMock,
+        createOpenAI: createOpenAIMock,
+        createGoogleGenerativeAI: createGoogleGenerativeAIMock,
+        createMCPClient: createMCPClientMock,
+      },
     },
-  },
-  http: {
-    fetch: async () => new Response(),
-    toCurl: () => '',
-  },
-};
+    http: {
+      fetch: async () => new Response(),
+      toCurl: () => '',
+    },
+    ...overrides,
+  };
+}
 
 function createUsage(overrides: Partial<LanguageModelUsage> = {}): LanguageModelUsage {
   return {
@@ -165,7 +168,7 @@ describe('core.ai.agent (refactor)', () => {
           stepLimit: 2,
         },
       },
-      baseContext,
+      createTestContext(),
     );
 
     expect(result.responseText).toBe('Hello agent');
@@ -211,13 +214,12 @@ describe('core.ai.agent (refactor)', () => {
       close: async () => {},
     });
 
-    const contextWithTools: ExecutionContext = {
-      ...baseContext,
+    const contextWithTools: ExecutionContext = createTestContext({
       metadata: {
-        ...baseContext.metadata,
+        ...createTestContext().metadata,
         connectedToolNodeIds: ['tool-node-1'],
       },
-    };
+    });
 
     try {
       const result = await runComponentWithRunner(
@@ -303,7 +305,7 @@ describe('core.ai.agent (refactor)', () => {
           stepLimit: 2,
         },
       },
-      baseContext,
+      createTestContext(),
     );
 
     const toolMessage = result.conversationState.messages.find(