fix unit tests

Author: charleslien

.agents/__tests__/context-pruner.test.ts

@@ -307,22 +307,22 @@ describe('context-pruner edge cases', () => {
     content,
   })
 
-  const createTerminalToolMessage = (command: string, output: string): any => ({
+  const createTerminalToolMessage = (
+    command: string,
+    output: string,
+  ): ToolMessage => ({
     role: 'tool',
-    content: {
-      type: 'tool-result',
-      toolCallId: 'test-id',
-      toolName: 'run_terminal_command',
-      output: [
-        {
-          type: 'json',
-          value: {
-            command,
-            stdout: output,
-          },
+    toolCallId: 'test-id',
+    toolName: 'run_terminal_command',
+    content: [
+      {
+        type: 'json',
+        value: {
+          command,
+          stdout: output,
         },
-      ],
-    },
+      },
+    ],
   })
 
   const runHandleSteps = (messages: Message[]) => {
@@ -441,21 +441,21 @@ describe('context-pruner edge cases', () => {
     // Create content large enough to exceed 200k token limit to trigger pruning
     const largeContent = 'x'.repeat(150000)
 
-    const createToolMessage = (toolName: string, size: number): any => ({
+    const createToolMessage = (
+      toolName: string,
+      size: number,
+    ): ToolMessage => ({
       role: 'tool',
-      content: {
-        type: 'tool-result',
-        toolCallId: 'test-id',
-        toolName,
-        output: [
-          {
-            type: 'json',
-            value: {
-              data: 'a'.repeat(size),
-            },
+      toolCallId: 'test-id',
+      toolName,
+      content: [
+        {
+          type: 'json',
+          value: {
+            data: 'a'.repeat(size),
           },
-        ],
-      },
+        },
+      ],
     })
 
     const messages = [

common/src/util/__tests__/messages.test.ts

@@ -210,14 +210,14 @@ describe('convertCbToModelMessages', () => {
       })
 
       expect(result).toEqual([
-        {
+        expect.objectContaining({
           role: 'user',
           content: [
             expect.objectContaining({
               type: 'text',
             }),
           ],
-        },
+        }),
       ])
       expect((result as any)[0].content[0].text).toContain('<tool_result>')
     })
@@ -243,14 +243,14 @@ describe('convertCbToModelMessages', () => {
       })
 
       expect(result).toEqual([
-        {
+        expect.objectContaining({
           role: 'user',
           content: [
             expect.objectContaining({
               type: 'file',
             }),
           ],
-        },
+        }),
       ])
     })
 
@@ -260,7 +260,10 @@ describe('convertCbToModelMessages', () => {
           role: 'tool',
           toolName: 'test_tool',
           toolCallId: 'call_123',
-          content: [toolJsonContent({ result1: 'success' })],
+          content: [
+            toolJsonContent({ result1: 'success' }),
+            toolJsonContent({ result2: 'also success' }),
+          ],
         },
       ]
 
@@ -269,6 +272,7 @@ describe('convertCbToModelMessages', () => {
         includeCacheControl: false,
       })
 
+      console.dir({ result }, { depth: null })
       // Multiple tool outputs are aggregated into one user message
       expect(result).toEqual([
         expect.objectContaining({

packages/agent-runtime/src/__tests__/run-programmatic-step.test.ts

@@ -330,10 +330,7 @@ describe('runProgrammaticStep', () => {
           }
           options.toolResults.push(toolResult)
 
-          options.state.messages.push({
-            role: 'tool',
-            content: toolResult,
-          })
+          options.state.messages.push(toolResult)
         }
         // Return a value to satisfy the call
         return {}

packages/agent-runtime/src/tools/tool-executor.ts

@@ -299,10 +299,7 @@ export function executeToolCall<T extends ToolName>(
     toolResults.push(toolResult)
 
     if (!excludeToolFromMessageHistory) {
-      state.messages.push({
-        role: 'tool' as const,
-        content: toolResult,
-      })
+      state.messages.push(toolResult)
     }
 
     // After tool completes, resolve any pending creditsUsed promise