Merge branch 'main' into brandon/benchify

brandonkachen · brandonkachen · commit b16754d8c51e · 2025-09-30T16:06:25.000-07:00
diff --git a/.agents/base-max.ts b/.agents/base-max.ts
@@ -9,8 +9,8 @@ const definition: SecretAgentDefinition = {
   ...base('anthropic/claude-sonnet-4.5', 'max'),
   spawnableAgents: [
     'file-explorer',
-    'researcher-web-gpt-5',
-    'researcher-docs-gpt-5',
+    'researcher-web-sonnet',
+    'researcher-docs-sonnet',
     'implementation-planner-max',
     'thinker',
     'reviewer-max',
diff --git a/.agents/base.ts b/.agents/base.ts
@@ -6,7 +6,7 @@ import type { SecretAgentDefinition } from './types/secret-agent-definition'
 const definition: SecretAgentDefinition = {
   id: 'base',
   publisher,
-  ...base('anthropic/claude-4.5-sonnet', 'normal'),
+  ...base('anthropic/claude-sonnet-4.5', 'normal'),
 }
 
 export default definition
diff --git a/.agents/docs-researcher.ts b/.agents/docs-researcher.ts
@@ -6,7 +6,7 @@ import type { SecretAgentDefinition } from './types/secret-agent-definition'
 const definition: SecretAgentDefinition = {
   id: 'docs-researcher',
   publisher,
-  ...docsResearcher('google/gemini-2.5-flash'),
+  ...docsResearcher('x-ai/grok-4-fast'),
 }
 
 export default definition
diff --git a/.agents/implementation-planner/implementation-planner-max.ts b/.agents/implementation-planner/implementation-planner-max.ts
@@ -4,7 +4,7 @@ import { type SecretAgentDefinition } from '../types/secret-agent-definition'
 const definition: SecretAgentDefinition = {
   id: 'implementation-planner-max',
   publisher,
-  model: 'anthropic/claude-4.5-sonnet',
+  model: 'anthropic/claude-sonnet-4.5',
   displayName: 'Implementation Planner Max',
   spawnerPrompt:
     'Creates the best possible implementation plan by generating several different plans in parallel and selecting the best one. Includes full code changes.',
diff --git a/.agents/implementation-planner/implementation-planner.ts b/.agents/implementation-planner/implementation-planner.ts
@@ -8,7 +8,7 @@ const definition: SecretAgentDefinition = {
   id: 'implementation-planner',
   displayName: 'Implementation Planner',
   publisher,
-  model: 'anthropic/claude-4.5-sonnet',
+  model: 'anthropic/claude-sonnet-4.5',
   spawnerPrompt:
     'Creates comprehensive implementation plans with full code changes by exploring the codebase, doing research on the web, and thinking deeply. You can also use it get a deep answer to any question. Use this agent for tasks that require thinking.',
   inputSchema: {
diff --git a/.agents/implementation-planner/plan-selector.ts b/.agents/implementation-planner/plan-selector.ts
@@ -7,7 +7,7 @@ import {
 const definition: SecretAgentDefinition = {
   id: 'plan-selector',
   publisher,
-  model: 'anthropic/claude-4.5-sonnet',
+  model: 'anthropic/claude-sonnet-4.5',
   displayName: 'Plan Selector',
   spawnerPrompt:
     'Expert at evaluating and selecting the best plan from multiple options based on quality, feasibility, and simplicity.',
diff --git a/.agents/prompts/base-prompts.ts b/.agents/prompts/base-prompts.ts
@@ -254,7 +254,7 @@ export const baseAgentUserInputPrompt = (
       `Important: you *must* read as many files with the read_files tool as possible from the results of the file picker agents. Don't be afraid to read ${isLite ? '8' : '20'} files. The more files you read, the better context you have on the codebase and the better your response will be.`,
 
       isMax &&
-        `You must spawn the implementation-planner-max agent for medium to hard coding tasks. It will help you write the best possible code changes.`,
+        `Important: You must spawn the implementation-planner-max agent for all medium-to-hard coding tasks. It will help you come up with the best possible code changes.`,
 
       'If the users uses "@agent-id" or "@AgentName" in their message, you must spawn that agent. If you don\'t know what input parameters that agent expects, use the lookup_agent_info tool to get the agent metadata. Spawn all the agents that the user mentions.',
 
@@ -309,6 +309,9 @@ export const baseAgentUserInputPrompt = (
       (isLite || isFlash || isGeminiPro) &&
         `You must use the spawn_agents tool to spawn agents to help you complete the user request. You can spawn as many agents as you want. It is a good idea to spawn a file explorer agent first to explore the codebase. ${isLite ? '' : 'Finally, you must spawn the reviewer agent to review your code changes.'}`,
 
+      isMax &&
+        `Important: You must spawn the implementation-planner-max agent for all medium-to-hard coding tasks. It will help you come up with the best possible code changes.`,
+
       !isGPT5 &&
         'Finally, you must use the end_turn tool at the end of your response when you have completed the user request or want the user to respond to your message.',
 
diff --git a/.agents/researcher/sonnet/researcher-codebase-explorer-sonnet.ts b/.agents/researcher/sonnet/researcher-codebase-explorer-sonnet.ts
@@ -4,14 +4,10 @@ import researcherCodebaseExplorer from '../researcher-codebase-explorer'
 
 const definition: SecretAgentDefinition = {
   ...researcherCodebaseExplorer,
-  id: 'researcher-codebase-explorer-gpt-5',
+  id: 'researcher-codebase-explorer-sonnet',
   publisher,
-  displayName: 'Codebase Explorer GPT-5',
-  model: 'openai/gpt-5',
-  reasoningOptions: {
-    enabled: true,
-    effort: 'medium',
-  },
+  displayName: 'Codebase Explorer Sonnet',
+  model: 'anthropic/claude-sonnet-4.5',
 }
 
 export default definition
diff --git a/.agents/researcher/sonnet/researcher-docs-sonnet.ts b/.agents/researcher/sonnet/researcher-docs-sonnet.ts
@@ -4,13 +4,10 @@ import researcherDocs from '../researcher-docs'
 
 const definition: SecretAgentDefinition = {
   ...researcherDocs,
-  id: 'researcher-docs-gpt-5',
+  id: 'researcher-docs-sonnet',
   publisher,
-  displayName: 'Docs Researcher GPT-5',
-  model: 'openai/gpt-5',
-  reasoningOptions: {
-    effort: 'medium',
-  },
+  displayName: 'Docs Researcher Sonnet',
+  model: 'anthropic/claude-sonnet-4.5',
 }
 
 export default definition
diff --git a/.agents/researcher/sonnet/researcher-sonnet.ts b/.agents/researcher/sonnet/researcher-sonnet.ts
@@ -4,20 +4,15 @@ import researcher from '../researcher-grok-4-fast'
 
 const definition: SecretAgentDefinition = {
   ...researcher,
-  id: 'researcher-gpt-5',
+  id: 'researcher-sonnet',
   publisher,
-  displayName: 'Researcher GPT-5',
-  model: 'openai/gpt-5',
-  reasoningOptions: {
-    enabled: true,
-    effort: 'medium',
-  },
+  displayName: 'Researcher Sonnet',
+  model: 'anthropic/claude-sonnet-4.5',
 
   spawnableAgents: [
     'file-explorer',
-    'researcher-codebase-explorer-gpt-5',
-    'researcher-web-gpt-5',
-    'researcher-docs-gpt-5',
+    'researcher-web-sonnet',
+    'researcher-docs-sonnet',
   ],
 }
 
diff --git a/.agents/researcher/sonnet/researcher-web-sonnet.ts b/.agents/researcher/sonnet/researcher-web-sonnet.ts
@@ -4,13 +4,10 @@ import researcherWeb from '../researcher-web'
 
 const definition: SecretAgentDefinition = {
   ...researcherWeb,
-  id: 'researcher-web-gpt-5',
+  id: 'researcher-web-sonnet',
   publisher,
-  displayName: 'Web Researcher GPT-5',
-  model: 'openai/gpt-5',
-  reasoningOptions: {
-    effort: 'medium',
-  },
+  displayName: 'Web Researcher Sonnet',
+  model: 'anthropic/claude-sonnet-4.5',
 }
 
 export default definition
diff --git a/.agents/reviewer/reviewer-max.ts b/.agents/reviewer/reviewer-max.ts
@@ -1,14 +1,12 @@
 import { reviewer } from './reviewer-factory'
+import { publisher } from '../constants'
 
 import type { SecretAgentDefinition } from '../types/secret-agent-definition'
 
 const definition: SecretAgentDefinition = {
-  ...reviewer('openai/gpt-5'),
+  ...reviewer('anthropic/claude-sonnet-4.5'),
   id: 'reviewer-max',
-  reasoningOptions: {
-    enabled: true,
-    effort: 'medium',
-  },
+  publisher,
 }
 
 export default definition
diff --git a/.agents/web-researcher.ts b/.agents/web-researcher.ts
@@ -6,7 +6,7 @@ import type { SecretAgentDefinition } from './types/secret-agent-definition'
 const definition: SecretAgentDefinition = {
   id: 'web-researcher',
   publisher,
-  ...webResearcher('google/gemini-2.5-flash'),
+  ...webResearcher('x-ai/grok-4-fast'),
 }
 
 export default definition
diff --git a/backend/src/__tests__/xml-stream-parser.test.ts b/backend/src/__tests__/xml-stream-parser.test.ts
@@ -711,8 +711,9 @@ describe('processStreamWithTags', () => {
 
     expect(events).toEqual([
       {
-        name: undefined,
-        error: 'Tool not found: undefined',
+        name: 'parse_error',
+        error:
+          'Unknown tool undefined for tool call: {\n  "param1": "value1"\n}',
         type: 'error',
       },
     ])
diff --git a/backend/src/llm-apis/vercel-ai-sdk/ai-sdk.ts b/backend/src/llm-apis/vercel-ai-sdk/ai-sdk.ts
@@ -119,6 +119,7 @@ export const promptAiSdkStream = async function* (
     if (chunk.type !== 'text-delta') {
       const flushed = stopSequenceHandler.flush()
       if (flushed) {
+        content += flushed
         yield {
           type: 'text',
           text: flushed,
@@ -181,6 +182,7 @@ export const promptAiSdkStream = async function* (
 
       const stopSequenceResult = stopSequenceHandler.process(chunk.text)
       if (stopSequenceResult.text) {
+        content += stopSequenceResult.text
         yield {
           type: 'text',
           text: stopSequenceResult.text,
@@ -190,6 +192,7 @@ export const promptAiSdkStream = async function* (
   }
   const flushed = stopSequenceHandler.flush()
   if (flushed) {
+    content += flushed
     yield {
       type: 'text',
       text: flushed,
diff --git a/backend/src/main-prompt.ts b/backend/src/main-prompt.ts
@@ -205,7 +205,7 @@ export const mainPrompt = async (
     localAgentTemplates,
   })
 
-  logger.debug({ agentState }, 'Main prompt finished')
+  logger.debug({ agentState, output }, 'Main prompt finished')
 
   return {
     sessionState: {
diff --git a/backend/src/run-agent-step.ts b/backend/src/run-agent-step.ts
@@ -562,7 +562,18 @@ export const loopAgentSteps = async (
     while (true) {
       totalSteps++
       if (!checkLiveUserInput(userId, userInputId, clientSessionId)) {
-        break
+        logger.debug(
+          {
+            userId,
+            userInputId,
+            clientSessionId,
+            totalSteps,
+            runId,
+            agentState,
+          },
+          'User input no longer live (likely cancelled)',
+        )
+        throw new Error('User input no longer live (likely cancelled)')
       }
 
       const startTime = new Date()
diff --git a/backend/src/xml-stream-parser.ts b/backend/src/xml-stream-parser.ts
@@ -101,7 +101,10 @@ export async function* processStreamWithTags(
     }
 
     const toolName = parsedParams[toolNameParam] as keyof typeof processors
-    const processor = processors[toolName] ?? defaultProcessor(toolName)
+    const processor =
+      typeof toolName === 'string'
+        ? processors[toolName] ?? defaultProcessor(toolName)
+        : undefined
     if (!processor) {
       trackEvent(
         AnalyticsEvent.UNKNOWN_TOOL_CALL,
@@ -114,6 +117,11 @@ export async function* processStreamWithTags(
           autocompleted,
         },
       )
+      onError(
+        'parse_error',
+        `Unknown tool ${JSON.stringify(toolName)} for tool call: ${contents}`,
+      )
+      return
     }
 
     trackEvent(AnalyticsEvent.TOOL_USE, loggerOptions?.userId ?? '', {
diff --git a/common/src/old-constants.ts b/common/src/old-constants.ts
@@ -194,7 +194,7 @@ export const geminiModels = {
 export type GeminiModel = (typeof geminiModels)[keyof typeof geminiModels]
 
 export const openrouterModels = {
-  openrouter_claude_sonnet_4_5: 'anthropic/claude-4.5-sonnet',
+  openrouter_claude_sonnet_4_5: 'anthropic/claude-sonnet-4.5',
   openrouter_claude_sonnet_4: 'anthropic/claude-4-sonnet-20250522',
   openrouter_claude_opus_4: 'anthropic/claude-opus-4.1',
   openrouter_claude_3_5_haiku: 'anthropic/claude-3.5-haiku-20241022',
diff --git a/common/src/tools/params/tool/run-terminal-command.ts b/common/src/tools/params/tool/run-terminal-command.ts
@@ -45,7 +45,6 @@ export const runTerminalCommandParams = {
       process_type: z
         .enum(['SYNC', 'BACKGROUND'])
         .default('SYNC')
-        .optional()
         .describe(
           `Either SYNC (waits, returns output) or BACKGROUND (runs in background). Default SYNC`,
         ),
diff --git a/evals/git-evals/run-git-evals.ts b/evals/git-evals/run-git-evals.ts
@@ -141,7 +141,7 @@ You must decide whether to:
   - In this case, just put an empty string for next_prompt
 
 If deciding to continue, include a clear, focused prompt for Codebuff in next_prompt. Note that Codebuff does not have access to the spec, so you must describe the changes you want Codebuff to make in a way that is clear and concise.
-Explain your reasoning in detail.`,
+Explain your reasoning in detail. Do not ask Codebuff to git commit changes.`,
             },
           ],
           schema: AgentDecisionSchema,
@@ -273,14 +273,9 @@ function getCodebuffFileStates(
   evalCommitSha: string,
   projectPath: string,
 ): string {
-  // Stage all changes (including new files) before generating diff
-  execFileSync('git', ['add', '.'], {
-    cwd: projectPath,
-    stdio: ['ignore', 'pipe', 'pipe'],
-  })
-
-  // Get diff of staged files to include new files
-  return execFileSync('git', ['diff', '--staged'], {
+  // Get all changes since the commit before the target commit
+  // This captures both uncommitted changes and any commits made during the eval
+  return execFileSync('git', ['diff', `${evalCommitSha}^`], {
     cwd: projectPath,
     stdio: ['ignore', 'pipe', 'pipe'],
   }).toString()
diff --git a/npm-app/release/package.json b/npm-app/release/package.json
@@ -1,6 +1,6 @@
 {
   "name": "codebuff",
-  "version": "1.0.498",
+  "version": "1.0.500",
   "description": "AI coding agent",
   "license": "MIT",
   "bin": {
diff --git a/npm-app/src/__tests__/image-upload.test.ts b/npm-app/src/__tests__/image-upload.test.ts
@@ -253,5 +253,33 @@ describe('Image Upload Functionality', () => {
       expect(result.success).toBe(false)
       expect(result.error).toContain('Unsupported image format')
     })
+
+    test('should normalize unicode escape sequences in provided paths', async () => {
+      const actualFilename = 'Screenshot 2025-09-29 at 4.09.19 PM.png'
+      const filePath = path.join(TEST_DIR, actualFilename)
+      writeFileSync(filePath, MINIMAL_PNG)
+
+      const variations = [
+        'Screenshot 2025-09-29 at 4.09.19\\u{202f}PM.png',
+        'Screenshot 2025-09-29 at 4.09.19\\u202fPM.png',
+      ]
+
+      for (const candidate of variations) {
+        const result = await processImageFile(candidate, TEST_DIR)
+        expect(result.success).toBe(true)
+        expect(result.imagePart?.filename).toBe(actualFilename)
+      }
+    })
+
+    test('should handle shell-escaped characters in paths', async () => {
+      const spacedFilename = 'My Screenshot (Final).png'
+      const filePath = path.join(TEST_DIR, spacedFilename)
+      writeFileSync(filePath, MINIMAL_PNG)
+
+      const result = await processImageFile('My\\ Screenshot\\ \(Final\).png', TEST_DIR)
+
+      expect(result.success).toBe(true)
+      expect(result.imagePart?.filename).toBe(spacedFilename)
+    })
   })
 })
diff --git a/npm-app/src/agents/agent-utils.ts b/npm-app/src/agents/agent-utils.ts
@@ -1,5 +1,6 @@
 import * as fs from 'fs'
 import * as path from 'path'
+import * as os from 'os'
 
 import { getProjectRoot } from '../project-files'
 
@@ -42,3 +43,10 @@ export function getAllTsFiles(dir: string): string[] {
 export function getAgentsDirectory(): string {
   return path.join(getProjectRoot(), ...agentTemplatesSubdir)
 }
+
+/**
+ * Get the user's home agents directory path
+ */
+export function getUserAgentsDirectory(): string {
+  return path.join(os.homedir(), ...agentTemplatesSubdir)
+}
diff --git a/npm-app/src/agents/load-agents.ts b/npm-app/src/agents/load-agents.ts
diff --git a/npm-app/src/utils/image-handler.ts b/npm-app/src/utils/image-handler.ts
diff --git a/sdk/package.json b/sdk/package.json
diff --git a/sdk/src/run-state.ts b/sdk/src/run-state.ts
diff --git a/sdk/src/run.ts b/sdk/src/run.ts
diff --git a/web/src/app/docs/[category]/[slug]/page.tsx b/web/src/app/docs/[category]/[slug]/page.tsx
diff --git a/web/src/app/docs/[category]/page.tsx b/web/src/app/docs/[category]/page.tsx
diff --git a/web/src/app/docs/layout.tsx b/web/src/app/docs/layout.tsx
diff --git a/web/src/app/store/store-client.tsx b/web/src/app/store/store-client.tsx
diff --git a/web/src/components/docs/doc-sidebar.tsx b/web/src/components/docs/doc-sidebar.tsx

Original file line number	Diff line number	Diff line change
`@@ -6,7 +6,7 @@ import type { SecretAgentDefinition } from './types/secret-agent-definition'`
`6`	`6`	`const definition: SecretAgentDefinition = {`
`7`	`7`	`id: 'base',`
`8`	`8`	`publisher,`
`9`		`- ...base('anthropic/claude-4.5-sonnet', 'normal'),`
	`9`	`+ ...base('anthropic/claude-sonnet-4.5', 'normal'),`
`10`	`10`	`}`
`11`	`11`
`12`	`12`	`export default definition`
Original file line number	Diff line number	Diff line change
`@@ -6,7 +6,7 @@ import type { SecretAgentDefinition } from './types/secret-agent-definition'`
`6`	`6`	`const definition: SecretAgentDefinition = {`
`7`	`7`	`id: 'docs-researcher',`
`8`	`8`	`publisher,`
`9`		`- ...docsResearcher('google/gemini-2.5-flash'),`
	`9`	`+ ...docsResearcher('x-ai/grok-4-fast'),`
`10`	`10`	`}`
`11`	`11`
`12`	`12`	`export default definition`