Version Packages

[github-actions]

This PR was opened by the Changesets release GitHub action. When you're ready to do a release, you can merge this and the packages will be published to npm automatically. If you're not ready to do a release yet, that's fine, whenever you add more changesets to main, this PR will be updated.

Releases

@browserbasehq/stagehand@3.5.0

Minor Changes

• #2127 78bcde8 Thanks @seanmcguire12! - Add ignoreDefaultArgs option to selectively remove chrome-launcher's built-in default flags (e.g. --disable-extensions) when running locally

Patch Changes

• #2107 8fc16d2 Thanks @miguelg719! - Fix Anthropic CUA triple_click action mapping.

• #2126 ebbdcd3 Thanks @seanmcguire12! - fix(core): import ToolSet from ai public export

• #2120 12703a6 Thanks @miguelg719! - Fix structuredOutputMode for newer Anthropic models

• #2116 cb586a1 Thanks @seanmcguire12! - include "[selected]" or "[checked]" state in snapshot

@browserbasehq/stagehand-evals@2.0.2

Patch Changes

• Updated dependencies [8fc16d2, 78bcde8, ebbdcd3, 12703a6, cb586a1]:
  • @browserbasehq/stagehand@3.5.0

@browserbasehq/stagehand-server-v3@3.6.5

Patch Changes

• Updated dependencies [8fc16d2, 78bcde8, ebbdcd3, 12703a6, cb586a1]:
  • @browserbasehq/stagehand@3.5.0

@browserbasehq/stagehand-server-v4@3.6.5

Patch Changes

• Updated dependencies [8fc16d2, 78bcde8, ebbdcd3, 12703a6, cb586a1]:
  • @browserbasehq/stagehand@3.5.0

[cubic-dev-ai]

No issues found across 9 files

Confidence score: 5/5

• Automated review surfaced no issues in the provided summaries.
• No files require special attention.

Architecture diagram

sequenceDiagram
    participant Client as User / Eval Script
    participant Server as Stagehand Server (v3/v4)
    participant Core as @browserbasehq/stagehand
    participant LLM as Anthropic API (CUA)
    participant Browser as Playwright / Browserbase

    Note over Client,Browser: Runtime flow for Anthropic Computer Use Actions (CUA)

    Client->>Server: POST /act (Instruction: "Triple click title")
    Server->>Core: stagehand.act()
    
    Core->>Core: Capture Screenshot & DOM
    Core->>LLM: Send state + Computer Use prompt
    
    LLM-->>Core: Response: tool_use (triple_click)
    
    Note right of Core: Logic fixed in this release
    Core->>Core: CHANGED: Map "triple_click" to browser-specific action
    
    Core->>Browser: mouse.click(x, y, { clickCount: 3 })
    Browser-->>Core: Click performed
    
    Core-->>Server: Action successful
    Server-->>Client: 200 OK (Task complete)

    alt Action Validation (Evals)
        Client->>Core: Run @browserbasehq/stagehand-evals
        Core->>LLM: Verify triple_click behavior
        LLM-->>Client: Pass/Fail
    end

Loading