[CANONICALIZATION 4/4] Move server into it's own package (Resolves STG-1047)

[pirate]

Stack Position

main
   └── pr/1-event-bus-infrastructure
         └── pr/2-api-schemas-openapi
               └── pr/3-session-store-p2p-server
                     └── pr/4-test-infrastructure  ← THIS PR

What

This moves the fastify server implementation into it's own package.

There's also a few cleanup lines to remove disableAPI from old test files and stuff.

[changeset-bot]

⚠️ No Changeset found

Latest commit: 09a20df

Merging this PR will not cause a version bump for any packages. If these changes should not result in a new version, you're good to go. If these changes should result in a version bump, you need to add a changeset.

This PR includes no changesets

When changesets are added to this PR, you'll see the packages that this PR includes changesets for and the associated semver types

Click here to learn what changesets are, and how to add one.

Click here if you're a maintainer who wants to add a changeset to this PR

[greptile-apps]

Greptile Overview

Greptile Summary

Adds comprehensive integration test infrastructure for P2P server testing with support for both local and remote test targets.

Key Changes

• Global setup (global-setup.stagehand.ts) manages local P2P server lifecycle during test runs
• Integration test suite (integration/integration.test.ts) covers session management, navigation, extraction, observation, action replay, and agent execution
• Test harness factory (integration/support/stagehandClient.ts) configures Stagehand clients for different test targets
• Vitest configuration updated with environment loading and global setup hooks
• Removed deprecated disableAPI option from existing test files

Critical Issues Found

• stagehandClient.ts: Contains two critical bugs that will cause runtime failures:
  1. Invalid model name openai/gpt-5-mini (should be openai/gpt-4o-mini)
  2. Missing return statement for remote target - function returns undefined for remote tests
• These bugs will prevent remote target tests from running successfully

Confidence Score: 1/5

• This PR has critical bugs that will cause runtime failures when tests are executed
• Score reflects two critical logic errors in stagehandClient.ts: invalid model name and missing return statement for remote target. These will cause immediate runtime errors when integration tests run, especially for remote targets. While the overall architecture is sound, these bugs must be fixed before merging.
• Critical attention needed for packages/core/tests/integration/support/stagehandClient.ts due to runtime-breaking bugs

Important Files Changed

File Analysis

Filename	Score	Overview
packages/core/tests/integration/support/stagehandClient.ts	1/5	Critical bugs: invalid model name gpt-5-mini and missing return for remote target causing runtime errors
packages/core/tests/integration/integration.test.ts	2/5	Integration test suite with runtime errors due to stagehandClient bugs, plus minor indentation inconsistency
packages/core/tests/global-setup.stagehand.ts	5/5	Vitest global setup for managing local/remote test servers, properly handles teardown

Sequence Diagram

sequenceDiagram
    participant Test as Integration Test
    participant Setup as Global Setup
    participant Client as Test Harness
    participant Server as P2P Server
    participant Stagehand as Stagehand Instance
    participant Browser as Browser Session

    Note over Setup,Server: Test Initialization (beforeAll)
    Setup->>Setup: Check STAGEHAND_TEST_TARGET env
    alt target === "local"
        Setup->>Stagehand: new Stagehand({env: "LOCAL"})
        Setup->>Stagehand: init()
        Setup->>Server: createServer({host, port})
        Setup->>Server: listen()
        Setup->>Setup: provide("STAGEHAND_BASE_URL", localUrl)
    else target === "remote"
        Setup->>Setup: provide("STAGEHAND_BASE_URL", remoteUrl)
    end

    Note over Test,Browser: Test Execution
    Test->>Client: createStagehandHarness(target)
    Client->>Client: inject("STAGEHAND_BASE_URL")
    Client->>Client: set process.env.STAGEHAND_API_URL
    Client->>Stagehand: new Stagehand({env: "BROWSERBASE"})
    Test->>Stagehand: init()
    Stagehand->>Server: POST /v1/start
    Server->>Browser: Create session
    Browser-->>Server: Session ID
    Server-->>Stagehand: Session details
    Stagehand-->>Test: Ready

    Note over Test,Browser: Test Operations
    Test->>Stagehand: page.goto(url)
    Stagehand->>Server: POST /v1/navigate
    Server->>Browser: Navigate
    Browser-->>Server: Response
    Server-->>Stagehand: Navigation result
    Stagehand-->>Test: Success

    Test->>Stagehand: extract(prompt, schema)
    Stagehand->>Server: POST /v1/extract
    Server->>Browser: Get page content
    Server->>Server: LLM processing
    Server-->>Stagehand: Extracted data
    Stagehand-->>Test: Structured result

    Test->>Stagehand: observe(instruction)
    Stagehand->>Server: POST /v1/observe
    Server->>Browser: Find elements
    Server->>Server: LLM analysis
    Server-->>Stagehand: Actions
    Stagehand-->>Test: Action list

    Test->>Stagehand: act(action)
    Stagehand->>Server: POST /v1/act
    Server->>Browser: Execute action
    Browser-->>Server: Result
    Server-->>Stagehand: Execution result
    Stagehand-->>Test: Success

    Note over Test,Browser: Test Cleanup
    Test->>Stagehand: close()
    Stagehand->>Server: POST /v1/end
    Server->>Browser: Terminate session
    Browser-->>Server: Closed
    Server-->>Stagehand: Cleanup complete

    Note over Setup,Server: Global Teardown
    Setup->>Server: server.close()
    Setup->>Stagehand: stagehand.close()

Loading

[greptile-apps]

_{9 files reviewed, 5 comments}

_{Edit Code Review Agent Settings | Greptile}

[cubic-dev-ai]

1 issue found across 9 files

Prompt for AI agents (all 1 issues)

Check if these issues are valid — if so, understand the root cause of each and fix them.


<file name="packages/core/tests/integration/support/stagehandClient.ts">

<violation number="1" location="packages/core/tests/integration/support/stagehandClient.ts:47">
P1: createStagehandHarness only returns when the target is local, so remote test runs receive `undefined` and immediately crash. Add logic that instantiates and returns a Stagehand client for the remote target instead of exiting this function without a value.</violation>
</file>

_{Reply to cubic to teach it or ask questions. Re-run a review with @cubic-dev-ai review this PR}