fix: validate browser process before reporting launch success

[eureka928]

Summary

Fixes #1377 — POST /browser/login returned "success": true even when the Electron process exited immediately (e.g., sandbox misconfiguration, npx not found). The frontend (Cookies.tsx) also polled GET /browser/status which didn't exist, causing silent 404s.

Changes

Backend (tool_controller.py)

• Add _login_browser_process module-level variable to track the subprocess
• Add _wait_for_cdp_ready() helper that polls both process liveness and CDP port availability
• Rewrite open_browser_login() to validate process state instead of blind asyncio.sleep(3)
• Catch FileNotFoundError when npx is missing on PATH
• Use run_in_executor for blocking readline() to avoid stalling the event loop
• Add GET /browser/status endpoint the frontend already expects

Frontend (Cookies.tsx)

• Check response?.success instead of just response (object is truthy even when success: false)
• Show toast.error with the backend error message on failure

Before / After Proof

BEFORE (old behavior on main)

The old code blindly awaited asyncio.sleep(3) then returned success unconditionally:

# old code (main branch)
await asyncio.sleep(3)
return {
    "success": True,   # <-- always true, even if process crashed
    "pid": process.pid,
    ...
}

GET /browser/status did not exist → frontend polls got silent 404 errors.

AFTER (this PR)

Tested locally with uv run uvicorn main:api --port 5001:

=== TEST 1: GET /browser/status (no browser launched yet) ===
{"is_open":false}                                              ✅ New endpoint works

=== TEST 2: POST /browser/login (Electron crashes) ===
{"success":false,"error":"Browser process exited before CDP    ✅ Correctly reports failure
 became ready. Exit code: 1"}

=== TEST 3: GET /browser/status (after failed launch) ===
{"is_open":false}                                              ✅ Clean state after crash

=== TEST 4: POST /browser/login (retry) ===
{"success":false,"error":"Browser process exited before CDP    ✅ Consistent on retry
 became ready. Exit code: 1"}

Backend logs showing the process crash is detected and reported:

13:44:52 - tool_controller - INFO - [PROFILE USER LOGIN] Launching Electron browser with CDP on port 9223
13:44:52 - tool_controller - INFO - [ELECTRON OUTPUT] npm warn Unknown project config "shamefully-hoist"...
13:44:52 - tool_controller - INFO - [ELECTRON OUTPUT] [721291:...] FATAL:setuid_sandbox_host.cc(163)
  The SUID sandbox helper binary was found, but is not configured correctly...
13:44:53 - tool_controller - INFO - [ELECTRON OUTPUT] electron exited with signal SIGTRAP
13:44:53 - tool_controller - ERROR - [PROFILE USER LOGIN] Electron process exited with code: 1

Test plan

• cd backend && uv run pytest — 402 passed, 11 skipped
• TypeScript type-check — no new errors (pre-existing ones in skillsStore.ts only)
• Manual: POST /browser/login returns success: false when Electron is unavailable
• Manual: GET /browser/status returns {"is_open": false} when no browser is running
• Manual: Retry after failure returns consistent results
• Manual: With working Electron, success: true only after CDP port opens or process is confirmed alive

[eureka928]

@Wendong-Fan would you review this PR?
Thank you

[eureka928]

@fengju0213 would you review this please?
Thank you for your time

[nitpicker55555]

@eureka928 can you provide video for the problem and the video after your solution?