Slow Attachment and Frequent Disconnects

What website was slow?

The extension will show as loading until you start using the cli/MCP and the extension can connect with the websocket relay

What website was slow?

The extension will show as loading until you start using the cli/MCP and the extension can connect with the websocket relay

What website was slow?

The extension will show as loading until you start using the cli/MCP and the extension can connect with the websocket relay

What website was slow?

The extension will show as loading until you start using the cli/MCP and the extension can connect with the websocket relay

What website was slow?

The extension will show as loading until you start using the cli/MCP and the extension can connect with the websocket relay

1. It's app.crisp.chat, it's chat handling app (like intercom)

2. also sometimes it shows like this:

Additional Info:

1. It's app.crisp.chat, it's chat handling app (like intercom)

2. also sometimes it shows like this:

Additional Info:

1. It's app.crisp.chat, it's chat handling app (like intercom)

2. also sometimes it shows like this:

Additional Info:

1. It's app.crisp.chat, it's chat handling app (like intercom)

2. also sometimes it shows like this:

Additional Info:

@remorses Managed to get my first reliable session of the day with Playwriter.

I cloned the repo locally and built + installed unpacked extension, then had an agent debug.

Here's what it said fixes were:

1. I fixed CDP child-session detach handling in the extension relay logic in extension/src/background.ts. Specifically, Target.detachFromTarget now resolves the child sessionId from params.sessionId (the shape Playwright actually sends), maps it to the parent tab via childSessions, and no-ops unknown child sessions instead of throwing.
2. That removed the recurring relay error (No tab found for method Target.detachFromTarget ...) that was showing up right before your earlier Frame has been detached failures, and after this change page.goto(...) became stable.
3. I rebuilt the unpacked extension from sourc, so generated output in extension/dist/background.js matches source (not hand-edited drift).
4. I also added tab-group styling retries/paint nudges in extension/src/background.ts, but this appears cosmetic only; it did not correlate with command reliability.

Practical Bottom Line

The core reliability fix was the Target.detachFromTarget session routing change. The tab-group gray/unlabeled chip behavior is annoying UI paint behavior, but not what was breaking Playwriter execution.

@remorses Managed to get my first reliable session of the day with Playwriter.

I cloned the repo locally and built + installed unpacked extension, then had an agent debug.

Here's what it said fixes were:

1. I fixed CDP child-session detach handling in the extension relay logic in extension/src/background.ts. Specifically, Target.detachFromTarget now resolves the child sessionId from params.sessionId (the shape Playwright actually sends), maps it to the parent tab via childSessions, and no-ops unknown child sessions instead of throwing.
2. That removed the recurring relay error (No tab found for method Target.detachFromTarget ...) that was showing up right before your earlier Frame has been detached failures, and after this change page.goto(...) became stable.
3. I rebuilt the unpacked extension from sourc, so generated output in extension/dist/background.js matches source (not hand-edited drift).
4. I also added tab-group styling retries/paint nudges in extension/src/background.ts, but this appears cosmetic only; it did not correlate with command reliability.

Practical Bottom Line

The core reliability fix was the Target.detachFromTarget session routing change. The tab-group gray/unlabeled chip behavior is annoying UI paint behavior, but not what was breaking Playwriter execution.

@remorses Managed to get my first reliable session of the day with Playwriter.

I cloned the repo locally and built + installed unpacked extension, then had an agent debug.

Here's what it said fixes were:

1. I fixed CDP child-session detach handling in the extension relay logic in extension/src/background.ts. Specifically, Target.detachFromTarget now resolves the child sessionId from params.sessionId (the shape Playwright actually sends), maps it to the parent tab via childSessions, and no-ops unknown child sessions instead of throwing.
2. That removed the recurring relay error (No tab found for method Target.detachFromTarget ...) that was showing up right before your earlier Frame has been detached failures, and after this change page.goto(...) became stable.
3. I rebuilt the unpacked extension from sourc, so generated output in extension/dist/background.js matches source (not hand-edited drift).
4. I also added tab-group styling retries/paint nudges in extension/src/background.ts, but this appears cosmetic only; it did not correlate with command reliability.

Practical Bottom Line

The core reliability fix was the Target.detachFromTarget session routing change. The tab-group gray/unlabeled chip behavior is annoying UI paint behavior, but not what was breaking Playwriter execution.

@remorses Managed to get my first reliable session of the day with Playwriter.

I cloned the repo locally and built + installed unpacked extension, then had an agent debug.

Here's what it said fixes were:

1. I fixed CDP child-session detach handling in the extension relay logic in extension/src/background.ts. Specifically, Target.detachFromTarget now resolves the child sessionId from params.sessionId (the shape Playwright actually sends), maps it to the parent tab via childSessions, and no-ops unknown child sessions instead of throwing.
2. That removed the recurring relay error (No tab found for method Target.detachFromTarget ...) that was showing up right before your earlier Frame has been detached failures, and after this change page.goto(...) became stable.
3. I rebuilt the unpacked extension from sourc, so generated output in extension/dist/background.js matches source (not hand-edited drift).
4. I also added tab-group styling retries/paint nudges in extension/src/background.ts, but this appears cosmetic only; it did not correlate with command reliability.

Practical Bottom Line

The core reliability fix was the Target.detachFromTarget session routing change. The tab-group gray/unlabeled chip behavior is annoying UI paint behavior, but not what was breaking Playwriter execution.

@tyler-dot-earth thank you for diving deeper into the issue, can you open a PR with the changes the agent made so I can take a look? You can ask your agent to use the gh cli to do so

@tyler-dot-earth thank you for diving deeper into the issue, can you open a PR with the changes the agent made so I can take a look? You can ask your agent to use the gh cli to do so

@tyler-dot-earth thank you for diving deeper into the issue, can you open a PR with the changes the agent made so I can take a look? You can ask your agent to use the gh cli to do so

@tyler-dot-earth thank you for diving deeper into the issue, can you open a PR with the changes the agent made so I can take a look? You can ask your agent to use the gh cli to do so

@tyler-dot-earth thank you for diving deeper into the issue, can you open a PR with the changes the agent made so I can take a look? You can ask your agent to use the gh cli to do so

here's what it changed - https://github.com/remorses/playwriter/pull/59

@tyler-dot-earth thank you for diving deeper into the issue, can you open a PR with the changes the agent made so I can take a look? You can ask your agent to use the gh cli to do so

here's what it changed - https://github.com/remorses/playwriter/pull/59