Pete Miller

Yes probably a different file is neccessary because ConversationContext will be inside an iframe and AIChatContext will be on the outside (and AIChatContext and ConversationContext might each be split in...

This is the error format: ``` { "error": { "message": "The model `gpt-4-turbo-preview` does not exist or you do not have access to it.", "type": "invalid_request_error", "param": null, "code": "model_not_found"...

> it doesn't appear that we're exposing mojo interfaces on the JS side to any of the wallet endpoints But, shouldn't we? Looks to me like we're sending simple data...

> In general, I'm okay with this approach. However, I think I'd prefer to handle this at an API level - both of the use cases we have here are...

> My preferred approach (if feasible) would be via a separate command that does nothing else but run the type checks, similarly to npm run presubmit I don't think we...

@fallaciousreasoning I've "simplified" the bindings. Still have some browser tests todo but the problem seems fixed now. I had to add knowledge about the currently bound conversation's UUID to the...

It _does_ work in the full-page UI (so it's related to the `isStandalone` flag - either we allow the tab tracker to be bound in all UI or we disable...

> we won't be getting the max page content length until the first request completes we won't be getting ~the max page content length~ whether the content was truncated until...

Also the name `truncated_content_percentage` isn't obvious which way round it's referring to (90% was truncated or 90% was used) and technically the way it's being used here is the opposite...

This PR actually seems to change 2 somewhat separate areas and so I think needs another github issue attached: 1. Page truncation warning, including percentage of content truncated 2. Long...