Hey! Another quick question about a use case we're setting up, want to check what already

exists before we build around it.

We want to track what the market is saying about us: crawl a set of public

pages daily, land them in a knowledge base, and have an agent produce a weekly

trend summary. The knowledge base + agent + schedule trigger side looks like a

great fit already.

Three things we're unsure about:

1. Does the Web Crawler render JavaScript? A lot of the pages we care about are

SPAs where the content only appears after client-side routing, so a static

fetch gets an empty shell. Is there a headless-render mode we're missing?

2. The crawler seems to stay on the same host as the seed URL. Is that a

deliberate design decision? We'd love one connector with multiple seed

origins, or an allowlist of extra hosts, instead of one connector per site.

3. If we do the fetching ourselves, is POST /api/knowledge-files the right way

to push content into a knowledge base from an external pipeline? Or is that

more of an internal endpoint we shouldn't build on?

Happy to be told we're holding it wrong on any of these.

Is there anyway to prevent new models on a provider from being auto enabled? E.g. we use openrouter and when a new model is released, we don't want it to appear until we approve it. We have guardrails in Openrouter to prevent them from using the model, but it would be nice to not have to constantly hide them.

Could also add a switch to use the /models/user endpoint rather than /models that would return the filtered list from openrouter

Hey, RBAC question.

We run nested teams and want each team to self-manage its own agents/gateways/skills, with someone leading one team but a plain member in another.

Can't express that today: *:team-admin is scoped by team membership, not by team_member.role, so leading one team gives the same authority in every team you're in (and their parents).

#7807 looks like it replaces that model. Is that the direction, and where does it leave #5450?