cli: opt-in telemetry categories + config echo

[Sayan-]

What changes

• kernel browsers create/update --telemetry:
  • --telemetry=all -> default category set
  • --telemetry=off -> disable
  • --telemetry=console,network -> capture exactly those categories (opt-in, replace semantics). The old name=on/off DSL is removed.
• settableCategories now covers all 9 (console, network, page, interaction, control, connection, system, screenshot, captcha). monitor is not settable (auto-managed, flows when a CDP category is captured).
• After create/update, the CLI echoes the categories telemetry will capture, so the effect of an opt-in selection is obvious.
• telemetry stream --categories filter set now covers every event category incl. monitor; event category is read from the SDK's typed Category field (removed the JSON/prefix fallback and the stale monitor->system mapping).

Test plan

• CI
• manual: --telemetry=all / =off / =console,network; confirm the echoed config; telemetry stream --categories monitor

---

Note

Medium Risk
Changes user-facing --telemetry semantics (breaking for scripts using name=on/off) and how stream filtering maps categories; behavior is localized to CLI with solid test coverage.

Overview
Reworks browser --telemetry on create/update to opt-in category selection with replace semantics: all (default set), off, or a comma list like console,network that captures only those categories. The old name=on/off DSL is removed.

Settable categories expand to nine (control, connection, system, screenshot, captcha, etc.); monitor stays non-settable but is included in telemetry stream --categories. After create/update, the CLI prints which categories telemetry will capture when --telemetry was used.

Streaming uses the SDK’s typed Category field (drops JSON/prefix fallback and monitor→system remapping). Tests and flag help text are updated for the new behavior.

^{Reviewed by Cursor Bugbot for commit ce2f532. Bugbot is set up for automated code reviews on this repo. Configure here.}

[firetiger-agent]

Created a monitoring plan for this PR.

What this PR does: Changes the --telemetry flag for browsers create and browsers update from a name=on|off format to a simple opt-in list (console,network). Expands the available categories from 4 to 9. After a create or update, the CLI now echoes which categories are capturing.

Intended effect:

• POST /browsers 400 rate: baseline ~0–14 errors/hr; confirmed if no new surge in 400s after release
• API telemetry error logs: baseline ~0–2 WARN/ERROR/hr mentioning "telemetry"; confirmed if stays near zero with no new patterns

Risks:

• Breaking change for scripted --telemetry users — any automation using the old network=on,page=off syntax will get a CLI-side parse error before reaching the API; alert if browser creation request volume drops >30% sustained for 2+ hours vs prior-week same-hour baseline
• Opt-in replaces partial update — old network=off sent a targeted disable; new network sends an opt-in-only replace; users expecting selective disables may silently enable unintended categories; alert if any user-reported unexpected telemetry behavior surfaces
• --categories=api / --categories=system stream filter removed — users streaming with these old category names will get a CLI validation error; no server-side signal, manual checkpoint on first user invocation with old filter

Status updates will be posted automatically on this PR as monitoring progresses.

View monitor