Skip to content

BOT-2328: feat(agents): give remote agents an Activity tab - #1330

Merged
klross merged 2 commits into
mainfrom
katieross/bot-2328-remote-agents-lost-their-activity-feed
Oct 20, 2026
Merged

klross merged 2 commits into
mainfrom
katieross/bot-2328-remote-agents-lost-their-activity-feed

Conversation

@klross

@klross klross commented Oct 19, 2026 •

Copy link
Copy Markdown
Contributor

Why?

Remote agents lost their activity feed in #1040. Since then an agent's Activity tab comes from its type, and the remote agent type never supplied one. So "View activity" on a remote agent lands on Settings or Info, and owners can't see what their remote agents are doing.

The remote agent type now supplies its own Activity tab, next to its Settings tab in adapter-type/, the same way Codex, Claude Code and Goose supply theirs. It shows the turns the agent published over the relay while Buzz has been open, starting with the conversation it was opened from, with a button to show all conversations. It reads the Agent Activity plugin's feed and says when that plugin is off. Nothing else in the app changes.

Activity tab showing a turn the agent published

The screenshot is from the browser test, where a test agent type registers the plugin's Activity tab.

Known limit: turns from before Buzz was opened don't show, since #1347 removed saved activity a couple hours ago. Remote agents have no local session files, so their history has to come from Beekeeper; that's BOT-2351.

Linear: BOT-2328

🤖 Generated with Claude Code

@klross klross changed the title BOT-2328: feat(agents): give every agent the app's Activity tab BOT-2328: feat(agents): give remote agents an Activity tab Oct 19, 2026
@klross
klross marked this pull request as ready for review October 19, 2026 21:39
@klross
klross requested review from a team as code owners October 19, 2026 21:39
@klross
klross force-pushed the katieross/bot-2328-remote-agents-lost-their-activity-feed branch from 227bf91 to ef65403 Compare October 20, 2026 00:14
@klross
klross marked this pull request as draft October 20, 2026 00:18
@klross
klross force-pushed the katieross/bot-2328-remote-agents-lost-their-activity-feed branch from 2e8d490 to d9c0feb Compare October 20, 2026 00:33
@klross

klross commented Oct 20, 2026 •

Copy link
Copy Markdown
Contributor Author

🤖 @salman1993 never mind on the revert; I closed #1384 and #1347 stays. This PR is live-only on main now, and remote history is tracked in BOT-2351.

@klross
klross marked this pull request as ready for review October 20, 2026 00:41
@klross
klross force-pushed the katieross/bot-2328-remote-agents-lost-their-activity-feed branch from d9c0feb to ba753d5 Compare October 20, 2026 00:46
@klross
klross changed the base branch from main to katieross/bot-2328-revert-1347 October 20, 2026 00:46
Since #1040 an agent's Activity tab comes from its type, and the remote
agent type never supplied one, so View activity on a remote agent landed
on Settings or Info.

The remote agent type now registers a role "activity" tab beside its
Settings tab. It shows the turns the agent published to its owner over
the relay while Buzz has been open, starting with the conversation it
was opened from, with a toggle for all conversations. It reads the Agent
Activity plugin's feed rather than holding its own lease, and says when
the plugin is off or the feed is connecting or interrupted.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Katie Ross <katieross@squareup.com>
@klross
klross force-pushed the katieross/bot-2328-remote-agents-lost-their-activity-feed branch from ba753d5 to 6a645e6 Compare October 20, 2026 00:59
@klross
klross changed the base branch from katieross/bot-2328-revert-1347 to main October 20, 2026 00:59

@wesbillman wesbillman left a comment

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

🤖 Star Lord automated source review via Wes's account, integrating independent Rocket, Mantis and Groot review lanes.

Reviewed head 6a645e6 against base ac42eb6 (feature merge base c2c7f6f).

Two low-severity recovery issues are commented inline. The existing feed remains the lifecycle and retention owner; the documented session-only history limitation is accepted for this scope. Assessment: 8/10 until the recovery messages/control behavior below are corrected; these are non-blocking comments, not a merge verdict.

Separate public-material finding: this public PR description includes internal issue-tracker links in the Known limit paragraph and final Linear line. Please replace those links with a public description or public issue reference. The URLs expose internal workspace/issue identifiers even if their destinations require authentication. I did not access the private issues. The attached browser-fixture screenshot was inspected; no additional disclosure finding in that image.

Validation: source-only review of all ten feature files and relevant callers, session ownership, scope filtering, recovery and focus transitions. No tests, builds, app launches, live remote-agent workflow, runtime performance measurement or CI assessment. The browser test source mounts this tab on a fixture type; it does not establish live Builderlab integration. Native behavior, rendered accessibility, both themes and human acceptance remain unverified.

Comment thread src/bundled/builderlab/adapter-type/Activity.tsx
Comment thread src/bundled/builderlab/adapter-type/Activity.tsx
turnId: "published",
payload,
});
app.observer(frame("turn_started", {}), key);

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

🤖 Could we wait for app.relay.hasRoute("primary", "observer") before publishing these frames? Activity delivery is best effort, so publishing before the subscription is ready can legitimately lose them. Waiting here would make the test deterministic without changing the product’s delivery behavior.

Copy link
Copy Markdown
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

🤖 2286526 waits for the observer route before publishing.

@cloud-on-prem cloud-on-prem left a comment

Copy link
Copy Markdown

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

🤖 Approved at Prem’s direction.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Signed-off-by: Katie Ross <katieross@squareup.com>

@wesbillman wesbillman left a comment

Copy link
Copy Markdown
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

🤖 Star Lord automated source follow-up review via Wes's account.

Head 2286526; base ac42eb6. Compared the complete endpoint trees with previously reviewed 6a645e6: only a comment and observer-route readiness assertion were added in tests/browser/local-agent-activity.spec.mjs. Production code and all prior feature patches are unchanged, so this uses the small unchanged-feature solo follow-up exception.

No new actionable finding in this follow-up (9/10). The retrying assertion waits for the actual owner-only observer subscription before sending best-effort activity, preserves the existing delivery/UI assertions, and adds no browser cases or fixed delay.

The two recovery/focus comments and public-description finding in the previous review remain unresolved and are not duplicated here. The feature's prior 8/10 assessment therefore remains. I rechecked the public description and attached fixture screenshot; no additional image disclosure finding.

Validation is source-only: pinned test, fixture subscription/delivery helpers, unchanged recovery paths and relevant guidance. No tests, builds, app launches or live remote-agent execution. A single hosted-check snapshot showed one WebKit shard running and Windows native validation skipped; it does not establish complete CI success. Runtime timing, rendered failure/retry focus, accessibility and native integration remain unverified. This is a non-blocking COMMENT, not approval or merge authorization.

@klross
klross merged commit 3e7d5be into main Oct 20, 2026
67 of 69 checks passed
@klross
klross deleted the katieross/bot-2328-remote-agents-lost-their-activity-feed branch October 20, 2026 03:14
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

4 participants