Fix Eden page reload/crash when accessed via Tailscale#403
Fix Eden page reload/crash when accessed via Tailscale#403criztiano wants to merge 3 commits intobuilderz-labs:mainfrom
Conversation
…undary on ReactFlow and charts
criztiano
force-pushed
the
task/9871aa4e
branch
2 times, most recently
from
9ec6570 to
a26542a
Compare
Phase 1 (WebSocket): Already robust - ping/pong keepalive + backoff reconnection in websocket.ts Phase 2 (useSmartPoll): Add request deduplication (skip if in-flight), exponential backoff retry on error (3 attempts: 2s/4s/8s), error state return for UI affordances, proper retry cleanup on unmount/tab hide Phase 3 (useServerEvents): Add visibility-based SSE reconnect - when tab regains focus and EventSource is closed, reset backoff counter and reconnect immediately Phase 4 (Error Boundaries): Wrap additional recharts ResponsiveContainer instances in cost-tracker-panel with ErrorBoundary; ReactFlow in agent-network already wrapped Phase 5: task-board-panel fetchData interval 30s -> 60s to reduce connection pressure over Tailscale; SSE handles real-time task updates
criztiano
force-pushed
the
task/9871aa4e
branch
from
a26542a to
cef497b
Compare
ReviewThis PR adds agent workspace scaffolding files (AGENTS.md, SOUL.md, workspace-state.json, etc.) alongside the Eden page fix. These are separate concerns:
No CI runs have executed. Please rebase onto main and ensure |
- Re-added team-panel.tsx and all /api/team routes (lost during upstream merge) - Skills catalog with fixed taxonomy (9 categories, no creep) - Skill browser: category filter pills, tag-based search, active skills at top - Model selector: native select dropdown (PropertyChip removed — not in upstream) - Doc editor: textarea (BlockEditor removed — not in upstream) - skills-catalog.json with 112 categorized skills - skills-monitor.py: dumb cron that diffs disk vs catalog, files tasks for Roach - Daily cron at 9AM Rome
|
Thanks for the contribution! A few things before we can review this properly: Scope concerns — This is labeled as a fix for Eden page reload/crash via Tailscale, but the diff is +2,319 lines across 20 files and includes:
Could you:
Happy to review once the scope is narrowed down to the actual bug fix. |
0xNyk
left a comment
0xNyk
left a comment
There was a problem hiding this comment.
Review: DO NOT MERGE
Summary
This PR claims to fix an Eden page crash when accessed via Tailscale, but 95% of the changes are unrelated to that fix. It also commits a binary database file.
Critical Issues
-
Binary
eden.dbcommitted — A SQLite database file should never be in the repository. This must be removed and added to.gitignore. -
Massive scope creep — The PR contains:
- New
team-panel.tsx(1,392 lines) — entire new feature - 5 new API routes (
team/*) - Agent config artifacts (
AGENTS.md,SOUL.md,IDENTITY.md, etc.) - Changes to
use-smart-poll.tsandcost-tracker-panel.tsx
None of these are related to the Tailscale fix described in the PR title.
- New
-
No root cause explanation — It's unclear which specific change actually fixes the Eden page reload/crash. The PR description doesn't identify the root cause.
-
No tests, no CI validation — No test coverage for any of the changes.
Recommendation
Please close this PR and split into separate, focused PRs:
- PR-A: Tailscale fix only — identify the specific change that fixes the crash, with a clear root cause explanation
- PR-B: Team Panel feature — the new team panel as a standalone feature PR with tests
Each PR should have a clear description, passing CI, and appropriate test coverage.
|
Thanks for the crash recovery work! The ErrorBoundary and useSmartPoll hook from this PR are already on main ( |
|
ErrorBoundary + useSmartPoll already on main — no remaining work to cherry-pick. |
2 participants
Task: 9871aa4e