fix: OpenClaw 2026.5.12 chat protocol + Free→Paid tier auto-detect + tier-change UX + codex plugin auto-install - #132
Merged
yalexx merged 4 commits intoMay 16, 2026
Conversation
…upgrade UX - ChatApp/ChatPopup: bump WebSocket connect handshake to protocol v4 (2026.5.12 rejects v3 with code=1002), unbreaks the in-page chat - status/route.ts: remove the localTier !== null guard so Free→Paid portal upgrades are visible without re-login; mapPortalTier now gates non-null returns on a paid subscription plan, defending the stale-deviceTier case the old guard worked around - TierUpgradeCelebration: announce upgrade (Free→Pro→Max) once each via client-kv flag, plus a Free-plan downgrade notice with a Re-subscribe CTA; mounted in page.tsx - desktop-translations*: 12 new keys × 10 locales - status.test.ts: replace 1 test (skip-portal-when-null) with 3 new ones (Free portal query, Free→Pro upgrade detection, bogus deviceTier defence)
…el is configured OpenClaw 2026.5.12 split the codex agent harness out of the core gateway into a separate npm package (`@openclaw/codex`) and only auto-installs it during `openclaw onboard --auth-choice openai-codex…`. ClawBox's configure route writes `openclaw.json` directly (see the schema-drift note in `src/app/setup-api/ai-models/configure/route.ts`), so devices that pick a Codex model in the chat picker never trigger the install and every chat attempt fails with: Embedded agent failed before reply: Requested agent harness "codex" is not registered. This is the exact failure surfaced in #(this PR's issue): a device on 2026.5.12 with `agents.defaults.model.primary = "openai-codex/gpt-5.5"` errors on the first reply. Fix in `scripts/gateway-pre-start.sh` because it's the single chokepoint that runs on every gateway start — self-healing for existing devices on upgrade, fresh installs, and manual config edits. Detection mirrors OpenClaw's own `modelSelectionShouldEnsureCodexPlugin` (provider === "openai-codex" on the primary model OR any auth profile). Idempotent: skips the install when `@openclaw/codex/package.json` is already on disk. Verified on Jetson: 1. `openclaw plugins uninstall codex --force` 2. `systemctl restart clawbox-gateway` 3. Pre-start detects codex model + missing plugin → installs. 4. Gateway comes up with `codex` in its loaded plugins list.
|
Warning Rate limit exceeded
You’ve run out of usage credits. Purchase more in the billing tab. ⌛ How to resolve this issue?After the wait time has elapsed, a review can be triggered using the We recommend that you space out your commits to avoid hitting the rate limit. 🚦 How do rate limits work?CodeRabbit enforces hourly rate limits for each developer per organization. Our paid plans have higher rate limits than the trial, open-source and free plans. In all cases, we re-allow further reviews after a brief timeout. Please see our FAQ for further information. ℹ️ Review info⚙️ Run configurationConfiguration used: Path: .coderabbit.yaml Review profile: ASSERTIVE Plan: Pro Run ID: 📒 Files selected for processing (2)
📝 WalkthroughWalkthroughThis PR introduces a tier upgrade/downgrade celebration modal UI, refines portal tier resolution logic to always query when a token exists, upgrades WebSocket protocol to v4 across client connections, and adds automatic Codex plugin detection and installation to the gateway startup script. ChangesTier Upgrade Celebration UI
Portal Tier Resolution Logic
WebSocket Protocol v4 Upgrade
Gateway Codex Plugin Auto-Installation
Playwright and tests
Estimated code review effort🎯 3 (Moderate) | ⏱️ ~20 minutes Possibly related PRs
Suggested labels
Suggested reviewers
Poem
🚥 Pre-merge checks | ✅ 4 | ❌ 1❌ Failed checks (1 warning)
✅ Passed checks (4 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. ✨ Finishing Touches🧪 Generate unit tests (beta)
Tip 💬 Introducing Slack Agent: The best way for teams to turn conversations into code.Slack Agent is built on CodeRabbit's deep understanding of your code, so your team can collaborate across the entire SDLC without losing context.
Built for teams:
One agent for your entire SDLC. Right inside Slack. Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 1
Caution
Some comments are outside the diff and can’t be posted inline due to platform limitations.
⚠️ Outside diff range comments (1)
src/tests/routes/ai-models/status.test.ts (1)
382-406:⚠️ Potential issue | 🟠 Major | ⚡ Quick winStale test contradicts the new portal lookup behavior.
This test expects
fetchSpy.not.toHaveBeenCalled()when aclawaiTokenis present withlocalTier=null. However, the route change at line 258 now queries the portal wheneverclawaiTokenexists, regardless oflocalTier. The test at lines 180-201 confirms this new behavior.The comment on lines 383-385 referencing "defence-in-depth keeps the portal call skipped" describes the old behavior that this PR removes.
Proposed fix
it("returns clawaiAccountTier=null but clawaiConfigured=true for a Free user chatting via OpenAI", async () => { - // Free user with a paired clawai token but no paid local picker - // → defence-in-depth keeps the portal call skipped, so - // clawaiAccountTier stays null. clawaiConfigured is true so the - // hook reports loggedIn=true (Free users have a paired account). + // Free user with a paired clawai token — portal is queried and + // returns Free, so clawaiAccountTier stays null. clawaiConfigured + // is true so the hook reports loggedIn=true (Free users have a + // paired account). mockReadConfig.mockResolvedValue({ auth: { profiles: { "openai:default": { provider: "openai", mode: "token" }, "deepseek:default": { provider: "deepseek", mode: "api_key" }, }, }, agents: { defaults: { model: { primary: "openai/gpt-5" } } }, models: { providers: { deepseek: { apiKey: "claw_test456" } } }, } as never); mockGetConfigValue.mockResolvedValue(null); + fetchSpy.mockResolvedValue(new Response( + JSON.stringify({ tier: "free", deviceTier: null }), + { status: 200 }, + )); const res = await GET(); const body = await res.json(); expect(body.clawaiTier).toBeNull(); expect(body.clawaiAccountTier).toBeNull(); expect(body.clawaiConfigured).toBe(true); - expect(fetchSpy).not.toHaveBeenCalled(); + expect(fetchSpy).toHaveBeenCalledTimes(1); });🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the rest with a brief reason, keep changes minimal, and validate. In `@src/tests/routes/ai-models/status.test.ts` around lines 382 - 406, The test's comment and assertion still assert the old behavior that the portal call is skipped; update the test in the case using mockReadConfig / mockGetConfigValue (the Free user with a paired clawai token and localTier=null) to reflect the new route behavior: remove or rewrite the "defence-in-depth keeps the portal call skipped" comment and change the assertion from expect(fetchSpy).not.toHaveBeenCalled() to expect(fetchSpy).toHaveBeenCalled() (the GET() route now queries the portal whenever a clawaiToken exists regardless of localTier).
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@scripts/gateway-pre-start.sh`:
- Around line 309-318: The code should defensively ensure cfg["auth"] is a dict
before calling .get to avoid AttributeError; change how profiles is derived so
it uses the existing pattern (e.g., auth = cfg.get("auth"); use auth if
isinstance(auth, dict) else {}), then set profiles = auth.get("profiles", {}) or
{}; keep the rest of the uses_codex logic but rely on profiles being a dict so
profiles.items() is safe and reference the variables primary, profiles, and
uses_codex when making the change.
---
Outside diff comments:
In `@src/tests/routes/ai-models/status.test.ts`:
- Around line 382-406: The test's comment and assertion still assert the old
behavior that the portal call is skipped; update the test in the case using
mockReadConfig / mockGetConfigValue (the Free user with a paired clawai token
and localTier=null) to reflect the new route behavior: remove or rewrite the
"defence-in-depth keeps the portal call skipped" comment and change the
assertion from expect(fetchSpy).not.toHaveBeenCalled() to
expect(fetchSpy).toHaveBeenCalled() (the GET() route now queries the portal
whenever a clawaiToken exists regardless of localTier).
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Path: .coderabbit.yaml
Review profile: ASSERTIVE
Plan: Pro
Run ID: 75f905d8-e57d-4913-a053-c8939b47473c
📒 Files selected for processing (11)
scripts/gateway-pre-start.shsrc/app/page.tsxsrc/app/setup-api/ai-models/status/route.tssrc/components/ChatApp.tsxsrc/components/ChatPopup.tsxsrc/components/TierUpgradeCelebration.tsxsrc/lib/desktop-translations-part1.tssrc/lib/desktop-translations-part2.tssrc/lib/desktop-translations-part3.tssrc/lib/desktop-translations.tssrc/tests/routes/ai-models/status.test.ts
Comment on lines
+309
to
+318
| primary = (cfg.get("agents", {}).get("defaults", {}).get("model", {}) or {}).get("primary") or "" | ||
| profiles = cfg.get("auth", {}).get("profiles", {}) or {} | ||
| uses_codex = ( | ||
| isinstance(primary, str) and primary.lower().startswith("openai-codex/") | ||
| ) or any( | ||
| (isinstance(k, str) and k.lower().startswith("openai-codex:")) or | ||
| (isinstance(v, dict) and isinstance(v.get("provider"), str) | ||
| and v["provider"].lower() == "openai-codex") | ||
| for k, v in profiles.items() | ||
| ) |
There was a problem hiding this comment.
Add defensive isinstance check for auth to match existing pattern.
Line 310 will raise AttributeError if cfg["auth"] exists but is not a dict (e.g., None or a corrupted value), since the exception handler only catches FileNotFoundError and JSONDecodeError. This would silently skip the codex install, causing confusing "harness not registered" errors.
The existing code at line 131 uses the defensive pattern—this should match:
Proposed fix
primary = (cfg.get("agents", {}).get("defaults", {}).get("model", {}) or {}).get("primary") or ""
-profiles = cfg.get("auth", {}).get("profiles", {}) or {}
+auth = cfg.get("auth")
+profiles = auth.get("profiles", {}) if isinstance(auth, dict) else {}
+profiles = profiles if isinstance(profiles, dict) else {}
uses_codex = (As per coding guidelines: "Review for proper error handling" — the current pattern is inconsistent with line 131's defensive approach.
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@scripts/gateway-pre-start.sh` around lines 309 - 318, The code should
defensively ensure cfg["auth"] is a dict before calling .get to avoid
AttributeError; change how profiles is derived so it uses the existing pattern
(e.g., auth = cfg.get("auth"); use auth if isinstance(auth, dict) else {}), then
set profiles = auth.get("profiles", {}) or {}; keep the rest of the uses_codex
logic but rely on profiles being a dict so profiles.items() is safe and
reference the variables primary, profiles, and uses_codex when making the
change.
yalexx
approved these changes
May 16, 2026
There was a problem hiding this comment.
Actionable comments posted: 3
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@playwright.config.ts`:
- Around line 3-4: The parsed port value assigned to the variable port can be
NaN and is then used to build baseURL and in webServer.command; validate port
after parsing (e.g., Number(...) stored in portRaw) and ensure
Number.isFinite(port) && Number.isInteger(port) && port > 0, otherwise set port
to a safe default (3100) or throw a clear error; then use the validated port
variable when constructing baseURL and in webServer.command (referencing the
port and baseURL identifiers).
In `@src/components/TierUpgradeCelebration.tsx`:
- Around line 188-193: The dialog overlay in TierUpgradeCelebration lacks
keyboard dismissal and a description link for assistive tech; update the
component to (1) generate stable IDs (useId or similar) for titleId and a new
descriptionId and add aria-describedby={descriptionId} on the root dialog, (2)
add a useEffect that registers a keydown listener to call onClose when Escape is
pressed and cleans up on unmount, and (3) change the overlay onClick to only
call onClose when the click target equals the overlay (e.target ===
e.currentTarget) so inner content clicks don't close it; ensure the description
element uses the descriptionId and avoid injecting unescaped HTML (use safe
text) to prevent XSS.
- Around line 74-120: The component TierUpgradeCelebration currently uses local
useState for dialog lifecycle; replace that with the shared window reducer
pattern from useWindows.ts: remove useState<DialogState|null> dialog and
setDialog usage and instead register this modal with the useWindows reducer
(import useWindows and the window action creators), dispatch the "open" action
when you need to show the upgrade/downgrade dialog (use the same payload shape {
kind: "upgrade", tier } or { kind: "downgrade-free" }) and dispatch the "close"
action in onClose (and update SEEN_KEY via the reducer close handler or via the
close action side-effect). Ensure rankOf/tier/loading logic still computes when
to dispatch open (replace setDialog calls with dispatch open) and keep
kv.set(SEEN_KEY, ...) behavior triggered on close via the window close handler
or by dispatching an action that includes the seen value; reference
TierUpgradeCelebration, useClawboxLogin, SEEN_KEY, FREE_SEEN_VALUE, rankOf and
integrate with the existing useWindows reducer API for
open/minimize/maximize/close semantics.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Path: .coderabbit.yaml
Review profile: ASSERTIVE
Plan: Pro
Run ID: 1b6be83b-5e3b-487a-8077-0ed3c4c02a72
📒 Files selected for processing (4)
playwright.config.tssrc/components/TierUpgradeCelebration.tsxsrc/tests/routes/ai-models/status.test.tssrc/tests/unit/translations.test.ts
Comment on lines
+3
to
+4
| const port = Number(process.env.PLAYWRIGHT_PORT || process.env.PORT || 3100); | ||
| const baseURL = `http://localhost:${port}`; |
There was a problem hiding this comment.
Validate parsed port before using it in URL/command.
On Line 3, Number(...) can produce NaN, which then breaks both baseURL and webServer.command (Line 42-43). Add a finite integer check and fallback/error.
Proposed fix
-const port = Number(process.env.PLAYWRIGHT_PORT || process.env.PORT || 3100);
+const rawPort = process.env.PLAYWRIGHT_PORT ?? process.env.PORT;
+const parsedPort = rawPort ? Number.parseInt(rawPort, 10) : 3100;
+const port = Number.isInteger(parsedPort) && parsedPort > 0 ? parsedPort : 3100;
const baseURL = `http://localhost:${port}`;Also applies to: 42-43
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@playwright.config.ts` around lines 3 - 4, The parsed port value assigned to
the variable port can be NaN and is then used to build baseURL and in
webServer.command; validate port after parsing (e.g., Number(...) stored in
portRaw) and ensure Number.isFinite(port) && Number.isInteger(port) && port > 0,
otherwise set port to a safe default (3100) or throw a clear error; then use the
validated port variable when constructing baseURL and in webServer.command
(referencing the port and baseURL identifiers).
Comment on lines
+74
to
+120
| export default function TierUpgradeCelebration() { | ||
| const { tier, loading } = useClawboxLogin(); | ||
| const { t } = useT(); | ||
| const [dialog, setDialog] = useState<DialogState | null>(null); | ||
|
|
||
| useEffect(() => { | ||
| if (loading) return; | ||
| const seen = kv.get(SEEN_KEY); | ||
| const currentSeenValue = tier ?? FREE_SEEN_VALUE; | ||
|
|
||
| // First observation on this browser/device is a baseline, not a | ||
| // transition. Without this guard, already-paid accounts see the | ||
| // celebration every time this feature reaches a fresh client cache. | ||
| if (seen === null) { | ||
| kv.set(SEEN_KEY, currentSeenValue); | ||
| return; | ||
| } | ||
|
|
||
| const currentRank = rankOf(tier); | ||
| const seenRank = rankOf(seen); | ||
|
|
||
| // Upgrade to a paid tier we haven't celebrated yet. | ||
| if (currentRank > seenRank && (tier === "flash" || tier === "pro")) { | ||
| setDialog({ kind: "upgrade", tier }); | ||
| return; | ||
| } | ||
| // Downgrade from any paid tier to Free. | ||
| if (currentRank === 0 && seenRank > 0) { | ||
| setDialog({ kind: "downgrade-free" }); | ||
| return; | ||
| } | ||
| // Intermediate downgrade (Max → Pro) or no-change tick: silently | ||
| // sync `seen` so a later climb back to the same tier doesn't re-fire | ||
| // the celebration the user already saw. | ||
| if (seen !== currentSeenValue) kv.set(SEEN_KEY, currentSeenValue); | ||
| }, [tier, loading]); | ||
|
|
||
| if (!dialog) return null; | ||
|
|
||
| const onClose = () => { | ||
| if (dialog.kind === "upgrade") { | ||
| kv.set(SEEN_KEY, dialog.tier); | ||
| } else { | ||
| kv.set(SEEN_KEY, FREE_SEEN_VALUE); | ||
| } | ||
| setDialog(null); | ||
| }; |
There was a problem hiding this comment.
🛠️ Refactor suggestion | 🟠 Major | 🏗️ Heavy lift
Use the shared useWindows.ts reducer pattern for this modal’s lifecycle state.
dialog open/close is managed locally with useState, which diverges from the required window-state reducer pattern for component windows/modals.
As per coding guidelines, src/components/**/*.tsx: Use the useWindows.ts reducer pattern for managing window state (open, minimize, maximize, close, focus).
🧰 Tools
🪛 ESLint
[error] 97-97: Error: Calling setState synchronously within an effect can trigger cascading renders
Effects are intended to synchronize state between React and external systems such as manually updating the DOM, state management libraries, or other platform APIs. In general, the body of an effect should do one or both of the following:
- Update external systems with the latest state from React.
- Subscribe for updates from some external system, calling setState in a callback function when external state changes.
Calling setState synchronously within an effect body causes cascading renders that can hurt performance, and is not recommended. (https://react.dev/learn/you-might-not-need-an-effect).
/home/jailuser/git/src/components/TierUpgradeCelebration.tsx:97:7
95 | // Upgrade to a paid tier we haven't celebrated yet.
96 | if (currentRank > seenRank && (tier === "flash" || tier === "pro")) {
97 | setDialog({ kind: "upgrade", tier });
| ^^^^^^^^^ Avoid calling setState() directly within an effect
98 | return;
99 | }
100 | // Downgrade from any paid tier to Free.
(react-hooks/set-state-in-effect)
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@src/components/TierUpgradeCelebration.tsx` around lines 74 - 120, The
component TierUpgradeCelebration currently uses local useState for dialog
lifecycle; replace that with the shared window reducer pattern from
useWindows.ts: remove useState<DialogState|null> dialog and setDialog usage and
instead register this modal with the useWindows reducer (import useWindows and
the window action creators), dispatch the "open" action when you need to show
the upgrade/downgrade dialog (use the same payload shape { kind: "upgrade", tier
} or { kind: "downgrade-free" }) and dispatch the "close" action in onClose (and
update SEEN_KEY via the reducer close handler or via the close action
side-effect). Ensure rankOf/tier/loading logic still computes when to dispatch
open (replace setDialog calls with dispatch open) and keep kv.set(SEEN_KEY, ...)
behavior triggered on close via the window close handler or by dispatching an
action that includes the seen value; reference TierUpgradeCelebration,
useClawboxLogin, SEEN_KEY, FREE_SEEN_VALUE, rankOf and integrate with the
existing useWindows reducer API for open/minimize/maximize/close semantics.
Comment on lines
+188
to
+193
| <div | ||
| className="fixed inset-0 z-[100001] flex items-center justify-center bg-black/70 backdrop-blur-sm p-4" | ||
| role="dialog" | ||
| aria-modal="true" | ||
| aria-labelledby={titleId} | ||
| onClick={onClose} |
There was a problem hiding this comment.
Add Escape dismissal and dialog description linkage for keyboard/screen-reader completeness.
♿ Suggested accessibility patch
return (
<div
className="fixed inset-0 z-[100001] flex items-center justify-center bg-black/70 backdrop-blur-sm p-4"
role="dialog"
aria-modal="true"
aria-labelledby={titleId}
+ aria-describedby={`${titleId}-description`}
+ onKeyDown={(e) => {
+ if (e.key === "Escape") onClose();
+ }}
onClick={onClose}
>
@@
- <p className="text-sm text-white/65 leading-relaxed">{body}</p>
+ <p id={`${titleId}-description`} className="text-sm text-white/65 leading-relaxed">{body}</p>As per coding guidelines, src/components/**: React 19 components with Tailwind CSS v4. Review for accessibility, proper state management, and XSS prevention.
Also applies to: 215-219
🤖 Prompt for AI Agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
In `@src/components/TierUpgradeCelebration.tsx` around lines 188 - 193, The dialog
overlay in TierUpgradeCelebration lacks keyboard dismissal and a description
link for assistive tech; update the component to (1) generate stable IDs (useId
or similar) for titleId and a new descriptionId and add
aria-describedby={descriptionId} on the root dialog, (2) add a useEffect that
registers a keydown listener to call onClose when Escape is pressed and cleans
up on unmount, and (3) change the overlay onClick to only call onClose when the
click target equals the overlay (e.target === e.currentTarget) so inner content
clicks don't close it; ensure the description element uses the descriptionId and
avoid injecting unescaped HTML (use safe text) to prevent XSS.
KrasimirKralev
added a commit
that referenced
this pull request
May 17, 2026
… path, useReducer + a11y (#134) Follow-ups to PR #132 implementing CodeRabbit's four review prompts on our code. scripts/gateway-pre-start.sh - Guard `profiles.items()` behind an isinstance check (#132 review on line 318): `cfg["auth"]` may be `None` or a corrupted scalar on a hand-edited config, which would have raised AttributeError on the previous `.get("profiles", {})` call and silently skipped the codex install. Now matches the same defensive pattern used at line 131 for openrouter. - Derive `CODEX_PLUGIN_DIR` from `$(dirname "$OPENCLAW_CONFIG")` instead of hardcoding `/home/clawbox/.openclaw/...` (#133 review on line 323). Works for non-default clawbox users / per-user installs without changing behaviour on the default Jetson layout. src/components/TierUpgradeCelebration.tsx - Switched from `useState` to a local `useReducer` (#132 review on line 120). Silences the ESLint `set-state-in-effect` warning and matches the spirit of the "reducer pattern" coding guideline. Deliberately NOT routed through `useWindows.ts` — that hook is for desktop windows with z-order / minimize / maximize, none of which apply to a transient centred modal, and forcing a fake appId / icon / defaultWidth through it would be worse than the status quo. `ClawBoxLoginModal.tsx` follows the same local-state shape and was not flagged. - Accessibility improvements (#132 review on line 193): * `useId()` for stable, collision-free title + description IDs — drops the hand-rolled "tier-upgrade-title" / "tier-downgrade-title" strings and the `titleId` field on the CONTENT table. * Added `aria-describedby` linking the body paragraph. * `Escape` keydown handler with cleanup on unmount. * Backdrop click guarded with `e.target === e.currentTarget` instead of relying on the inner card's `stopPropagation`. Smoke-tested on Jetson: - `openclaw plugins uninstall codex --force` → `systemctl restart clawbox-gateway` → pre-start detects the missing plugin via the dynamically-resolved path, installs it, gateway comes up with `codex` in its loaded-plugins list.
3 tasks
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Four related fixes that unblock the in-page chat and tighten the
tier-state UX after upgrading to OpenClaw 2026.5.12.
1. Chat protocol v3 → v4 (fix)
OpenClaw 2026.5.12 dropped support for the v3 gateway protocol and
rejects v3 connect frames with
code=1002 reason=protocol mismatch.The in-page chat panel was sending
minProtocol: 3, maxProtocol: 3and rendered "Could not connect to gateway" in the UI. Bumped both
chat surfaces to v4.
src/components/ChatApp.tsxsrc/components/ChatPopup.tsx2. Free → Paid portal upgrade auto-detection (fix)
/setup-api/ai-models/statuswas short-circuiting the portal callwhenever the local picker tier was
null, so a Free user whoupgraded on the portal didn't see the new tier until they re-paired
ClawBox AI. The original guard's intent — preventing a portal-side
deviceTier="flash"bug from promoting a Free user — is nowenforced inside
mapPortalTier, which requires a paid subscriptiontierbefore honouring anydeviceTierstamp. The guard on thecall-site is removed, so the next 30 s poll (worst case ~2.5 min
after the portal's own 60 s reconcile + 120 s device-side cache TTL)
flips the device without a re-login.
src/app/setup-api/ai-models/status/route.tssrc/tests/routes/ai-models/status.test.ts— 1 test replaced(skip-portal-when-null no longer applies), 2 new tests added
(Free→Pro upgrade detection, bogus deviceTier defence)
3. Tier-change announcement modal (feat)
A new
TierUpgradeCelebrationmodal mounted on the desktop root.Watches
useClawboxLogin()and announces transitions:Re-subscribe CTA
Each transition is announced once via a
clawai_tier_seenflagstored through
client-kv, so a poll re-reporting the same tierdoesn't re-open the dialog. A real cancel + resubscribe flow does
re-celebrate, because the downgrade resets the flag.
src/components/TierUpgradeCelebration.tsx(new)src/app/page.tsx(mount, gated behind the existingkv.init())desktop-translations*.ts4. Auto-install @openclaw/codex when a codex model is configured (fix)
OpenClaw 2026.5.12 also split the codex agent harness out of the
core gateway into a separate npm package (`@openclaw/codex`) and
only auto-installs it on the `openclaw onboard --auth-choice
openai-codex…` path. Because our `configure/route.ts` writes
`openclaw.json` directly (see the schema-drift note already in
that file), devices that pick a Codex model in the chat picker
never trigger the install and every reply fails with:
`Embedded agent failed before reply: Requested agent harness "codex" is not registered.`
Fix lives in `scripts/gateway-pre-start.sh` — single chokepoint,
runs on every gateway start, self-heals existing devices on update.
Detection mirrors OpenClaw's own `modelSelectionShouldEnsureCodexPlugin`
(checks the primary model + every auth profile for the
`openai-codex` provider). Idempotent — skips the install when
`@openclaw/codex/package.json` is already on disk.
How was this tested?
All four changes verified live on a Jetson dev install before
opening this PR:
go from
[ws] protocol mismatch code=1002to[ws] webchat connected.on the next status poll, no re-login. Real downgrade Pro → Free
also detected.
open chat tab, and the Free-downgrade modal fired after the real
cancel.
--force` followed by `systemctl restart clawbox-gateway` —
pre-start detected the codex model + missing plugin → installed →
gateway came up with `codex` in its loaded plugins list.
Checklist
bun run lint(will run in CI)bun run test(will run in CI)bun run build(will run in CI)Notes for the reviewer
added in feat(ai-models): auto-tier wizard + Free-tier UX gates #122 with the comment "Drop this guard once the portal
gates deviceTier by subscription." `mapPortalTier` now provides
that gating from the device side regardless of the portal's
contract, so we can drop the guard safely.
matching the rest of `gateway-pre-start.sh`'s style. Open to
switching to `openclaw plugins inspect codex` if you'd rather
go through the CLI than read the JSON directly.
(flash/pro) plus Free. If OpenClaw introduces an "Ultra" tier or
similar later, both the rank table and the CONTENT lookup will
need extending — trivially, but worth flagging.
fuchsia gradient used for upgrades; happy to swap to celebratory
styling if you'd rather the entire flow feel positive.
Summary by CodeRabbit
New Features
Bug Fixes
Tests