merge: upstream/main, adopt upstream attention indicators, drop our unread

Pulls in 24 upstream commits including session list status indicators
(attention + scheduled), lightbox image preview, agent flavor icons in
session header, Claude effort options, contextWindow propagation,
permissionMode persistence across hub restart, voice picker overhaul,
unique assistant-ui message IDs, and a heap of smaller fixes.

Upstream's PR tiann#699 (session-list status indicators) overlaps with the
unread badge we shipped in d530ee2. Per user direction, dropped our
implementation in favor of upstream's classifySessionAttention (which
distinguishes permission > input > background > unread per session)
plus the localStorage-backed sessionLastSeen tracker. Our cross-device
"unreadAt" metadata field, hub markSessionUnread/Read, /mark-read
route, NotificationHub mark-unread hooks, and SessionChat auto-clear
effect are all removed.

Kept from the fork:
  - Recent (12h) section + cross-device pin (server-side metadata.pinnedAt)
  - YOLO default for cursor / opencode permission mode
  - Continue + auto-discovery fallback for resume in spawnSession
  - usage/accountStatus on Session, UsageBar in settings, claude-mem
    auto-status, Claude 4.6/4.7 1M presets, etc.

Conflict resolutions of note:
  - SessionList.tsx: combined our Recent+Pin rendering with upstream's
    SessionAttentionIndicator; SessionItem now accepts both
    isPinned/onTogglePin and showDetailedStatus.
  - syncEngine.spawnSession: kept our useContinue + listAgentSessions
    fallback, took upstream's preferredPermissionMode formula and
    threaded it through the spawn call.
  - applySessionConfig: kept our legacy `applied.model` backfill
    BEFORE upstream's strict requested-key check.
  - StatusBar: kept our usage/accountStatus block, took upstream's
    opencode reasoning label extension.
  - SessionHeader: took upstream's AgentFlavorIcon, kept our
    getAgentDisplayName for the label.
  - UserMessage: kept our timestamp / queued / metadata-toggle bubble,
    replaced upstream's simpler bubble; reintroduced
    toggleMetadata/onMetadataKeyDown callbacks that were stale in the
    fork commit.
  - settings/index.tsx: kept our UsageBar/formatResetTime + upstream's
    PlayIcon/StopIcon, deduped useAppContext import and `api`
    destructure.
  - MessageAttachments: kept upstream's ImagePreview lightbox but with
    overflow-clip (so trackpad scroll doesn't latch on the image).
  - permissionModePersistence test mock: extended spawnSession mock
    signature with _sandbox/_continueLatest so the new 13th-arg
    permissionMode is captured.
  - modelOptions.test: expected arrays now include our claude-opus-4-6
    and 4-7 1M presets.

Build 48 -> 49.

via [HAPI](https://hapi.run)

Co-Authored-By: HAPI <noreply@hapi.run>

Author: yoyoworms

.github/pr-assets/pr-agent-flavor-icons-before.jpg

@@ -11,13 +11,7 @@ Thank you for your interest in contributing to HAPI! We welcome bug fixes, featu
 
 ## AI-Generated Code Policy (Vibe Coding)
 
-If you are using AI tools to generate code ("Vibe Coding"), please note:
-
-**We only accept code generated by the GPT-5.4 model.**
-
-Pull requests containing code generated by other AI models (including other GPT versions, Claude, Gemini, or any other LLM) will be rejected. This policy helps us maintain consistency and quality standards across the codebase.
-
-If you're unsure whether your AI-assisted code meets this requirement, please open an issue to discuss before submitting.
+If you used AI tools to generate code ("Vibe Coding"), please disclose the model in your PR description. Code is judged on merit — focused scope, tests, and the other expectations below apply equally regardless of how the code was authored.
 
 ## Pull Request Guidelines
 

.github/pr-assets/pr-agent-flavor-icons-claude.jpg

@@ -155,6 +155,45 @@ describe('AcpSdkBackend', () => {
         });
     });
 
+    it('captures model metadata from configOptions when models block is missing', async () => {
+        const backend = new AcpSdkBackend({ command: 'opencode' });
+        const backendInternal = backend as unknown as {
+            transport: { sendRequest: (method: string, params: unknown) => Promise<unknown>; close: () => Promise<void> } | null;
+        };
+        backendInternal.transport = {
+            sendRequest: async (method) => {
+                if (method === 'session/new') {
+                    return {
+                        sessionId: 'opencode-session-config-options',
+                        configOptions: [
+                            {
+                                id: 'model',
+                                category: 'model',
+                                currentValue: 'opencode/big-pickle',
+                                options: [
+                                    { value: 'opencode/big-pickle', name: 'OpenCode Zen/Big Pickle' },
+                                    { value: 'deepseek/deepseek-chat', name: 'DeepSeek/DeepSeek Chat' }
+                                ]
+                            }
+                        ]
+                    };
+                }
+                return null;
+            },
+            close: async () => {}
+        };
+
+        const sessionId = await backend.newSession({ cwd: '/tmp/x', mcpServers: [] });
+
+        expect(backend.getSessionModelsMetadata(sessionId)).toEqual({
+            availableModels: [
+                { modelId: 'opencode/big-pickle', name: 'OpenCode Zen/Big Pickle' },
+                { modelId: 'deepseek/deepseek-chat', name: 'DeepSeek/DeepSeek Chat' }
+            ],
+            currentModelId: 'opencode/big-pickle'
+        });
+    });
+
     it('returns undefined session metadata when session/new omits models', async () => {
         const backend = new AcpSdkBackend({ command: 'gemini' });
         const backendInternal = backend as unknown as {
@@ -213,6 +252,67 @@ describe('AcpSdkBackend', () => {
         });
     });
 
+
+
+    it('captures and sets OpenCode thought-level config option', async () => {
+        const backend = new AcpSdkBackend({ command: 'opencode' });
+        const calls: Array<{ method: string; params: unknown }> = [];
+        const backendInternal = backend as unknown as {
+            transport: { sendRequest: (method: string, params: unknown) => Promise<unknown>; close: () => Promise<void> } | null;
+        };
+        backendInternal.transport = {
+            sendRequest: async (method, params) => {
+                calls.push({ method, params });
+                if (method === 'session/new') {
+                    return {
+                        sessionId: 's1',
+                        configOptions: [{
+                            id: 'effort',
+                            name: 'Effort',
+                            category: 'thought_level',
+                            type: 'select',
+                            currentValue: 'low',
+                            options: [
+                                { value: 'low', name: 'Low' },
+                                { value: 'high', name: 'High' }
+                            ]
+                        }]
+                    };
+                }
+                if (method === 'session/set_config_option') {
+                    return {
+                        configOptions: [{
+                            id: 'effort',
+                            category: 'thought_level',
+                            currentValue: 'high',
+                            options: [{ value: 'high', name: 'High' }]
+                        }]
+                    };
+                }
+                return null;
+            },
+            close: async () => {}
+        };
+
+        await backend.newSession({ cwd: '/tmp/x', mcpServers: [] });
+        expect(backend.getThoughtLevelConfigOption('s1')).toMatchObject({
+            id: 'effort',
+            currentValue: 'low',
+            options: [{ value: 'low', name: 'Low' }, { value: 'high', name: 'High' }]
+        });
+
+        await backend.setConfigOption('s1', 'effort', 'high');
+
+        expect(calls).toContainEqual({
+            method: 'session/set_config_option',
+            params: { sessionId: 's1', configId: 'effort', value: 'high' }
+        });
+        expect(backend.getThoughtLevelConfigOption('s1')).toMatchObject({
+            id: 'effort',
+            currentValue: 'high'
+        });
+    });
+
     it('emits turn_complete after trailing tool updates from the same turn', async () => {
         backendStatics.UPDATE_QUIET_PERIOD_MS = 25;
         backendStatics.UPDATE_DRAIN_TIMEOUT_MS = 200;
@@ -284,4 +384,65 @@ describe('AcpSdkBackend', () => {
             'turn_complete'
         ]);
     });
+
+    it('combines OpenCode usage_update and prompt usage into a usage message', async () => {
+        backendStatics.UPDATE_QUIET_PERIOD_MS = 25;
+        backendStatics.UPDATE_DRAIN_TIMEOUT_MS = 200;
+        backendStatics.PRE_PROMPT_UPDATE_QUIET_PERIOD_MS = 1;
+        backendStatics.PRE_PROMPT_UPDATE_DRAIN_TIMEOUT_MS = 50;
+
+        const backend = new AcpSdkBackend({ command: 'opencode' });
+        const backendInternal = backend as unknown as {
+            transport: {
+                sendRequest: (...args: unknown[]) => Promise<unknown>;
+                close: () => Promise<void>;
+            } | null;
+            handleSessionUpdate: (params: unknown) => void;
+        };
+
+        const messages: AgentMessage[] = [];
+        backendInternal.transport = {
+            sendRequest: async () => {
+                setTimeout(() => {
+                    backendInternal.handleSessionUpdate({
+                        sessionId: 'session-1',
+                        update: {
+                            sessionUpdate: 'usage_update',
+                            used: 13_879,
+                            size: 65_536,
+                        }
+                    });
+                }, 0);
+
+                await sleep(5);
+
+                return {
+                    stopReason: 'end_turn',
+                    usage: {
+                        totalTokens: 13_892,
+                        inputTokens: 8_119,
+                        outputTokens: 2,
+                        thoughtTokens: 11,
+                        cachedReadTokens: 5_760
+                    }
+                };
+            },
+            close: async () => {}
+        };
+
+        await backend.prompt('session-1', [{ type: 'text', text: 'hello' }], (message) => {
+            messages.push(message);
+        });
+
+        expect(messages).toContainEqual({
+            type: 'usage',
+            inputTokens: 8_119,
+            outputTokens: 2,
+            cacheReadTokens: 5_760,
+            thoughtTokens: 11,
+            totalTokens: 13_892,
+            contextTokens: 13_879,
+            contextWindow: 65_536
+        });
+    });
 });