feat: agent YAML parsing, engine API fixes & AIOX cockpit enhancements#5
Open
rafaelscosta wants to merge 68 commits intomainfrom
Open
feat: agent YAML parsing, engine API fixes & AIOX cockpit enhancements#5rafaelscosta wants to merge 68 commits intomainfrom
rafaelscosta wants to merge 68 commits intomainfrom
Conversation
5 integrations (Engine, WhatsApp, Supabase, API Keys, Voice) with live health checks, status cards, and setup modals for each service. New /integrations view with sidebar navigation. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
… registry, secrets vault - 22 unit tests for integrationStore, useIntegrationStatus, IntegrationCard - Auto-redirect to /integrations on first run when nothing is connected - Sidebar status dot (red/yellow/green) on Integrations icon - WhatsApp Send Test button for end-to-end webhook validation - Engine-side integration CRUD routes (/integrations) - Encrypted secrets vault (AES-256-GCM) with /integrations/secrets API - SQLite migration 005_integrations for integrations + secrets tables - SPA engineApi methods for integrations and secrets Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- 3 new integration cards with health checks and setup modals - Telegram: bot token config, webhook setup, status check - Google Drive: OAuth Client ID config, auth status indicator - Google Calendar: shared OAuth config, Calendar API scope guidance - Updated store, hook, hub, and tests for 8 total integrations Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Item 1 — Google OAuth PKCE flow: - Engine /auth/google/* routes (url, callback, refresh, disconnect, status) - SPA GoogleOAuthCallback component handles redirect - Google Drive/Calendar modals with "Connect with Google" button - Tokens stored in engine secrets vault (AES-256-GCM) - Disconnect button revokes tokens and clears vault Item 2 — Telegram engine routes: - /telegram/status — bot info + webhook status from Telegram API - /telegram/webhook — receive updates with secret token verification - /telegram/webhook/setup — set webhook URL via Bot API - /telegram/send — send messages to any chat_id - /telegram/test — verify bot is active - /telegram/events — SSE stream with 100-event buffer Item 3 — SPA ↔ Engine sync: - integration-sync.ts — helper library for engine-first operations - syncStatusToEngine() — push status to engine DB (best-effort) - storeSecret/getSecretValue — engine vault with localStorage fallback - startGoogleOAuth/completeGoogleOAuth — full OAuth redirect flow - Health checks query engine /auth/google/status first, fallback local - URL sync for /integrations and /auth/google/callback paths Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
All API route files used path.resolve(cwd, '..', '..') which resolved to /Downloads/ where Turbopack encountered an invalid symlink (cpl2/venv/bin/python). Fixed to use single parent (..) which correctly resolves to the apps/ project root. Also added outputFileTracingRoot to next.config.ts as safety net. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Auto-Experiment #1 — console.log cleanup (metric: lower_is_better) - Baseline: 41 console.log statements - After: 17 (legitimate lifecycle/connection logs kept) - Delta: -24 (-58.5%) Removed: debug traces in use-chat, use-execute, page.tsx, StoryDetailModal, workflows API SSE handlers Converted: retry logs → console.warn, failure logs → console.error, dev-only API log → console.debug Build: PASS | Tests: 65/65 PASS Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
…ional Auto-Experiment #2: Replace unsafe :any with proper TypeScript types. - AgentList: AgentSummary for all agent references - AgentProfileModal: custom AgentProfile interface extending PlatformAgent - ChatContainer: AgentWithUI, Message, ChatSession, AgentCommand types - MarkdownRenderer: MdChildren type, explicit component prop types - ActivityPanel: AgentWithUI, CommandCategory interface (3 intentional :any kept for heterogeneous callbacks) - CategoryManager: explicit squad types, filter type guards - use-execute: Record<string, unknown> casts replacing (as any) - use-workflows: proper unknown→cast pattern for SSE data Build passes, 65/65 tests pass. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Auto-Experiment #3: Bundle size optimization via code splitting. Lazy loading (React.lazy + Suspense): - page.tsx: 11 view panels lazy-loaded (only active view loads) - AppLayout: AgentExplorer, GlobalSearch, KeyboardShortcuts lazy-loaded (modal/overlay components only load when opened) Barrel export cleanup (export * → named exports): - stories, kanban, terminals, terminal, roadmap, qa, github, context, insights, settings, agents, workflow index.ts files - Enables better tree-shaking by making imports explicit Build passes, 65/65 tests pass. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Auto-Experiment #4: Test coverage expansion. New test files: - utils.test.ts (25 tests): cn, formatRelativeTime, generateId, truncate, capitalizeFirst - domain-taxonomy.test.ts (17 tests): resolveSquadDomain priority chain, getDomainColor/Bg/Border/Label with fallbacks, DOMAIN_ORDER completeness - execution-log-store.test.ts (17 tests): startExecution, endExecution duration calc, addLog ID incrementing, agent/tool/error logging, clearLogs Total: 65 → 124 tests (+90.8%) Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Auto-Experiment #5: Dead code detection and removal. Removed: - useWorkflowSchema hook (never used in any component) - useWorkflowStats hook (never used in any component) - Unused WorkflowSchema/WorkflowStats type imports - shortcutDefinitions array (35 lines, duplicate of KeyboardShortcuts.tsx) - validationRules object (67 lines, declared but never referenced) - Unused getSquadType type import in use-execute.ts De-exported (internal-only usage): - queryKeys in use-optimistic.ts (used internally, never imported) Total: ~130 lines of dead code removed. Build passes, 124/124 tests pass. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Auto-Experiment #6: Accessibility audit fixes. aria-labels added: - Header: search button, notification bell (dynamic count), dismiss button, agent explorer, workflow toggle, activity panel toggle - All toggle buttons now have aria-expanded - Notification badge marked aria-hidden (count in button label) - Notification dropdown: role="menu" + aria-label Keyboard support added: - AgentCard: role="button", tabIndex, onKeyDown (Enter/Space), focus ring - SquadCard: role="button", tabIndex, onKeyDown (Enter/Space), focus ring Form labels: - GlobalSearch input: aria-label added FAB components: - FAB: aria-label from label/icon prop, icons marked aria-hidden - FABIcon: aria-label from tooltip/icon prop - HelpFAB: aria-label="Ajuda" Build passes, 124/124 tests pass. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
…scriptions Auto-Experiment #7: React.memo / useMemo audit. MessageBubble.tsx: - Extract animation objects to module-level constants (stable references) - Extract MARKDOWN_PATTERN regex to module-level constant - Memoize MessageAttachments with React.memo - Add useMemo for image/file filtering in MessageAttachments ConversationHistory.tsx: - Convert destructured store to selective Zustand subscriptions (prevents re-renders on isStreaming/isLoading/error changes) - Wrap filteredSessions, sortedSessions, grouped in useMemo (eliminates O(n*m) filter+sort+group on every render) - Wrap handleSelect, handleNewConversation, handleClearAll in useCallback Build passes, 124/124 tests pass. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Create Next.js error.tsx for route-level error handling - Wrap ViewContent with ErrorBoundary + resetKeys=[activeView] - View crashes are isolated; switching views auto-resets error state Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- CI: add lint, typecheck, test jobs (was only typecheck + build) - CI: add metrics-gate job that runs after quality checks - Metrics script: measures :any count, test count, barrel exports, build size - Metrics baseline: snapshot with current values (3 any, 124 tests, 28 barrels) - --ci flag blocks regressions, --save creates new baseline Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Replace 3 `any` typed callbacks in CommandCategory interface with `unknown` + explicit casts. Type safety fully enforced. :any count: 3 → 0 Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Replace 28 `export *` in hooks/index.ts and stores/index.ts with explicit named exports for better tree-shaking. export * count: 28 → 0 Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Convert 8 components with 3+ destructured store props to individual selectors. Prevents unnecessary re-renders when unrelated state changes. Files: AppLayout, Header, Sidebar, MobileNav, ExecutionLogPanel, ChatContainer, GlobalSearch, use-execute Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- npm audit fix: resolve minimatch ReDoS and rollup path traversal - npm update: patch-level updates for 17 packages (tailwind, zustand, framer-motion, swr, mermaid, etc.) - 0 vulnerabilities remaining Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- src/lib/env.ts: centralized env access with typed defaults - .env.example: documents all env vars used by the dashboard - Client vars validated at import, server vars via getServerEnv() Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Add apiFetcher to api client — SWR-compatible wrapper with retry, timeout, and error classification. Replace 4 raw fetch() fetchers in use-agents, use-stories, use-squads, use-aios-status. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Install @next/bundle-analyzer, enable with ANALYZE=true - Add scripts: analyze, metrics, metrics:save, metrics:ci - Run `npm run analyze` to visualize chunk sizes Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
:any count: 3 → 0 export * barrels: 28 → 0 test count: 124 (stable) vulnerabilities: 4 → 0 Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Install highlight.js (required by globals.css @import) - Bundle audit: 5.4MB total chunks, largest 424KB - mermaid (72MB src) already lazy-loaded - lucide-react (40MB src) tree-shakes correctly via named imports - No further optimization needed at this scale Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
…enhancements Major features: - Vault: enterprise data management with 3-level navigation (overview → workspace → document), 4 tabs (dados, templates, taxonomias, c-suite), Zustand store with Supabase sync - Marketplace: browse, listing detail, seller dashboard, submit wizard, review queue, purchases - Brainstorm: room list, idea canvas, idea cards, organize panel, output preview Enhancements: - Integrations: setup wizard, config export/import, QR share, env generator, webhook alerts - Engine: agent logger, process pool improvements, rate limit tests - Dashboard: health cards, dependency graph, health sparkline - Terminals: live terminal card/output, SSE hook - GitHub/QA/Context/Monitor views refactored and improved - New hooks: useHealthCheck, useCapabilities, useEngineConnection, useMarketplace, etc. - Supabase services for brainstorm, marketplace, vault, roadmap - Multiple store additions and test coverage Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
…utes Includes marketplace-escrow-release edge function, dashboard API, integrations API routes, and knowledge files overview fixes. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
…d tests Includes useDashboardOverview, useMarketplaceAgents hooks, settings Supabase service, health report tests, and connection profile store tests. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
The comment `squads/*/agents/` prematurely closed the JSDoc block due to `*/`. Changed to `squads/[name]/agents/`. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
New test files: - toast-store.test.ts: timer tracking, auto-dismiss, clearAll cleanup - theme.test.ts: getSquadTheme, getTierTheme for all squad types - icons.test.ts: getIconComponent, EMOJI_TO_ICON, ICON_MAP - env.test.ts: client env defaults, getServerEnv() shape Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Prevent unbounded growth of execution log store by keeping only the last 500 entries. Matches monitor-store's MAX_EVENTS pattern. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
:any: 0 | tests: 272 | export*: 0 | vulns: 0 | TS errors: 0 Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Update docker-compose, nginx, vite config, and package dependencies. Add SSL setup and VPS provisioning scripts, deploy env example. Add Supabase migrations (brainstorm rooms, orchestration tasks, seed) and remove superseded migration files. Add tech audit plan and migration documentation. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
rafaelscosta
changed the title
…ault emojis - Parse rich agent data (persona, commands, principles, voiceDNA, anti-patterns, integration) from engine YAML content field in getAgent() - Fix engine response unwrapping in getRegistryAgent (squad→squadId, title→role) - Add React Query cache invalidation via engineStatus in useSquads/useAgents - Replace emoji strings with Lucide icons in Vault components (VaultOverview, WorkspaceDetail) using getIconComponent() - Route CommandsModal and AgentCommandsPanel to engine API (tasks/workflows) instead of non-existent /api/squads/:id/commands endpoint - Add "Engine offline" informative message in EmptyChat - Fix agents-api tests with proper engine URL mock - AIOX cockpit theme enhancements, new cockpit components, Geist font, HUD patterns, animations, and Glass→Cockpit migration groundwork Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
rafaelscosta
changed the title
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
AIOX Dashboard Premium — AI Orchestration Dashboard with: - React 19 + Vite 7 + TypeScript - Cockpit-style dark theme (AIOX Brandbook) - 37 squad visualization + agent registry - Real-time orchestration via SSE - Engine integration (Bun + Claude CLI) - Supabase persistence - E2E tests (Playwright) + unit tests (Vitest) - Storybook component docs - Docker deployment config Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
…Wave 1] Install shadcn infrastructure (CVA, Radix Slot) and create 3 core components (Button, Card, Input) using Tailwind + CSS vars mapped to AIOX Cockpit tokens. Add ds-preview view for visual comparison with existing Cockpit components. Zero breaking changes to existing UI. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
…mium dashboard merge Engine rewrite (~850 lines): - Bun/Hono server on :4002 with SQLite persistence - Claude CLI integration for AI planning and step execution - SSE streaming with 15 event types matching shared/api-contract.ts - Plan approval flow with polling, revision support - Agent discovery from filesystem (60+ agents) - Demo mode fallback when Claude CLI unavailable shadcn/ui Wave 1: - Button, Card, Input components with CVA variants - CSS variable bridge mapping AIOX tokens to shadcn vars - components.json for shadcn CLI compatibility - DSPreview comparison page Premium dashboard merge: - Marketing hub with traffic, content, funnels, creatives modules - Sales dashboard, traffic dashboard, creative gallery - Agent tech-sheet detailed view - Vault SSOT with spaces, sources, document upload - Enhanced vault store with full CRUD - Type system expansion (ViewType, VaultTab, DocumentType) Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Map --secondary and --accent to --aiox-blue (#0099FF) instead of lime, matching the existing dashboard palette where blue is the secondary interactive color. Adjust Button secondary variant to use translucent blue background with blue text/border. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
…ovable Rewrite Button, Card, Input to match standard shadcn/ui patterns from the lovable dashboard. Fix CSS vars bridge: secondary=surface-alt, accent=lime glow 10%, ring=lime 40%, radius=0.5rem. Add DEFAULT values for background/foreground/accent in Tailwind config so bg-background, text-foreground, border-input resolve correctly. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Add /ds-preview path to VIEW_PATHS so the view is accessible via URL. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Exclude engine/ from frontend tsconfig.app.json build - Inline vault-store types in src/services/api/vault.ts (break cross-project import) - Fix PackageBuilder.tsx engine import → use local api/vault types - Add backwards-compatible variants to CockpitButton (default/danger/outline) - Add accent prop alias to CockpitCard for backward compat - Add size prop to CockpitKpiCard (accepted, unused) - Relax CockpitTable generics (T=any), make header optional, add label alias - Add children support to CockpitSectionDivider - Add squad? field to 9 agent type interfaces - Fix VaultImportDialog missing VaultDocument fields - Fix TrafficDashboard CockpitTableColumn import - Fix OrchestrationActivityPanel variable shadowing - Fix sales-room/seed.ts duplicate property - Fix useTrafficData missing clicks field - Add marketing-hub to ViewType union - Make Celebration trigger optional with celebrating alias Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Register DSPreview lazy component in App.tsx viewMap - Add loading message for ds-preview view - Add Design System entry to sidebar EXTRA_ITEMS with Palette icon Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Phase 1 (Foundation): - Expanded types with VaultSpace, DataSource, DocumentQuality, 7-tab system - Supabase migration: vault_spaces, vault_sources, vault_documents_v2, vault_sync_jobs, vault_taxonomy, vault_context_packages, vault_activity - Engine SQLite cache + vault-store CRUD for all entities - File parsers (PDF, DOCX, XLSX, CSV, MD, TXT) with content hashing - Frontend API service, Supabase service, mock data with spaces/sources Phase 2 (Intelligence): - AI Services: classify, summarize, taxonomy, quality-score, tags via Claude CLI - Connector system: VaultConnector interface, URL scrape, AI memory parser - ETL sync runner: discover -> extract -> transform -> load pipeline - Context Package builder with markdown/JSON/YAML export - AI Memory Import wizard (Claude, ChatGPT, Gemini, Copilot) - Sync job SSE streaming for real-time progress - 37 engine routes total Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Copied 16 brandbook images (product mockups, apparel, jackets, agent cards, squad group) to public/creatives/ and replaced all placehold.co URLs in the CreativeGallery component. Images include: bag, bottle, cap, cards, flatlay, mug, bomber, hoodie, tshirt, LED jacket, neon jacket, tech jacket, agent ARIA/DEX/ORION, squad. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Single-key view navigation shortcuts (H, D, S, etc.) were firing while typing in the rejection notes textarea inside CockpitModal. Added two guards: - Check for SELECT and role="textbox" elements - Suppress all single-key shortcuts when target is inside a dialog, [role="dialog"], or [aria-modal="true"] Modifier shortcuts (Cmd+K, Cmd+B, etc.) still work everywhere. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
…utes Marketing Hub: - Overview with HeroKpiStrip, SecondaryMetrics, MarqueeTicker, SectionNumber - Traffic module with expanded campaign metrics - Content module: CarouselBuilder, ContentCalendar, ThumbnailCreator - Creatives module with updated studio layout - Funnels module: FunnelBuilder, TemplateGallery - Analytics: UnifiedDashboard with enhanced visualizations - Scenarios: BudgetSimulator, GoalCalculator, BreakEvenVisualizer - Charts system: 9 chart components (Area, Bar, Donut, Funnel, Heatmap, Scatter, Sparkline, Waterfall) with AIOX theme + ChartContainer + FilterBar - Shared: PlatformDistribution, MarketingKpiCard refactor Dashboard: - PlatformIntelligencePanel, IntegrationGraphPanel, KnowledgeSearchPanel - SquadHealthBadge component - usePlatformIntelligence hook Engine: - Content routes (content.ts) - Platform routes (platform.ts) Infra: - Migrate postcss.config.js -> postcss.config.cjs - New stores: graphStore, maturityStore, qualityGateStore - Engine API service expansion - recharts dependency added Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Raw JSON from Claude CLI stream-json was dumped as-is into the terminal, making it unreadable. Now the TerminalOutput component parses each line: - Detects JSON events (system/init, assistant, tool_use, tool_result, error, done, rate_limit, routed messages) - Extracts readable content from nested message structures - Renders with colored labels (INIT, AI, TOOL, RESULT, DONE, ERROR) - Shows formatted timestamps (HH:MM:SS) - Truncates long content with ellipsis - Falls back to raw ANSI/heuristic rendering for non-JSON lines Also updated TerminalCard to reuse TerminalOutput instead of inline rendering logic. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Lines from Claude CLI stream-json often contain multiple JSON objects
concatenated without separators (e.g. '{"type":"system",...}{"type":"init",...}').
Added splitJsonBlobs() that tracks brace depth to split them into
individual objects before parsing. Each JSON is now rendered as its own
structured row with label, timestamp, and formatted content.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
…tables The agent content parser only recognized ```yaml fenced blocks (used by squad agents) but not --- frontmatter (used by core .claude/agents/). Changes: - parseAgentContent now tries frontmatter (---...---) first, then fenced - Added fallback: extract *command patterns from markdown tables when no commands: array exists in YAML - Extract sampleTasks from triggers/activation_triggers YAML arrays Fixes both "Acoes & Comandos" and "Perfil do Agente" modals showing empty states for agents. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
The menu had a scroll loop: onMouseEnter set selectedIndex, which triggered scrollIntoView, which moved the list under the cursor, triggering mouseEnter on a new item, repeating infinitely. Fix: track whether selection came from keyboard or mouse. Only call scrollIntoView for keyboard-driven selection changes. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Core agents (aios-qa, aios-dev, etc.) have no squad assignment. sendMessage was requiring squadId to be non-null, causing messages to silently disappear with an early return. Changes: - Remove squadId from the required fields check - Fallback squadId to 'core' when passing to executeMutation - Handle undefined squadId in getSquadType call Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
…gistry
The frontend expected { agents: [...], count: N } with squadId field,
but the engine returned a raw array with squad field. This caused all
340 agents to appear without squadId, breaking:
- Chat message sending (squadId required)
- Agent profile modal (squadId needed for detail fetch)
- Commands modal (squad-based task/workflow queries)
- Agent search and filtering
All /agents routes now map squad -> squadId and wrap in { agents, count }.
Agent detail routes wrap in { agent: {...} } as frontend expects.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Replaced the 400 placeholder with a real implementation that: - Loads agent persona from filesystem (loadAgentContent) - Spawns Claude CLI with agent context + user message - Streams response via SSE (start, text, done, error events) - Falls back to demo mode when Claude CLI is unavailable This is the endpoint the chat UI calls for real-time agent responses. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Frontend expects:
- start: { executionId, agentId, agentName }
- text: { content: "string" }
- done: { usage: { inputTokens, outputTokens, totalTokens }, duration: N }
Engine was sending:
- start: { agentId, sessionId } (missing executionId/agentName)
- done: { content, durationMs } (wrong shape, caused [object Object] display)
Also: result event no longer re-sends content (already streamed via
assistant events) — prevents duplicated text in the chat bubble.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Added dual persistence layer for chat sessions and messages: Migration (20260321_chat_sessions.sql): - chat_sessions: id, agent_id, agent_name, squad_id, title, message_count - chat_messages: id, session_id, role, content, metadata (JSONB), attachments (JSONB) - RLS anon full CRUD, indexes on session_id + created_at Service (supabase/chat.ts): - upsertSession, listSessions, deleteSession - addMessage, updateMessage, getMessages, bulkInsertMessages - Graceful fallback when Supabase unavailable (same pattern as vault.ts) Store (chatStore.ts): - All mutations now fire-and-forget sync to Supabase - _syncFromSupabase() on init: merges remote sessions missing locally - Seeds Supabase with localStorage sessions on first run - localStorage remains primary — Supabase is the durable backup Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Welcome message "O que posso fazer por voce" now shows agent-specific suggestions from: commands (YAML/markdown), sampleTasks (triggers), or hardcoded fallbacks by agent/squad. Slash command menu (/) now includes dynamic commands: - Agent *commands extracted from markdown definitions - Squad tasks from engine registry - Squad workflows from engine registry - Plus the existing global system commands (/help, /clear, /export, etc.) ChatContainer fetches tasks/workflows via useRegistryTasks/useRegistryWorkflows and builds SlashCommand[] passed through ChatInput -> SlashCommandMenu. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Claude CLI assistant events can return message as either a JSON string or a direct object. extractTextFromAssistant now handles both: - string: JSON.parse then extract content[].text - object: directly extract content[].text or content string - undefined: return empty string This fixes [object Object] appearing in chat responses when the CLI returns message as an object instead of a JSON-encoded string. Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
rafaelscosta
force-pushed
the
feat/glass-ui-design-system-v2
branch
from
bd07b98 to
11b88f2
Compare
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
contentfield — modal tabs now show real data instead of empty statesgetRegistryAgent, add React Query cache invalidation viaengineStatusinuseSquads/useAgents, routeCommandsModalandAgentCommandsPanelto engine endpointsgetIconComponent()across VaultOverview and WorkspaceDetailTest plan
npx tsc --noEmit)npx vitest run)🤖 Generated with Claude Code