BTerminal/docs/progress.md

BTerminal v2 — Progress Log

Session: 2026-03-05

Research Phase (complete)

• Analyzed current BTerminal v1 codebase (2092 lines Python, GTK3+VTE)
• Queried Memora — no existing BTerminal memories
• Researched Claude Agent SDK — found structured streaming, subagent tracking, hooks
• Researched Tauri + xterm.js ecosystem — found 4+ working projects
• Researched terminal latency benchmarks — xterm.js acceptable for AI output
• Researched 32:9 ultrawide layout patterns
• Evaluated GTK4 vs Tauri vs pure Rust — Tauri wins for this use case
• Created task_plan.md with 8 phases
• Created findings.md with 7 research areas

Technology Decision (complete)

• Decision: Tauri 2.x + Solid.js + Claude Agent SDK + xterm.js
• Rationale documented in task_plan.md Phase 0

Adversarial Review (complete)

• Spawned devil's advocate agent to attack the plan
• Identified 5 fatal/critical issues:
  1. Node.js sidecar requirement unacknowledged
  2. SDK 0.2.x instability — need abstraction layer
  3. Three-tier observation overengineered → simplified to two-tier
  4. Solid.js ecosystem too small → switched to Svelte 5
  5. Missing: packaging, error handling, testing, responsive design
• Revised plan (Rev 2) incorporating all corrections
• Added error handling strategy table
• Added testing strategy table
• Defined MVP boundary (Phases 1-4)
• Added responsive layout requirement (1920px degraded mode)

Phase 1 Scaffolding (complete)

• Created feature branch v2-mission-control
• Initialized Tauri 2.x + Svelte 5 project in v2/ directory
• Rust backend stubs: main.rs, lib.rs, pty.rs, sidecar.rs, watcher.rs, session.rs
• Svelte frontend: App.svelte with Catppuccin Mocha CSS variables, component stubs
• Node.js sidecar scaffold: agent-runner.ts with NDJSON communication pattern
• Tauri builds and launches (cargo build --release verified)
• Dev scripts: npm run dev, npm run build, npm run tauri dev/build
• 17 operational rules added to .claude/rules/
• Project meta files: CLAUDE.md, .claude/CLAUDE.md, TODO.md, CHANGELOG.md
• Documentation structure: docs/README.md, task_plan.md, phases.md, findings.md, progress.md

Phase 2: Terminal Pane + Layout (complete)

• Rust PTY backend with portable-pty (PtyManager: spawn, write, resize, kill)
• PTY reader thread emitting Tauri events (pty-data-{id}, pty-exit-{id})
• Tauri commands: pty_spawn, pty_write, pty_resize, pty_kill
• xterm.js terminal pane with Canvas addon (explicit, no WebGL)
• Catppuccin Mocha theme for xterm.js (16 ANSI colors)
• FitAddon with ResizeObserver + 100ms debounce
• PTY bridge adapter (spawnPty, writePty, resizePty, killPty, onPtyData, onPtyExit)
• CSS Grid tiling layout with 5 presets (1-col, 2-col, 3-col, 2x2, master-stack)
• Layout store with Svelte 5 $state runes and auto-preset selection
• Sidebar with session list, layout preset selector, new terminal button
• Keyboard shortcuts: Ctrl+N new terminal, Ctrl+1-4 focus pane
• PaneContainer with header bar (title, status, close)
• Empty state welcome screen with Ctrl+N hint
• npm dependencies: @xterm/xterm, @xterm/addon-canvas, @xterm/addon-fit
• Cargo dependencies: portable-pty, uuid

Phase 3: Agent SDK Integration (in progress)

• Rust SidecarManager: spawn Node.js, stdio NDJSON, query/stop/shutdown (sidecar.rs, 218 lines)
• Node.js agent-runner: spawns claude -p --output-format stream-json, manages sessions (agent-runner.ts, 176 lines)
• Tauri commands: agent_query, agent_stop, agent_ready in lib.rs
• Sidecar auto-start on app launch
• SDK message adapter: full stream-json parser with 9 typed message types (sdk-messages.ts, 234 lines)
• Agent bridge: Tauri IPC adapter for sidecar communication (agent-bridge.ts, 53 lines)
• Agent dispatcher: routes sidecar events to agent store (agent-dispatcher.ts, 87 lines)
• Agent store: session state with messages, cost tracking (agents.svelte.ts, 91 lines)
• AgentPane component: prompt input, message rendering, stop button, cost display (AgentPane.svelte, 420 lines)
• UI integration: Ctrl+Shift+N for new agent, sidebar agent button, TilingGrid routing

Architecture decision: Uses claude CLI with --output-format stream-json instead of Agent SDK query() API. Avoids SDK npm dependency and version churn while getting identical structured output.

Bug Fix: Svelte 5 Rune File Extensions (2026-03-06)

• Diagnosed blank screen / "rune_outside_svelte" runtime error
• Root cause: store files used .ts extension but contain Svelte 5 $state/$derived runes, which only work in .svelte and .svelte.ts files
• Renamed: layout.ts -> layout.svelte.ts, agents.ts -> agents.svelte.ts, sessions.ts -> sessions.svelte.ts
• Updated all import paths in 5 files to use .svelte suffix (e.g., from './stores/layout.svelte')

Phase 3 Polish (2026-03-06)

• Sidecar crash detection: dispatcher listens for sidecar-exited event, marks running sessions as error
• Restart UI: "Restart Sidecar" button in AgentPane error bar, calls agent_restart command
• Auto-scroll lock: scroll handler disables auto-scroll when user scrolls >50px from bottom, "Scroll to bottom" button appears

Phase 4: Session Management + Markdown Viewer (2026-03-06)

• rusqlite 0.31 (bundled) + dirs 5 + notify 6 added to Cargo.toml
• SessionDb: SQLite with WAL mode, sessions table + layout_state singleton
• Session CRUD: list, save, delete, update_title, touch (7 Tauri commands)
• Frontend session-bridge.ts: typed invoke wrappers for all session/layout commands
• Layout store wired to persistence: addPane/removePane/focusPane/setPreset all persist
• restoreFromDb() on app startup restores panes in layout order
• FileWatcherManager: notify crate watches files, emits Tauri "file-changed" events
• MarkdownPane component: marked.js rendering, Catppuccin-themed styles, live reload
• Sidebar "M" button opens file picker for .md/.markdown/.txt files
• TilingGrid routes markdown pane type to MarkdownPane component

Phase 5: Agent Tree + Polish (2026-03-06, complete)

• Agent tree visualization (SVG): AgentTree.svelte component with horizontal tree layout, bezier edges, status-colored nodes; agent-tree.ts utility (buildAgentTree, countTreeNodes, subtreeCost)
• Agent tree toggle in AgentPane: collapsible tree view shown when tool_call messages exist
• Global status bar: StatusBar.svelte showing terminal/agent pane counts, active agents with pulse animation, total tokens and cost
• Notification system: notifications.svelte.ts store (notify, dismissNotification, max 5 toasts, 4s auto-dismiss) + ToastContainer.svelte (slide-in animation, color-coded by type)
• Agent dispatcher notifications: toast on agent_stopped (success), agent_error (error), sidecar crash (error), cost result (success with cost/turns)
• Settings dialog: SettingsDialog.svelte modal (default shell, cwd, max panes, theme flavor) with settings-bridge.ts adapter
• Settings backend: settings table (key/value) in session.rs, Tauri commands settings_get/set/list in lib.rs
• Keyboard shortcuts: Ctrl+W close focused pane, Ctrl+, open settings dialog
• CSS grid update: app.css grid-template-rows '1fr' -> '1fr auto' for status bar row
• App.svelte: integrated StatusBar, ToastContainer, SettingsDialog components

Phase 6: Packaging + Distribution (2026-03-06)

• Created install-v2.sh — build-from-source installer with 6-step dependency check process
  • Checks Node.js 20+, Rust 1.77+, system libs (WebKit2GTK, GTK3, GLib, etc.)
  • Prompts to install missing packages via apt
  • Builds with npx tauri build, installs to ~/.local/bin/bterminal-v2
  • Creates desktop entry and installs SVG icon
• Updated v2/src-tauri/tauri.conf.json: bundle targets ["deb", "appimage"], category, descriptions, deb depends, appimage settings
• Regenerated all icons in v2/src-tauri/icons/ from bterminal.svg as RGBA PNGs (32x32, 128x128, 256x256, 512x512, .ico)
• Created .github/workflows/release.yml — CI workflow triggered on v* tags
  • Ubuntu 22.04 runner, caches Rust/npm deps
  • Builds .deb + AppImage, uploads as GitHub Release artifacts via softprops/action-gh-release@v2
• Build verified: .deb (4.3 MB), AppImage (103 MB) both built successfully
• Tauri auto-update plugin deferred (needs signing key + update server)

Phase 5 continued: SSH, ctx, themes, detached mode, auto-updater (2026-03-06)

• ctx integration: Rust ctx.rs (read-only CtxDb with SQLITE_OPEN_READ_ONLY), 5 Tauri commands (ctx_list_projects, ctx_get_context, ctx_get_shared, ctx_get_summaries, ctx_search), ctx-bridge.ts adapter, ContextPane.svelte (project selector, tabs for entries/summaries/search)
• SSH session management: SshSession struct + ssh_sessions table in session.rs, 3 Tauri commands (ssh_session_list/save/delete), ssh-bridge.ts adapter, SshDialog.svelte (create/edit modal with validation), SshSessionList.svelte (grouped by folder, color dots)
• TilingGrid SSH routing: SSH pane type routes to TerminalPane with shell=/usr/bin/ssh and constructed args array
• Catppuccin theme flavors: themes.ts with all 4 palettes (Latte/Frappe/Macchiato/Mocha), theme.svelte.ts reactive store, SettingsDialog flavor dropdown, TerminalPane uses getXtermTheme(), persistence via SQLite settings
• Detached pane mode: detach.ts utility (isDetachedMode, getDetachedConfig from URL params), App.svelte renders single pane in full-viewport without chrome when ?detached=1
• Syntax highlighting: highlight.ts with Shiki lazy singleton (13 preloaded languages, catppuccin-mocha theme), integrated into MarkdownPane and AgentPane text messages
• Tauri auto-updater plugin: tauri-plugin-updater (Rust) + @tauri-apps/plugin-updater (npm) + updater.ts frontend utility
• AgentPane markdown rendering: text messages now rendered as markdown with Shiki highlighting
• New npm dependencies: shiki, @tauri-apps/plugin-updater
• New Rust dependency: tauri-plugin-updater

Session: 2026-03-06 (continued) — Polish, Testing, Extras

Terminal Copy/Paste + Theme Hot-Swap

• Copy/paste in TerminalPane via Ctrl+Shift+C/V (attachCustomKeyEventHandler: C copies selection, V reads clipboard and writes to PTY)
• Theme hot-swap: onThemeChange() callback registry in theme.svelte.ts, TerminalPane subscribes and updates term.options.theme on flavor change
• All open terminals now update immediately when theme flavor changes

Agent Tree Enhancements

• Click tree node -> scroll to corresponding message (handleTreeNodeClick in AgentPane, scrollIntoView with smooth behavior)
• Subtree cost display: yellow cost text below each tree node label (subtreeCost from agent-tree.ts, NODE_H increased from 32 to 40)
• Each message div has id="msg-{msg.id}" for scroll targeting

Session Resume

• Follow-up prompt input appears after session completes or errors (if sdkSessionId exists)
• startQuery(text, resume=true) passes resume_session_id to SDK via agent_query
• Styled .follow-up input + button in done-bar and error-bar sections

Pane Drag-Resize Handles

• Splitter overlays in TilingGrid (positioned with fixed CSS, outside grid to avoid layout interference)
• Column splitters: vertical bars between grid columns with mousemove drag
• Row splitters: horizontal bars between grid rows with mousemove drag
• customColumns/customRows $state override preset grid-template; reset on preset change
• Supports 2-col, 3-col, and 2-row layouts with min 10% / max 90% ratio clamping
• .dragging class disables user-select during drag; splitter hover shows accent color at 40% opacity

Auto-Update Workflow Enhancement

• release.yml: TAURI_SIGNING_PRIVATE_KEY + PASSWORD env vars passed to build step
• Generates latest.json for auto-updater (version, pub_date, platform URL, signature from .sig file)
• Uploads latest.json alongside .deb and .AppImage as release artifacts

Deno Sidecar Evaluation

• Created agent-runner-deno.ts proof-of-concept (Deno.Command for claude CLI, TextLineStream for NDJSON)
• Same NDJSON protocol as Node.js version; compiles to single binary with deno compile
• Not yet integrated with Rust SidecarManager; needs real-world testing

Testing Infrastructure

• Vitest added: vitest ^4.0.18 dev dependency, vite.config.ts test config, npm run test script
• sdk-messages.test.ts: tests for adaptSDKMessage() — init, text, thinking, tool_use, tool_result, cost, unknown types
• agent-tree.test.ts: tests for buildAgentTree(), countTreeNodes(), subtreeCost()
• Cargo: tempfile 3 dev dependency added
• session.rs tests: SessionDb CRUD (sessions, SSH sessions, settings, layout), uses tempfile::tempdir()
• ctx.rs tests: CtxDb error handling with missing database (conn=None)

Session: 2026-03-06 (continued) — Session Groups, Auto-Update Key, Deno Sidecar, Tests

Auto-Update Signing Key

• Generated Tauri signing keypair (minisign)
• Set pubkey in tauri.conf.json updater.pubkey field (base64 encoded)
• Private key to be stored as TAURI_SIGNING_PRIVATE_KEY GitHub secret by user

Session Groups/Folders

• Added group_name column to sessions table (session.rs, ALTER TABLE migration with pragma_table_info check)
• Session struct: added group_name field with #[serde(default)]
• SessionDb: update_group(id, group_name) method + save/list queries updated
• Tauri command: session_update_group registered in lib.rs
• Frontend: updateSessionGroup() in session-bridge.ts
• Layout store: Pane.group? field, setPaneGroup(id, group) function in layout.svelte.ts
• SessionList.svelte: grouped sidebar with collapsible headers (arrow + count), right-click to set group via prompt()
• Svelte 5 snippet pattern used for paneItem to avoid duplication across grouped/ungrouped rendering

Deno Sidecar Integration (upgraded from PoC)

• SidecarCommand struct in sidecar.rs: { program, args } abstracts runtime choice
• resolve_sidecar_command(): Deno-first resolution (checks agent-runner-deno.ts + deno availability), Node.js fallback
• Both sidecar runners bundled in tauri.conf.json resources array
• Graceful fallback: if Deno binary not found in PATH, falls back to Node.js with log warning

E2E/Integration Tests

• layout.test.ts (30 tests): addPane, removePane, focusPane, setPreset, autoPreset, getGridTemplate, getPaneGridArea, renamePaneTitle, setPaneGroup
• agent-bridge.test.ts (11 tests): Tauri IPC mock pattern with vi.hoisted(), invoke/listen wrappers
• agent-dispatcher.test.ts (18 tests): sidecar event routing, crash detection, vi.useFakeTimers() for async setTimeout
• sdk-messages.test.ts rewrite (25 tests): removed unused ErrorContent import
• E2E scaffold: v2/tests/e2e/README.md documenting WebDriver approach
• Total: 104 vitest tests + 29 cargo tests, all passing

Build status: TypeScript 0 errors, Rust 0 errors (1 pre-existing warning), all tests green.

Session: 2026-03-06 (continued) — Agent Teams / Subagent Support

Agent Teams Frontend Support

• Agent store: AgentSession extended with parentSessionId?, parentToolUseId?, childSessionIds[] for parent-child hierarchy
• Agent store: createAgentSession() accepts optional parent param, registers bidirectional parent/child links
• Agent store: findChildByToolUseId(parentId, toolUseId), getChildSessions(parentId) query functions
• Agent store: removeAgentSession() cleans up parent's childSessionIds on removal
• Agent dispatcher: SUBAGENT_TOOL_NAMES detection ('Agent', 'Task', 'dispatch_agent') on tool_call events
• Agent dispatcher: spawnSubagentPane() creates child session + layout pane, auto-groups under parent title
• Agent dispatcher: toolUseToChildPane Map routes messages with parentId to correct child pane
• Agent dispatcher: handleAgentEvent() splits messages — parentId-bearing go to child panes, rest to parent
• AgentPane: parent link bar (SUB badge + navigate-to-parent button)
• AgentPane: children bar (chips per child subagent, status-colored, clickable to focus)
• SessionList: subagent panes show '↳' icon instead of '*'

Design: No separate sidecar process per subagent. Parent's sidecar handles all; routing is purely frontend based on SDK's parentId field.

Session: 2026-03-06 (continued) — Subagent Tests, Cost Aggregation, Signing Key

Subagent Dispatcher Tests

• 10 new tests in agent-dispatcher.test.ts for subagent routing:
  • spawn subagent pane on Agent/Task tool_call
  • skip non-subagent tool_calls (Read, etc.)
  • deduplicate panes for same toolUseId
  • reuse existing child session from findChildByToolUseId
  • route messages with parentId to child pane
  • route child init message (sets model, marks running)
  • route child cost message (updates cost, marks done)
  • fallback title when input has no prompt/name
  • fallback group when parent pane not found
• Total: 28 dispatcher tests (18 existing + 10 new), 114 vitest tests overall
• New mocks added: mockCreateAgentSession, mockFindChildByToolUseId, mockAddPane, mockGetPanes, layout.svelte mock

Subagent Cost Aggregation

• getTotalCost(id) recursive helper in agents.svelte.ts — aggregates costUsd, inputTokens, outputTokens across parent + all children via childSessionIds
• AgentPane done-bar: shows "(total: $X.XXXX)" in yellow when child sessions exist and total > parent cost
• .total-cost CSS class: var(--ctp-yellow), 10px font-size

TAURI_SIGNING_PRIVATE_KEY

• Set via gh secret set on DexterFromLab/BTerminal GitHub repo

Session: 2026-03-06 (continued) — Multi-Machine Architecture Design

Multi-Machine Support Architecture

• Designed full multi-machine architecture in docs/multi-machine.md (303 lines)
• Three-layer model: BTerminal (controller) + bterminal-relay (remote binary) + unified frontend
• WebSocket NDJSON protocol: RelayCommand/RelayEvent envelope wrapping existing sidecar format
• Authentication: pre-shared token + TLS, rate limiting, lockout
• Autonomous relay model: agents keep running when controller disconnects
• Reconnection with exponential backoff (1s-30s), state_sync on reconnect
• 4-phase implementation plan: A (extract bterminal-core crate), B (relay binary), C (RemoteManager), D (frontend)
• Updated TODO.md and docs/task_plan.md to reference the design

Next Steps

• Deno sidecar: test with real claude CLI, benchmark startup time vs Node.js
• E2E testing with Playwright/WebDriver (when display server available)
• Multi-machine Phase A: extract bterminal-core crate
• Test agent teams with CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1