CLI Evolution Phase 1: TUI, IDE plugins, KG, mesh, plan, extensions#11
Merged
ApiliumDevTeam merged 92 commits intodevfrom Mar 5, 2026
Merged
CLI Evolution Phase 1: TUI, IDE plugins, KG, mesh, plan, extensions#11ApiliumDevTeam merged 92 commits intodevfrom
ApiliumDevTeam merged 92 commits intodevfrom
Conversation
Implement auto-discovery of user-defined slash commands from .mayros/commands/*.md files with YAML frontmatter for description, argument hints, and allowed tools. Commands are injected into TUI autocomplete, help text, and dispatch (default case expands $ARGUMENTS before sendMessage). Project commands override user commands with the same name. 24 unit tests.
Implement auto-discovery of agent definitions from .mayros/agents/*.md files with YAML frontmatter for name, model, allowed-tools, workspace, and default flag. Agents are merged into listAllAgentEntries() and resolveAgentEntry() with config agents taking priority over markdown agents. Project agents override user agents with the same id. 15 unit tests.
Pass ctx.sessionKey through all observability hooks to the TraceEmitter so every trace event carries its session identifier. This enables session-scoped querying via DecisionGraph.buildFromSession() and the new mayros trace session command. All emit methods now accept an optional session parameter appended to the TraceEvent.
Register two new built-in subcommands in register.subclis.ts: - mayros trace: query, explain, aggregate, and inspect session decision trees from Cortex trace events. Subcommands: events, explain, stats, session, status. Connects directly to Cortex with fallback to plugin config for connection details. - mayros plan: semantic plan mode with four phases (explore, assert, approve, execute). Plan state is persisted as RDF triples in Cortex. Subcommands: start, explore, assert, show, approve, execute, done, list, status. Assertions can be verified via Cortex Proof of Logic.
New extension that scans TypeScript/JS files using regex patterns (consistent with skill-scanner.ts approach) and maps code entities to RDF triples for storage in Cortex. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
SHA-256 content hashing for change detection, plugin entry with code_index_query tool, CLI subcommands (run, status, query), and service registration. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
26 unit tests covering scanner regex extraction, RDF mapper namespace correctness, incremental hash computation, config parsing, and plugin metadata validation. Also fix unused import lint warnings. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
ProjectMemory class with convention/decision/finding storage and querying. New tools: project_convention_store, project_convention_query. Enhanced agent_end hook with auto-detection of project knowledge. Enhanced before_prompt_build with convention and finding injection. Added projectMemory config section. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
CompactionExtractor extracts typed knowledge from both user and assistant messages before context compaction: conventions, decisions, file changes, bug findings, and error patterns. 21 unit tests. Enhanced before_compaction hook to use structured extraction. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Knowledge graph CLI with 8 subcommands: search, conventions, decisions, code, explore, stats, status, explain. Provides unified access to personal memories, project conventions, code entities, and session findings. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Version bump reflecting Phase 2 additions: project memory, smart compaction, and cross-session recall capabilities. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Low-level git worktree operations for parallel agent isolation: createWorktree, removeWorktree, listWorktrees, pruneWorktrees, isWorktreePath, findWorktreeForPath. Uses execFileSync consistent with existing infra modules. Includes 17 tests with mocked git. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
TeamManager class for team lifecycle: create teams with shared namespaces, track member states, orchestrate merge via KnowledgeFusion. Follows PlanStore pattern with subject-per-team and delete-then-create updates. Includes 16 tests. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
WorkflowOrchestrator class with phase-sequential, agent-parallel execution model. Includes 3 built-in workflow definitions (code-review, feature-dev, security-review) and a registry for custom workflows. Cortex-backed state tracking with triple persistence. Includes 25 tests. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Extend agent-mesh plugin with 3 new tools (mesh_create_team, mesh_team_status, mesh_run_workflow) and team CLI subcommands (create, status, list, merge). Add TeamsConfig and WorktreeConfig types to config with parse functions and validation. Includes 12 new tests for config and registration. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
New subcli with 4 subcommands: run (execute a workflow), list (available definitions), status (run progress), and history (past runs). Reads Cortex config from agent-mesh plugin entry. Registered in register.subclis.ts for lazy loading. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Version bump reflecting Phase 3 additions: team manager, workflow orchestrator, built-in workflow definitions, mesh tools, team CLI, and workflow CLI. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Team manager: test for late-join member update, test for mixed member status merge (only completed members merged). Workflow orchestrator: test for mixed run states in listing, test for multi-phase aggregate computation, test for completion with no executed phases. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Command-level filtering for bash execution with domain allowlist, command blocklist, and dangerous pattern detection. Includes command parser with pipe/chain/subshell handling, 6 default dangerous patterns, and configurable enforce/warn/off modes. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Runtime permission dialogs with bash intent classification across 5 risk levels, policy persistence in Cortex, and audit trail. Supports exact, glob, and regex policy matching with auto-approve for safe operations. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Markdown-defined hooks in .mayros/hooks/ evaluated by LLM for policy enforcement. Includes safe recursive descent condition parser, response caching with session/global scopes, and dynamic hook registration. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Connect to MCP servers via 4 transports (stdio, SSE, HTTP, WebSocket), bridge MCP tools as Mayros tools, and register tool metadata in Cortex. Includes session lifecycle management with auto-reconnect. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Add dependency audit with 8 security scan rules, update checker for installed skills, category registry, hub_rate tool, and session_start hook for update notifications. Enhanced search with trust badges. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
WebSocket client connecting to Mayros gateway with JSON-RPC protocol, tree views for sessions/agents/skills, and webview panels for chat, plan mode, trace viewer, and knowledge graph browser.
Include tools/* in pnpm workspace packages for the VSCode extension.
Update bash-sandbox, interactive-permissions, llm-hooks, mcp-client, and vscode-extension versions to 0.5.0 for the ecosystem release.
- Add icon.svg and sidebar-icon.svg assets for VSCode extension - Add webview CSS files (chat, plan, trace, kg) with VS Code theme vars - Rename esbuild.config.mjs to .mts with proper type annotations - Split tsconfig: exclude webview from Node check, add tsconfig.webview.json - MCP client: switch from registerCommand to registerCli pattern - MCP client: add after_tool_call hook for centralized usage tracking
DependencyAuditor now persists audit reports as RDF triples in Cortex
with per-finding granularity. UpdateChecker persists version check
history. Both accept optional CortexClientLike in constructor with
graceful fallback when Cortex is unavailable.
Triple namespaces:
${ns}:skill:audit:${slug} — report summary + findings
${ns}:skill:update:${slug} — version check history
New RulesEngine class stores rules as RDF triples with hierarchical scope resolution (global → project → agent → skill → file). Rules are queryable, learnable, and priority-sorted. Includes CLI (mayros rules list/add/remove/learn/status), 21 tests, and sub-CLI registration. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
New AgentMemory class stores per-agent memories as RDF triples with type classification, project scoping, confidence tracking, and usage counting. Memories persist across sessions and are recalled by relevance. Adds memory: boolean field to MarkdownAgent for opt-in persistent memory per agent definition. Includes 15 tests. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Register code-cli as a subcli and wire no-subcommand fallback in run-main.ts so bare `mayros` launches the interactive TUI session. Includes 4 tests for option parsing and defaults.
Bare `mayros` now checks wizard.lastRunAt: if missing, redirects to `mayros onboard`; otherwise launches `mayros code`. Add onboard/code to ALLOWED_INVALID_COMMANDS so they work without valid config. Add 9 TUI slash commands (/plan, /kg, /trace, /team, /tasks, /workflow, /rules, /mailbox, /onboard) that delegate to the agent via structured messages, bridging Phase 2-5 features into the interactive session. Includes 3 first-run gate tests and 13 ecosystem command tests.
Implement `mayros batch run` for processing multiple prompts in parallel with configurable concurrency. Supports JSONL and text (--- separated) input formats, JSON-lines output, per-prompt timeout, and result streaming. Includes 18 tests covering input parsing and type validation. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Implement `mayros teleport export/import/inspect` for portable session bundles. Exports session transcript (base64), store metadata, and Cortex triples as a single JSON file. Import restores all components with optional session key remapping. Includes 19 tests covering export, import, validation, and round-trip integrity. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
New extension with pull-based delta sync protocol between Cortex instances. Includes sync-protocol (conflict detection, LWW resolution, reconciliation), peer-manager (Cortex-backed peer state, lifecycle), plugin entry (3 tools: sync_status/sync_now/sync_pair, auto-sync hooks), and sync CLI (status/pair/remove/now). 36 tests covering protocol logic, conflict strategies, and peer management. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Extends ProjectMemory with ingestMayrosMd() to parse markdown sections into Cortex triples (conventions, key files, build commands, decisions). Adds CortexClient.listTriplesSince() for timestamp-based delta queries. Integrates sync pairing into onboarding wizard and registers /batch, /teleport, /sync TUI slash commands with their handlers.
Introduces IntelliJ Platform plugin for Mayros with Gradle build config, WebSocket JSON-RPC client mirroring the VSCode extension protocol, application-level service for connection lifecycle, chat tool window with message history and streaming, and persistent settings state.
Completes the IntelliJ plugin with agents tree view, trace events table panel, quick settings panel, send-selection and explain-code editor context actions, TODO/FIXME gutter line markers, and WebSocket client unit tests.
Use execute() with details return instead of handler(), return content as typed array, and replace registerHook() with api.on() for hooks.
batch-cli: clear timeout handle to prevent memory leak in batch ops teleport: fix transcript filename on remap, use replaceAll for triple subject remapping, add Array.isArray guard in validateBundle, fix ~ path expansion with os.homedir() fallback sync-protocol: guard against NaN dates in resolveConflict and buildLocalDelta sort, change > to >= for inclusive delta boundary, simplify redundant reconcile condition cortex-client: fix listTriplesSince off-by-one (> to >=) + NaN guard project-memory: add fallback for empty slugify result, relax convention detection to sectionDepth >= 1 onboarding: add URL format validation for peer endpoint and alphanumeric check for peerId jetbrains: fix Timer leak in reconnect, clear event listeners on reconnect to prevent duplicates, use consistent session keys in editor actions, add Logger for WebSocket parse errors, add clearEventListeners
- Replace global clearEventListeners() with per-panel unsubscribe pattern (ChatPanel and TracesPanel track their own listeners) - Remove MayrosClient.clearEventListeners() (no longer needed) - Add randomUUID suffix to slugify fallback for collision safety - Add skipSync to onboarding test that was failing from sync step - Document inclusive >= boundary design decision in sync-protocol
…solation - project-memory: add continue after non-build cmdMatch to prevent bulletMatch fallthrough; normalize CRLF before splitting lines - teleport: use Buffer byte length for transcriptSize instead of string length; guard NaN port with Number.isFinite fallback - onboarding: use validated endpoint variable instead of dead code; trim peerId once into local variable - onboarding tests: add skipSync to remaining tests; mock onboard-mcp - JetBrains MayrosClient: @volatile on shared state fields; CopyOnWriteArrayList for event listeners; cancel reconnectTimer in disconnect(); snapshot pending requests before clear - JetBrains MayrosService: CopyOnWriteArrayList for listeners; @volatile client; @synchronized connect() - JetBrains panels: clear listener list on reconnect instead of calling off() on wrong (new) client instance
…nd JetBrains Fix output ordering in batch CLI, atomic session store writes in teleport, null-byte conflict keys and stable JSON comparison in sync protocol, sort-before-limit in project memory queries, ScheduledExecutorService for reconnect in JetBrains client, EDT-safe listener notifications, and project-scoped gutter session keys.
…ple remapping Clear timeout timer in syncWithPeer via try/finally, filter stats by active status only, add Disposable to ChatPanel and TracesPanel with listener cleanup, remap projectMemory triples and node references on teleport import, graceful executor shutdown in MayrosClient, safe asString in TracesPanel, sync-cli error handling, and peer ID validation feedback in onboarding.
Sync 46 extension packages, VSCode extension, and JetBrains plugin to version 0.1.4. Add missing SVG icon for JetBrains tool windows, add @service annotation to MayrosSettings for proper IDE lifecycle.
The cortex-sync extension lacked a mayros.plugin.json file, causing the manifest registry to emit error diagnostics during bundled plugin discovery. This polluted config validation results and caused two plugin-validation tests to fail. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
VSCode and JetBrains plugins are being published for the first time, so they start at 0.1.0 independently of the core Mayros version. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
…ugin Add Gradle 8.12 wrapper (gradlew) so the plugin can be built without a system Gradle install. Set Java source/target compatibility to 17 to match the Kotlin jvmTarget and IntelliJ Platform requirements. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
…xtension Port JetBrains-exclusive features: "Explain Code" and "Send Selection" context menu actions with editor selection support, CodeLens gutter markers for TODO/FIXME/HACK/mayros: comments, and upgrade the WebSocket client to gateway protocol v3 with challenge-response handshake.
…tBrains plugin Consolidate six separate tool windows into a single tabbed panel with a setup screen for disconnected state and auto-detection of gateway token from ~/.mayros/mayros.json. Add Skills, Plan, and Knowledge Graph panels, agent filter on Traces, HTML chat bubbles with theme-aware colors, and upgrade MayrosClient to gateway protocol v3 with challenge-response handshake.
Rename `mayros dashboard` → `mayros team-dashboard` and `mayros doctor` → `mayros diagnose` for clarity. Update ci-plugin and lsp-bridge workspace dependencies to use @apilium/mayros scope. Add gitignore entries for JetBrains and VSCode extension build artifacts.
Gateway propagates isHeartbeat flag on chat events so TUI can suppress heartbeat noise. History loader strips heartbeat prompt/response pairs. Slash commands without arguments now open interactive select overlays instead of printing usage text (/think, /verbose, /reasoning, /elevated, /activation, /style, /theme). Deduplicate /agents→/agent, /sessions→ /session, /models→/model aliases and remove /elev, /reset, /quit. Add elevatedLevel, groupActivation, and PendingImage to TUI state types.
New WelcomeScreen component with shield mascot avatar in golden gradient and two-column bordered panel showing quick-start tips and session info. Shown on first connection, subsequent session switches show plain text. Responsive single-column fallback for terminals narrower than 70 cols. Add Ctrl+V image paste support via clipboard capture — images are encoded as base64 attachments and sent alongside chat messages. Editor inserts an [Image #N] marker at cursor position. Code CLI now shows onboarding hint on first run and supports --clean flag to start with a blank chat while preserving session history. Gateway disconnect handler distinguishes ECONNREFUSED with a helpful start hint, and local pairing requests are auto-approved inline.
- hono 4.11.10 → 4.12.4 (cookie injection, SSE injection, serveStatic) - @hono/node-server 1.19.9 → 1.19.10 (encoded slash bypass) - dompurify ^3.3.1 → ^3.3.2 (XSS vulnerability) Add hono as direct dependency to ensure pnpm override propagates to transitive peer resolution through @buape/carbon.
Document all new capabilities in README: Terminal UI, IDE plugins, Knowledge Graph, Multi-Agent Mesh, Plan Mode, Extensions Ecosystem, Hooks System, and updated architecture diagram. Add .vscodeignore for clean VSCode extension packaging. Gitignore internal docs. Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
ApiliumDevTeam
added a commit
that referenced
this pull request
Mar 9, 2026
) ## Summary - **Terminal UI** — interactive TUI (`mayros code`) with 30+ slash commands, 3 themes, vim mode, image paste, headless CLI (`-p`), markdown agents/commands - **IDE plugins** — VSCode extension (sidebar + 4 webview panels) and JetBrains plugin (unified tabbed panel), both via Gateway WebSocket - **Knowledge Graph** — code indexer (incremental TS/JS scanning → RDF), project memory, smart compaction, cross-session recall - **Multi-Agent Mesh** — team manager, workflow orchestrator (3 built-in), agent mailbox, background tracker, git worktree isolation - **Plan Mode** — Cortex-backed semantic planning (explore → assert → approve → execute) - **Extensions** — bash-sandbox, interactive-permissions, llm-hooks, mcp-client, code-indexer, cortex-sync, CI/CD, LSP bridge - **Hooks** — 29 hook types, HTTP webhook dispatcher (HMAC-SHA256), async queue with dead-letter - **CLIs** — trace, plan, kg, workflow, dashboard, session, tasks, mailbox, rules, batch, teleport - **Security** — 18-layer architecture maintained across all new features - **README** — updated with all new capabilities, architecture diagram, extensions table ## Test plan - [ ] `pnpm build` compiles without errors - [ ] `pnpm test` passes (~9200+ tests across 1035+ files) - [ ] `mayros code` launches TUI with welcome screen - [ ] `mayros -p "hello"` streams response in headless mode - [ ] VSCode extension packages cleanly (`cd tools/vscode-extension && pnpm package`) - [ ] JetBrains plugin builds (`cd tools/jetbrains-plugin && ./gradlew build`) - [ ] `mayros kg stats` returns Cortex graph statistics - [ ] `mayros plan start "test"` creates a new plan - [ ] `mayros trace status` reports Cortex connectivity - [ ] `mayros workflow list` shows built-in workflow definitions - [ ] README renders correctly on GitHub (tables, code blocks, diagram)
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
1 participant
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
mayros code) with 30+ slash commands, 3 themes, vim mode, image paste, headless CLI (-p), markdown agents/commandsTest plan
pnpm buildcompiles without errorspnpm testpasses (~9200+ tests across 1035+ files)mayros codelaunches TUI with welcome screenmayros -p "hello"streams response in headless modecd tools/vscode-extension && pnpm package)cd tools/jetbrains-plugin && ./gradlew build)mayros kg statsreturns Cortex graph statisticsmayros plan start "test"creates a new planmayros trace statusreports Cortex connectivitymayros workflow listshows built-in workflow definitions