Skip to content

Releases

v0.3.4

Choose a tag to compare

@lwgray lwgray released this 17 May 15:12
b8ff90e

Bugfix release.

Fixed

  • Hex project IDs with all-digit prefixes loaded only a few tasks. Marcus projects use 32-character hexadecimal IDs. The fast task-loading path checked whether a project was a numeric Planka project by inspecting only the first 8 characters of its ID. A hex ID whose first 8 characters were coincidentally all decimal digits (e.g. 10519982...) was misclassified as a Planka project, routing it down a prefix-matching path that silently dropped every hex task ID. Affected projects showed only the 2-4 tasks recovered from conversation logs in the DAG, Swim Lane, and Board views instead of the full task list. The check now requires the entire ID to be numeric.

This release also brings main current with the v0.3.3 cost dashboard, which had not previously been merged to main.

v0.3.0 — Quality & Observability

Choose a tag to compare

@lwgray lwgray released this 22 Apr 19:30
0fa898c

The Quality & Observability release. Adds Epictetus quality dashboard, display_role task classification, design phase visibility in Swim Lanes, and Project Info drawer. Requires Marcus >= 0.3.0 for quality_assessments.

Added

  • Quality dashboard tab — visualizes Epictetus code audit reports with sub-tabs: Agents (effectiveness bars, root cause attributions), Scores (dimension bars, smoke tests, coordination), Issues (ghost code, cross-agent, findings), Recommendations (instruction quality, project/global scope filtering)
  • Display role system — three-tier task classification (work/structural/context) computed by the aggregator, controlling which tasks appear in each visualization view
  • Design phase in Swim Lane — dedicated "Marcus / Planning" lane with dashed ghost-node styling
  • Ghost nodes in DAG — design/structural tasks rendered as hollow rings with dashed stroke
  • Project Info drawer — context tasks surfaced in a dedicated drawer
  • QualityAssessment dataclass and aggregator method to load audit reports from marcus.db

Changed

  • Quality tab hides metrics sidebar and timeline scrubber for full viewport width
  • Metrics exclude structural and context tasks from completion counts
  • Dependency graph excludes context tasks; structural tasks preserve DAG topology

Fixed

  • auto_completed label no longer misclassifies work tasks as context
  • Context tasks filtered from agent swim lanes

v0.2.1 - CI/CD & Code Quality

Choose a tag to compare

@lwgray lwgray released this 30 Mar 13:13
18e437a

What's New

Added

  • Pre-commit hooks: mypy, black, isort, flake8, pydocstyle, bandit, detect-secrets
  • GitHub Actions CI/CD workflows: quality checks, test suite, Claude code review, version gate
  • Configuration files: .flake8, .coveragerc, pytest.ini, .secrets.baseline
  • Expanded dev dependencies (pytest-asyncio, pytest-cov, isort, flake8, bandit, httpx)
  • Custom pre-commit hooks for naive datetime detection and changelog reminders

Fixed

  • All existing lint, type annotation, and formatting issues across the codebase
  • Setuptools flat-layout package discovery error in CI
  • Deprecated license format in pyproject.toml
  • Historical API integration tests now skip gracefully when Marcus is not installed

Full Changelog: v0.2.0...v0.2.1

Cato v0.1 - Live Marcus Visualization Dashboard

Choose a tag to compare

@lwgray lwgray released this 10 Nov 23:19
42af3a7

Cato v0.1 - Initial Release

Overview

Cato is a live visualization dashboard for Marcus multi-agent orchestration system. This initial release provides real-time monitoring and visualization of agent tasks, dependencies, conversations, and parallelization metrics.

Key Features

🔴 Live Data Mode

  • Real-time connection to Marcus backend API
  • Automatic 60-second background refresh
  • On-demand refresh when clicking project dropdown
  • Always shows newest/most recently active project

📊 Visualization Views

  • Network Graph: Task dependencies and relationships
  • Agent Swim Lanes: Timeline view of agent task execution with playback
  • Conversation View: Agent message history with expandable details
  • Health Dashboard: System metrics and status (in development)

🎮 Playback Controls

  • Timeline scrubbing to review project execution
  • Play/pause with adjustable speeds (0.5x - 10x)
  • Preserved playback position during data refresh
  • Linear timeline (1.0 power scale) for intuitive visualization

🔧 Project Management

  • Project dropdown with automatic newest selection
  • Real-time project list updates
  • Single-project filtering
  • Graceful handling of deleted projects

Technical Details

  • Frontend: React + TypeScript + Zustand
  • Backend: Python FastAPI
  • Visualization: D3.js for network graphs
  • State Management: Zustand with persistent playback state

UX Improvements in v0.1

  • ✅ Removed mock mode complexity (live-only)
  • ✅ Simplified header controls
  • ✅ Background auto-refresh without UI clutter
  • ✅ Automatic newest project selection on load
  • ✅ Project list refreshes when dropdown clicked
  • ✅ Fixed power scale consistency (0.4 → 1.0)

Known Issues

  • Health dashboard view incomplete (type errors)
  • Some pre-existing TypeScript warnings

Installation

See README.md for installation and setup instructions.

Next Steps

See GitHub issues for planned enhancements and feature requests.


🤖 Generated with Claude Code