What's New

Full release history — every version's added features, fixes, and changes.

What’s New in Mr. Pumpkin

All notable changes to Mr. Pumpkin will be documented in this file.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

[0.5.13] - 2026-03-06

Added

  • Audio lip-sync recording tool (Issue #66) — analyze an audio file with Google Gemini and automatically generate a timeline with mouth viseme commands timed to speech.
    • skill/lipsync_cli.py — new CLI: python -m skill.lipsync_cli audio.mp3 --filename my_recording
    • skill/audio_analyzer.py — GeminiAudioProvider uploads audio to the Gemini File API, extracts per-word timing and emotion in two passes
    • Auto-generated timeline contains expression commands (from detected emotion) and mouth viseme commands (mouth_open, mouth_wide, mouth_rounded, mouth_closed, mouth_neutral) timed to individual words
    • Audio file is uploaded to the pumpkin server alongside the timeline — playback is fully synchronized
    • Supports .mp3, .wav, .ogg, .m4a, .aac, .flac for analysis; warns with ffmpeg hint for formats pygame cannot play back
    • --dry-run mode for preview without upload

Fixed

  • Gemini SDK v1.x API changes: files.upload() kwarg and .uri vs .name for Part.from_uri() calls in audio_analyzer.py
  • Mouth viseme commands wired into _execute_timeline_command() in pumpkin_face.py (were in vocabulary but not dispatched during timeline playback)
  • wiggle_nose timeline playback dispatch fixed (now aliases to _start_nose_twitch)
  • upload_audio handlers in pumpkin_face.py and timeline.py now accept .m4a, .aac, .flac

[0.5.12] - 2026-03-05

Added

  • Squad v0.8.20 upgrade — 12 new GitHub Actions workflow templates for squad automation (CI gating, label enforcement, branch protection, issue assignment, triage, preview/promotion/release pipelines, and label sync)
  • squad.config.ts — project-level Squad configuration covering model tiers, routing rules, casting universe allowlist, and platform settings
  • Blog post: “Built With Squad: An Entire Creative Coding Project Delivered by an AI Team” — the full development story of Mr. Pumpkin, including how parallel fan-out, reviewer gates, and persistent agent memory shaped the project. See blog for the full post.
  • docs/development-story.md — narrative of how Mr. Pumpkin was built by the AI team, covering architectural milestones and squad collaboration patterns
  • Updated squad identity, casting state, and project-conventions skill with session learnings

Fixed

  • mouth_wide viseme shape refined: increased grin upturn, curved arc distinct from neutral mouth, corrected width
  • Documentation site blog listing broken due to missing index.html required by jekyll-paginate
  • CI: added x86_64-linux platform to Gemfile.lock to resolve Linux build failures
  • Docs site nav layout: corrected search bar width overflow and mobile hamburger menu display (#61, #62)

[0.5.11] - 2026-03-04

Added

  • GitHub Pages static website — full Jekyll site deployed via GitHub Actions at https://mlinnen.github.io/mr-pumpkin/
    • Pages: Home (hero + stats), Features, Installation & Usage, Documentation, Blog, Contributing
    • Dark pumpkin theme with #FF6B00 orange accents; mobile-responsive with hamburger nav
    • Two-row header: 🎃 logo (left, links home) + search on top row; nav tabs on bottom row
    • On-site search powered by Lunr.js — no redirect to GitHub, searches content locally in the browser
    • Blog section with paginated post listing
  • Mouth speech control (Issue #59) — independent viseme-based mouth animation orthogonal to face expressions. Send viseme commands at ~20Hz from audio analysis to drive lip sync.
    • 5 viseme shapes: closed (M/B/P sounds), open (AH/AA sounds), wide (EE/IH sounds), rounded (OO/OH sounds), neutral (releases back to expression-driven mouth)
    • 6 commands: mouth_closed, mouth_open, mouth_wide, mouth_rounded, mouth_neutral, and mouth <viseme> (parameterized form)
    • Smooth animated transitions between shapes at configurable speed
    • See Timeline Schema for full command reference.

[0.5.10] - 2026-03-03

Added

  • Recording chaining via play_recording command (issue #55) — embed one recording inside another using stack-based nested playback; maximum nesting depth is 5 levels. See Timeline Schema for details.
  • help command over TCP and WebSocket (issue #56) — returns a plain-text listing of all available commands with syntax; safe to call at any time including during timeline playback.

[0.5.9] - 2026-03-02

Added

  • docs/what-is-new.md — full release history page linked from docs/index.md
  • docs/timeline-schema.md — full reference for timeline JSON format, command vocabulary, and timing rules

Fixed

  • CI release workflow now extracts the current version’s CHANGELOG section as the release notes body; falls back to --generate-notes if no entry found
  • CI release workflow now patches pre-existing GitHub releases with updated notes via gh release edit
  • CI release notes now correctly diff from the previous semver tag using --notes-start-tag

[0.5.8] - 2026-03-02

Added

  • AI-powered recording skill (skill/ package) — describe an animation in plain English and have it generated and uploaded automatically using Google Gemini
    • skill/cli.py — command-line interface (python -m skill "make the pumpkin look confused")
    • skill/generator.py — LLM-backed timeline generation with pluggable provider interface
    • skill/uploader.py — TCP upload client with dry-run and validation support
    • skill/timing_guidelines.md — LLM-facing timing reference used during generation
  • docs/recording-skill.md — user guide for the recording skill
  • docs/timeline-schema.md — full reference for the timeline JSON format, command vocabulary, and timing rules
  • Linked both new docs from docs/index.md

Fixed

  • TCP upload deadlock on Windows loopback — Nagle’s algorithm was batching the JSON payload and END_UPLOAD sentinel into one TCP segment, causing the server’s recv loop to miss the sentinel. Fixed with a line-buffered accumulation loop in pumpkin_face.py
  • CI release workflow now handles pre-existing GitHub release tags — uses gh release upload --clobber instead of failing on duplicate gh release create
  • CI release notes now correctly diff from the previous semver tag (not the previous tag on main) using --notes-start-tag

[0.5.7] - 2026-03-03

Added

  • Expanded documentation coverage with WebSocket examples and detailed integration guide
  • End-user documentation suite in docs/ folder
    • docs/index.md — documentation hub and quick-start guide
    • docs/installation.md — running Mr. Pumpkin as a system service (Linux systemd, Windows Task Scheduler/NSSM)
    • docs/building-a-client.md — comprehensive client integration guide with TCP and WebSocket protocols, full command reference, Python/Node.js/C# examples, timeline format, and troubleshooting

[0.5.6] - 2026-03-02

Added

  • End-user documentation in docs/ folder (issue #53)
    • docs/index.md — documentation hub and quick-start guide
    • docs/installation.md — running Mr. Pumpkin as a system service (Linux systemd, Windows Task Scheduler/NSSM)
    • docs/building-a-client.md — comprehensive client integration guide with TCP and WebSocket protocols, full command reference, Python/Node.js/C# examples, timeline format, and troubleshooting

[0.5.5] - 2026-03-02

Added

  • update.sh (Linux/macOS/Raspberry Pi) and update.ps1 (Windows) auto-update scripts
  • 5-phase update workflow: check version → download → stop → deploy → restart
  • GitHub API version check with gh CLI preferred, direct URL download fallback
  • Process detection via command line pattern match — preserves original launch arguments
  • Graceful stop with SIGTERM/Stop-Process, 5s timeout, force kill fallback
  • ZIP validation before overwriting files; logs all operations to mr-pumpkin-update.log
  • Suitable for cron job (Linux/Pi) or Windows Task Scheduler scheduling
  • docs/auto-update.md — detailed platform setup guide
  • README Auto-Update section with cron/Task Scheduler examples

[0.5.4] - 2026-03-01

Fixed

  • upload_timeline now works correctly from the WebSocket test client — WebSocket clients send filename and JSON inline in a single message, but the handler was routing through command_router which returned a TCP-only UPLOAD_MODE stub. _ws_handler now intercepts upload_timeline directly and calls file_manager.upload_timeline() to save the file.

[0.5.3] - 2026-03-01

Added

  • wiggle_nose command: alias for nose twitch animation, callable from TCP client
  • wiggle_nose added to timeline recording capture whitelist
  • 21 test cases covering wiggle_nose command routing, alias equivalence, and recording integration

[0.5.2] - 2026-03-01

Fixed

  • Expression commands now work correctly from TCP client — resolved Python __main__ vs module name circular import issue where command_handler.py imported Expression as a second class instance, causing enum equality checks to silently fail
  • Happy and sad mouth curves were visually swapped — corrected sine curve sign in _get_mouth_points (pygame Y increases downward; smile needs +, frown needs -)
  • TCP recv deadlock in client_example.py — added socket.shutdown(SHUT_WR) after send so the server’s blocking recv() unblocks for no-response commands (blink, roll, gaze, etc.)

[0.5.1] - 2026-03-01

Added

  • Version management via VERSION file for centralized version control

Changed

  • Replaced hardcoded version string with dynamic version loading from VERSION file
  • Squad upgraded from v0.5.3 to v0.5.4

[0.5.0] - 2026-02-27

🎃 Major Features

WebSocket Dual-Protocol Support

  • Added WebSocket server on port 5001 running parallel to TCP server (port 5000)
  • Both protocols share unified CommandRouter for protocol-agnostic command execution
  • Browser client support via standard JavaScript WebSocket API
  • Zero breaking changes — fully backward compatible with existing TCP clients

Added

  • WebSocket Server: Asyncio-based WebSocket server runs in daemon thread alongside TCP server
  • CommandRouter: Extracted 660-line command routing logic into shared command_handler.py module
  • Browser Test Client: websocket-test-client.html for interactive WebSocket validation
    • Connection status monitoring with automatic fallback
    • Quick test buttons for common commands
    • Timeline upload testing (multi-line JSON)
    • Full event logging with timestamps
  • Integration Tests: 27 comprehensive tests validating dual-protocol behavior
    • Identical response verification across protocols
    • Protocol switching and concurrent command execution
    • Error handling consistency
    • State synchronization (recordings, playback, expressions)
    • Connection resilience and large payload handling
    • Stress testing (50 rapid alternating commands)

Architecture

Milestone Implementation:

  • M1: Command router extraction (660-line refactor, 403/403 tests passing)
  • M2: WebSocket server setup (asyncio daemon thread, port 5001)
  • M3: Browser test client (HTML/JS validation tool)
  • M4: Integration testing (27 comprehensive test scenarios)
  • M5: Production release (v0.5.0 tagging and documentation)

Performance

  • Zero Regressions: All 403 existing tests continue to pass
  • New Test Coverage: 430 total tests (403 existing + 27 new integration tests)
  • Sub-millisecond Overhead: Command routing overhead negligible
  • Concurrent Client Support: Multiple TCP and WebSocket clients simultaneously

Documentation

  • Updated README with complete WebSocket usage examples
  • Documented all 27 integration test scenarios
  • Added browser test client documentation
  • Installation instructions unchanged (no new dependencies beyond websockets package)

Breaking Changes

None — This release is fully backward compatible. Existing TCP clients work without modification.

Upgrade Path

No migration required. Simply update to v0.5.0:

git pull origin main
pip install -r requirements.txt  # Installs websockets package
python pumpkin_face.py           # Both TCP and WebSocket servers start automatically

Contributors

  • Vi: Backend architecture (CommandRouter extraction, WebSocket server implementation)
  • Mylo: Integration test suite (27 comprehensive test scenarios, browser client validation)
  • Jinx: Milestone coordination, architectural review, release management

[0.4.1] - 2026-02-26

Added

  • Recording file upload with validation (#44)
  • Download timeline command handler
  • Playback controls and recording/list bug fixes (#42)

Changed

  • Systematic race condition delays in recording tests
  • Updated documentation for recording file structure

Fixed

  • Recording upload validation
  • Playback control edge cases

Earlier Versions

See git history for releases prior to 0.4.1.