Super-Goose v1.24.06 — AG-UI Event Bus + GPU/Studios/Marketplace

Super-Goose v1.24.06 — AG-UI Event Bus + Multi-Panel Wiring

Highlights

• AG-UI Event Bus — Broadcast channel (capacity=4096) with SSE multiplexing via tokio::select!
• GPU Detection Backend — nvidia-smi parsing with /api/system/gpu endpoint
• 3 New Panels — GPUPanel, StudiosPanel, MarketplacePanel with full API wiring + test suites
• AgentPanelContext Rewrite — All mock data removed, wired to live AG-UI stream
• Coach Review Enhancements — Error line counting + quality threshold enforcement
• 3-Agent Review — 7 post-session fixes for race conditions, ARIA, AbortController

What's New

#	Feature	Description
1	AG-UI Event Bus	Broadcast channel in AppState, SSE handler with tokio::select!
2	GPU Detection	nvidia-smi parsing backend + GPUPanel (Cluster/Jobs/Launch tabs)
3	Studios Panel	6 studio definitions, grid view, filter tabs, keyboard navigation
4	Marketplace Panel	5 tabs (Browse/My Cores/Sell/Extensions/Community), wired to core APIs
5	AgentPanelContext	Rewritten from mock data to live AG-UI stream via useAgUi()
6	Connector Builder	buildConnectors() merging extensions + well-known connectors
7	Coach Quality	Error line counting + min_quality_score threshold
8	Hook Handlers	Prompt & Agent handler implementations with formatters
9	StatusBar Migration	Mode selector with ARIA landmarks
10	Particle Optimization	MAX_PARTICLES=120, visibility-based animation loop

3-Agent Review Fixes

• Broadcast channel 256 to 4096 capacity (burst event safety)
• Unwrap to expect in HTTP production path
• MarketplacePanel race condition (Promise.all consolidation)
• GPUPanel AbortController (signal, abort on unmount, retry)
• StatusBar ARIA landmarks
• StudiosPanel focus restoration
• Eliminated last cargo warning

Code Stats

• 45 files changed, 6508 insertions, 712 deletions
• Vitest: 234 files, 3263 passed, 3 skipped, 0 failed
• tsc: CLEAN (0 errors)
• cargo check: CLEAN (0 warnings)
• Playwright E2E: 291 passed

Multi-Platform Builds

Trigger the Multi-Platform Release workflow from the Actions tab to generate:

• Windows: .exe installer + .zip portable
• macOS: .dmg (ARM) + .dmg (Intel)
• Linux: .deb + .rpm

Full Changelog

v1.24.05...v1.24.06

Super-Goose v1.24.05

Super-Goose v1.24.05 — 12 SOTA Agent Features

Highlights

• Native Extended Thinking — --thinking flag enables Anthropic's chain-of-thought reasoning with configurable budget
• Prompt Caching — 80-90% input cost reduction via Anthropic cache_control breakpoints (system prompt, tools, conversation history)
• Dead Code Removal — 4,454 lines removed (MCP Gateway module + stale tests)

12 Features

#	Feature	Description
1	Closed-Loop Reflexion	Agent learns from past failures via episodic memory
2	Active Budget Enforcement	--budget flag with real-time CostTracker
3	Bidirectional Output Guardrails	Scan LLM responses for secrets/PII
4	Structured Code-Test-Fix Loop	ShellTestRunner with real callbacks
5	/model Hot-Switch	Live model switching via Config::set_goose_model
6	Adaptive Compaction Manager	Context window management with stats
7	Cross-Session Search	goose session search command
8	Smart Project Auto-Detection	Rust/Node/Python/Go detection
9	Per-Tool Rate Limiting	Backpressure-aware rate limiter
10	Session Bookmarks	/bookmark save/list/restore/delete
11	Native Extended Thinking	Anthropic thinking blocks with --thinking flag
12	Prompt Caching Visibility	Cache hit/miss metrics in CostTracker

Code Changes

• 5 commits on main since v1.24.04
• 12 features verified as fully wired
• 4,454 lines of dead code removed
• Zero compiler warnings

Docker

docker pull ghcr.io/ghenghis/super-goose:v1.24.05
docker run -it ghcr.io/ghenghis/super-goose:v1.24.05 goose session

Full Changelog

v1.24.04...v1.24.05

v1.24.03 — Phase 6 Memory System + Candle Embeddings

Super-Goose v1.24.03

What's New — Phase 6: Memory System

• 4-Tier Memory Architecture — Working (0.70 decay), Episodic (0.90), Semantic (0.99), Procedural (0.98)
• Real Sentence Embeddings — all-MiniLM-L6-v2 via Candle/HuggingFace Hub (384-dim vectors)
• Hash Embedding Fallback — Deterministic hash-based embeddings when model unavailable
• Mem0 Dual-Write — Simultaneous local + external graph memory storage
• /memory Slash Command — stats, clear, save subcommands for runtime inspection
• 127 Memory Tests — All passing in release mode
• Project Scale — 161K+ lines of Rust, ~2,000 tests, 171 commits ahead of upstream

Memory Tiers

Tier	Decay	Description
Working	0.70	Active context, fast decay
Episodic	0.90	Session memories, event-based
Semantic	0.99	Long-term knowledge, vector search
Procedural	0.98	Learned patterns, skill memory

Docker

# Latest
docker pull ghcr.io/ghenghis/super-goose:latest

# This release
docker pull ghcr.io/ghenghis/super-goose:v1.24.03

# Run the CLI
docker run --rm -it ghcr.io/ghenghis/super-goose:latest goose

# Run the server (goosed) on port 3284
docker run -d -p 3284:3284 ghcr.io/ghenghis/super-goose:latest goosed

Downloads

Platform	CLI	Desktop
Docker	ghcr.io/ghenghis/super-goose:v1.24.03	—
Windows	goose-x86_64-pc-windows-msvc.zip	Goose-win32-x64.zip
macOS ARM	goose-aarch64-apple-darwin.tar.bz2	Goose.zip
macOS Intel	goose-x86_64-apple-darwin.tar.bz2	Goose_intel_mac.zip
Linux x86	goose-x86_64-unknown-linux-gnu.tar.bz2	.deb / .rpm
Linux ARM	goose-aarch64-unknown-linux-gnu.tar.bz2	—

Release Notes

• Docker image: ghcr.io/ghenghis/super-goose:v1.24.03 (linux/amd64)
• All 12/12 Release jobs + 3/3 Multi-Platform jobs + Docker: ALL GREEN
• Code signing disabled until certificates are configured
• Built from commit 68e78329dd on main
• Documentation: https://ghenghis.github.io/Super-Goose/

v1.24.02 — Stage 6 Integration + All-Platform Release

Super-Goose v1.24.02

What's New

• Stage 6 Integration — 16-tool governance layer with ResourceCoordinator (35 files, 17,811 lines)
• All-Platform Release Pipeline — builds for Windows, macOS (ARM + Intel), Linux (x86 + ARM)
• Docker Image — published to GitHub Container Registry (GHCR)
• Visual README Redesign — SVG artwork showing how all 16 tools work end-to-end
• GitHub Pages — full documentation site at https://ghenghis.github.io/Super-Goose/

Docker

Pull the official Super-Goose Docker image from GitHub Container Registry:

# Latest
docker pull ghcr.io/ghenghis/super-goose:latest

# This release
docker pull ghcr.io/ghenghis/super-goose:v1.24.02

# Run the CLI
docker run --rm -it ghcr.io/ghenghis/super-goose:latest goose

# Run the server (goosed) on port 3284
docker run -d -p 3284:3284 ghcr.io/ghenghis/super-goose:latest goosed

Image includes both goose (CLI) and goosed (server) binaries built from Rust in a minimal Debian runtime.

Downloads

Platform	CLI	Desktop
Docker	ghcr.io/ghenghis/super-goose:v1.24.02	—
Windows	goose-x86_64-pc-windows-msvc.zip	Goose-win32-x64.zip
macOS ARM	goose-aarch64-apple-darwin.tar.bz2	Goose.zip
macOS Intel	goose-x86_64-apple-darwin.tar.bz2	Goose_intel_mac.zip
Linux x86	goose-x86_64-unknown-linux-gnu.tar.bz2	.deb / .rpm
Linux ARM	goose-aarch64-unknown-linux-gnu.tar.bz2	—

Release Notes

• Docker image: ghcr.io/ghenghis/super-goose (linux/amd64)
• Code signing disabled until certificates are configured
• All 12/12 release workflow jobs passed GREEN
• Built from commit aa1d5c9e3 on main

v1.23.0 - Production Quality System Release

Goose v1.23.0 - Production Quality System Release

🎉 Major Features

Production Quality Enforcement System

• ✅ SonarQube Community Edition integrated and operational
• ✅ Zero Tolerance Quality Gate (0 blockers, 0 critical issues)
• ✅ Git Hooks blocking incomplete code (pre-commit + pre-push)
• ✅ CI/CD Pipeline with 4-stage quality enforcement
• ✅ Code Coverage generation (Vitest v8 + Cargo llvm-cov)
• ✅ 298/301 tests passing (99% pass rate)

AI Context System

• ✅ Epic .goosehints - Universal AI assistant rules
• ✅ Enhanced AGENTS.md - Comprehensive project context
• ✅ Research-backed best practices from 2,500+ repositories

Quality Metrics

• Code Duplication: 0.0% ✅
• Security Rating: A ✅
• Reliability Rating: A ✅
• Maintainability Rating: A ✅

📦 Downloads

Portable Version (Recommended)

• Unzip and run Goose.exe
• No installation required
• 192 MB download, 521 MB extracted

🚀 Quick Start

1. Download Goose-v1.23.0-win32-x64-portable.zip
2. Extract to any folder
3. Run Goose.exe

📖 Documentation

See full release notes in repository

v1.24.0 - LM Studio + Security Fixes

What's New

🔧 Fixed

• LM Studio Provider: Fixed all 8 compilation errors - now fully functional
  • ApiClient authentication with proper AuthMethod enum
  • CLI entry point corrections
  • Computer Use integration fixes

🔒 Security

• Fixed 158 Dependabot vulnerabilities (reduced to 9 stale alerts)
• Updated npm dependencies (tar 7.5.7, tmp 0.2.5, @modelcontextprotocol/sdk 1.26.0)
• Removed vulnerable temp test projects from repository

⬆️ Upstream Merges

• Merged 10 commits from block/goose main
• Gemini CLI parsing fixes
• Databricks fast models improvements
• Tailwind build path corrections

📝 Documentation

• Added Phase 7-8 comprehensive documentation for Claude Code

🎯 Technical Details

• 8 commits fixing LM Studio provider compilation errors
• 3 commits fixing Dependabot security alerts
• 1 merge commit with 10 upstream improvements

Full Changelog: v1.23.0...v1.24.0 (83 commits)

Built with ❤️ by Ghenghis
Co-Authored-By: Claude Sonnet 4.5 noreply@anthropic.com

Stable

stable

docs: updating guides to provide more clarity around moderator (#628)