v0.6.1

🚀 LoopTroop — What's New Since v0.6.0

A big round of bug fixes, stability improvements, and quality-of-life updates!

1. 🛑 Stop & Retry AI Models — Stop or restart slow, stuck, or blocked models individually during multi-model council rounds, without restarting everyone else.
2. 🔍 Better Error Pages — More detailed errors showing what went wrong, which model failed, and useful debugging information.
3. 🔄 Smoother Error Recovery — Fix an issue, retry or continue, and return to your workflow without getting stuck on the error screen.
4. 🛠️ More Reliable OpenCode Startup — Fixed startup failures, port conflicts, and process management issues.
5. 🤖 Improved Model Detection — More reliable model loading, refreshing, and provider detection.
6. 🐛 Fewer Unexpected Failures — Fixed OpenCode v2 issues that could incorrectly block or interrupt workflows.
7. 🧠 Smarter AI Response Handling — Better handling of incorrectly formatted model responses, reducing unnecessary failures.
8. ⏭️ Skip All AI Questions — Skip multiple pending questions across different models in one click, with improved timeout settings.
9. 💻 Better Windows & Linux Support — Fixed installation, compatibility, and reliability issues across platforms.
10. ⚡ Up to 20× Faster CLI — Much faster commands, a smaller package, and lighter builds. #259

✨ Overall: Faster, more stable, easier to recover from errors, and more control over your AI models!

Summary

• You can stop or retry one council model while the others keep drafting or voting, instead of waiting for a slow model's timeout.
• Error page recovery actions (Retry, Continue, Retry with extra note, Edit setup plan, and Cancel) automatically return the workspace to the resumed workflow status or canceled view and clear active error selection.
• OpenCode v2 scans and prompts tolerate private bookkeeping events safely and avoid automatic title calls that can block a finished response; OpenCode v1 remains supported.
• Ticket errors name failed readiness checks and show the five newest debug records, each collapsed in Technical details; live and saved debug logs redact credentials in every phase.
• OpenCode communication errors explain the failed operation and available cause, and coverage logs distinguish checks, response attempts, and reconnections.
• Interview batches recover safely when a model mixes field tags with YAML, preserving question text and recording the corrections in the ticket log.
• AI question actions stay visible, with a Skip all option for questions across every model tab.
• Development dependency updates fix command injection and denial-of-service vulnerabilities.
• AI question settings share Advanced rows across Configuration, Project, and New Ticket, with a nested wait that preserves edits and prevents invalid saves or starts.
• LoopTroop prefers configured and home OpenCode installations over PATH and reports which executable it selected.
• OpenCode v2 model discovery waits for providers; Configuration keeps model lists live, refreshes slow discovery, shows model counts, and collapses providers in the results.
• Published-install checks avoid Windows installer and startup hangs, keep PowerShell modules and OpenCode probes working, and apply CI repairs to existing releases.
• Windows PowerShell commands and installers load their standard modules when launched from PowerShell 7 through LoopTroop.
• Release publication waits for npm's package scans, includes the executable resolver needed to publish container tags, and reads npm metadata correctly during recovery and channel repair.
• LoopTroop starts OpenCode from ~/.opencode/bin again on Ubuntu, Fedora and other Linux desktops, where the OpenCode installer creates that folder group-writable.
• When another server already holds OpenCode's default port, LoopTroop starts its own OpenCode on the next free port and tells you where it went.
• looptroop doctor compares your Node with the newest release of your own Node line, says what gh is still needed for, and names the checks that actually stop LoopTroop from running.
• looptroop commands such as status and --help answer in about a third of a second instead of several seconds, and LoopTroop installs about 30 MB less.
• Token counts shown in the interface are now a quick estimate of four characters per token.

Container image

docker pull looptroopai/looptroop:0.6.1
docker pull ghcr.io/looptroop-ai/looptroop:0.6.1

Both registries serve the same linux/amd64 + linux/arm64 index,
sha256:a65238ec759a85a86c18f424a279fde806c79b6565a5fff407cc86da15794279.

The image ships without OpenCode, so a bare docker run exits at
startup by design. See Run it in a container
for how to point it at a server and reach the interface.

Full changelog