wt

The wt skill provides a robust management interface for LlamaFarm developers who need to run multiple, isolated instances of the platform simultaneously. By leveraging git worktrees, this tool creates fully autonomous development environments where each session maintains its own unique service ports, data directories, and background processes. This architecture effectively eliminates resource contention and port conflicts that typically occur when running multiple agent instances, making it the ideal solution for developers iterating on complex features, testing RAG workflows, or managing parallel AI agent sessions without disrupting the primary development branch. Whether you are debugging the server, verifying runtime model inference, or testing custom tool-calling logic, wt ensures each environment is self-contained and reproducible.

• Automated lifecycle management: Create, start, stop, and prune worktrees using simple CLI commands.

• Deterministic networking: Automatic port assignment for Server, Designer, and Runtime services to prevent collision.

• Isolated state: Dedicated data directories for each worktree ensuring clean experimentation and testing.

• Diagnostic utilities: Built-in doctor commands for troubleshooting stale PIDs, port binding errors, and service health checks.

• Workflow optimization: Seamless integration with Git branching to jump between feature environments instantly.

• Use wt create --go feat/my-feature to initialize a fully operational, isolated workspace from scratch.

• Monitor active services using wt list to view port mappings and operational status across all current worktrees.

• Leverage wt logs for specific components like rag, runtime, or server to perform targeted debugging without cross-pollination of logs.

• Always invoke wt gc to perform garbage collection on orphaned data directories and prune stale worktrees to maintain host system health.

• Note that wt relies on local port allocation (typically starting at 14345+offset); ensure your environment permits dynamic port binding for the specified ranges.

• The skill is specifically optimized for LlamaFarm development, utilizing underlying Git worktree structures to map different branches to the same repository root.