Drive end-to-end KonfAI deep-learning experiments (medical imaging — segmentation, synthesis, registration) through the konfai-mcp MCP server: inspect a dataset, author or adapt KonfAI YAML configs, validate them, launch and monitor train / prediction / evaluation jobs, then compare runs and iterate. Use when the user wants to train, predict, or evaluate a KonfAI model, onboard a dataset for KonfAI, author or debug a KonfAI config (invalid root, missing checkpoint, dataset mismatch), read live training metrics, or build a leaderboard — and whenever the konfai-mcp / mcp__konfai__* tools are in play. Triggers: "train a KonfAI model", "run a segmentation/synthesis experiment", "inspect this dataset for KonfAI", "validate my config", "evaluate the run", "leaderboard", "why did the job fail".
Review and release changes to the KonfAI framework itself (core konfai, konfai-apps, konfai-mcp) without breaking its load-bearing invariants. Use when modifying or reviewing framework code — patching/reconstruction, image geometry, streaming, the config-by-reflection engine, the model/criterion runtime, checkpoints/RESUME, DDP, storage backends — or when preparing a release, judging a performance change, or checking public-API/ecosystem compatibility (Slicer, HF bundles). Triggers: "review this KonfAI change", "is this safe to merge", "will this break geometry/patching/streaming", "prepare a KonfAI release", "did this regress performance", "check the wheel ships the model catalog", "does this break the Slicer/apps contract". This is the *maintainer* companion to konfai-cli (running workflows) and konfai-experiments (MCP-driven experiments).
Run KonfAI deep-learning workflows for medical imaging (segmentation, synthesis, registration) from the command line: author or adapt a YAML config, then train, resume, predict, and evaluate with the `konfai` CLI, or run a packaged model with the `konfai-apps` CLI. Use when the user wants to train / fine-tune / run inference / evaluate a KonfAI model from the terminal, adapt an example (Segmentation / Synthesis) config, understand the workspace outputs (Checkpoints / Predictions / Evaluations), reference a custom model or loss by classpath, or run and serve a published app (impact-synth, impact-seg, konfai-apps, konfai-apps-server). Triggers: "train a KonfAI model", "konfai TRAIN / PREDICTION / EVALUATION", "run konfai on the CLI", "konfai-apps infer", "run the segmentation example", "evaluate my predictions", "fine-tune this app".