feat: add TypoMixin for 'Did you mean?' suggestions on invalid commands

[anandhu-eng]

Summary

• Adds mlc/typo_mixin.py — a TypoMixin class that overrides ArgumentParser.error() to print a "Did you mean …?" hint when the user mistypes an action or target
• Wires it into mlc/main.py via TypoArgumentParser(TypoMixin, argparse.ArgumentParser); subparsers inherit the class automatically so both action-level and target-level typos are caught
• Candidates that start with the input (3+ chars) are suggested first, so abbreviations work: exp → experiment, dock → docker
• Target typos are checked against the chosen action's own targets (the pre-parser no longer validates target), so hints never suggest a target the action rejects
• No new runtime dependencies — uses difflib.get_close_matches from the Python stdlib

Example output

$ mlc rune script
Did you mean one of: 'prune', 'run'?
mlc: error: argument command: invalid choice: 'rune' …

$ mlc run scrip
Did you mean 'script'?
mlc run: error: argument target: invalid choice: 'scrip' …

$ mlc find cach
Did you mean 'cache'?
mlc find: error: argument target: invalid choice: 'cach' …

$ mlc dock run
Did you mean one of: 'docker', 'docker-run', 'doc'?
mlc: error: argument command: invalid choice: 'dock' …

Changes

File	Change
mlc/typo_mixin.py	New — TypoMixin with suggest() and error() override
mlc/main.py	Use TypoArgumentParser in build_pre_parser() and build_parser(); drop the pre-parser's shared target choices; add details/extra to the load subparser
tests/test_typo_mixin.py	New — 37 tests (unit + CLI subprocess)
docs/typo_detection.md	New — full documentation with examples and tuning guide
README.md	Mention typo detection in Key Features
.github/workflows/test-unit.yml	Update step description to reflect expanded test scope

Test plan

• python -m unittest discover -s tests -p "test_*.py" -v — all tests pass (130 total, 37 in test_typo_mixin.py)
• Existing tests (meta validation, cache mark-tmp) — no regressions
• Manually verified action typos: rune, pul, fidn, serach, lst
• Manually verified target typos: scrip, cach, rpo
• Verified prefix matches: exp → experiment, dock → docker first, remote → remote-*
• Verified target hints use per-action choices: mlc docker cach (no invalid cache hint), mlc load cfgg → cfg
• Scored cutoffs 0.5 / 0.55 / 0.6 on 100 realistic typos + 23 junk inputs; kept 0.6 (0.5 nearly doubled unwanted suggestions for +1 hit)
• Verified garbage input (e.g. xyzzy123) shows no hint — standard error only
• Exit code remains 2; usage block and error message unchanged

🤖 Generated with Claude Code

[github-actions]

MLCommons CLA bot All contributors have signed the MLCommons CLA ✍️ ✅

[github-actions]

🤖 AI PR Review Summary\n\nAdds a typo detection feature to the CLI by extending argparse.ArgumentParser with a mixin that suggests close matches for mistyped commands or targets. This improves user experience by providing helpful hints on invalid inputs. The change includes a new TypoMixin class, integration into the main parser, documentation, and tests. Risks are minimal as the mixin only overrides the error method and uses standard library difflib. Design is clean and modular, leveraging Python's MRO and argparse's parser_class inheritance to cover subcommands automatically.

[arjunsuresh]

LGTM