Summary

This pull request fixes several typos and bugs found in the documentation and build scripts.

Changes

README.md

• Line 135: Fixed typo in Basic Usage section where the command vllm-cl was missing the trailing i. Changed to vllm-cli.

Makefile

• Line 1: The .PHONY declaration was incomplete. Added missing targets: test-cov, format-check, type-check, ci-local, ci-matrix, and pre-commit. Without proper .PHONY declarations, make may skip execution if a file with the same name exists in the directory.

scripts/test_ci_locally.sh

• Line 54: Fixed pytest-cov detection logic. The original code used command -v pytest-cov which checks for a shell command, but pytest-cov is a Python package/pytest plugin, not a standalone executable. Changed to use python -c "import pytest_cov" for proper detection. Also added a skip message when pytest-cov is not installed.
• Line 93-104: Fixed pyproject.toml validation to be compatible with both Python 3.11+ (which has built-in tomllib) and earlier Python versions (which require the toml package). The original code only used import toml which would fail on Python 3.11+ if the toml package is not installed.

docs/profiles.md

• Line 8: Fixed grammar error. Changed Uses vLLM's defaults configuration to Uses vLLM's default configuration (singular form of "default" as adjective).
• Line 185: Updated the VLLM_ATTENTION_BACKEND environment variable reference table. Changed TRITON to TRITON_ATTN_VLLM_V1 to match the actual value used in the gpt_oss_ampere profile (line 71).

Testing

All changes are documentation and configuration fixes that do not affect runtime behavior. The script fixes improve compatibility across different Python versions and environments.