feat: add bench-ai CLI for natural language bench command help

[younis-ali]

Title

feat: add bench-ai CLI for natural-language bench command help

---

Description

Summary

Adds an optional bench-ai console entry point that turns plain-English questions into structured guidance and a suggested bench commands. It uses the existing requests dependency to call an OpenAI-compatible Chat Completions HTTP API. No new runtime dependencies.

Context is built from this machine’s bench Click CLI (full native subcommand paths + bench --help text) and, when run inside a Frappe bench, the same framework command appendix as bench --help (via get_frappe_help / get_env_frappe_commands). When not in a bench, a short static framework crib plus validation fallback still allows common commands (e.g. doctor) to be suggested conservatively.

Suggested lines are checked against the local CLI where possible; --run refuses invalid or placeholder commands. Default behavior is print-only.

What changed

• New module: bench/commands/ai.py the CLI, context builder, API client, parsing, validation.
• New console script in pyproject.toml: bench-ai = "bench.commands.ai:main".
• Tests: bench/tests/test_ai_command.py (mocked HTTP, parsing, validation, Click path collection).
• Docs: README section “bench-ai (optional)” under Basic Usage.

Configuration

Env varaibles

1. OPENAI_API_KEY or BENCH_AI_API_KEY: OPEN API key
2. BENCH_AI_BASE_URL: Default https://api.openai.com/v1
3. BENCH_AI_MODEL: Default gpt-4o-mini

How to test

pip install -e .
python -m unittest bench.tests.test_ai_command -v
export OPENAI_API_KEY=...   # optional for live test
bench-ai "back up all sites" --dry-run
bench-ai "what does bench doctor do?"

Dependencies

None added, uses existing requests (~2.32.x per pyproject.toml).

Documentation

README updated with install/setup, flags, and security note (review before --run).

Notes for reviewers

• Guidelines mention keeping PRs very small; this feature is intentionally one cohesive tool. I’m happy to split (e.g. core CLI + README in one PR, follow-up for crib/validation refinements) if you prefer cascading PRs.
• Facing strings are English CLI output; I did not wrap them in _() since bench CLI generally doesn’t i18n user messages the same way as Frappe apps—happy to adjust if project policy requires it.

Checklist

• Tests added / updated
• README updated
• CI considerations: unittest module (no pytest required)

[sonarqubecloud]

Quality Gate passed

Issues
0 New issues
0 Accepted issues

Measures
0 Security Hotspots
0.0% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud