Usage¶

mlenvdoctor exposes a CLI for daily use and a small Python API for tools that need structured diagnostic objects.

CLI Reference¶

Command	Use when	Example
`doctor`	You want the highest-signal summary	`mlenvdoctor doctor --guided`
`diagnose`	You need the complete table or exports	`mlenvdoctor diagnose --full --json issues.json`
`report`	You want JSON and HTML reports together	`mlenvdoctor report --output-dir mlenvdoctor-report`
`fix`	You want a safe plan or generated requirement files	`mlenvdoctor fix --plan`
`dockerize`	You need a Dockerfile for a model stack	`mlenvdoctor dockerize tinyllama --service`
`stack`	You want recommended dependency pins	`mlenvdoctor stack llm-training`
`mcp serve`	You want assistant/tool integration	`mlenvdoctor mcp serve`

Diagnose¶

Windows PowerShellmacOS / LinuxCIAutomation JSON

mlenvdoctor diagnose --full --json issues.json --html issues.html

mlenvdoctor diagnose --full --json issues.json --html issues.html

mlenvdoctor doctor --ci

mlenvdoctor doctor --json

Guided Recovery¶

mlenvdoctor doctor --guided

Use this for beginners, workshop environments, bootcamp labs, and teammate handoffs. It compresses many low-level checks into the top root-cause finding.

Python API¶

from mlenvdoctor.api import diagnose

issues = diagnose(full=True)

for issue in issues:
    if issue.status.startswith(("FAIL", "WARN")):
        print(issue.check_id, issue.severity, issue.fix)

The API returns DiagnosticIssue objects with stable fields:

Field	Meaning
`name`	Human-readable finding name
`status`	`PASS`, `WARN`, `FAIL`, or `INFO` with detail
`severity`	`critical`, `warning`, or `info`
`check_id`	Stable machine-readable check identifier
`category`	Runtime area such as `pytorch`, `gpu`, `docker`, or `system`
`fix`	Recommended action
`evidence`	Short evidence lines collected during the check

Use JSON for external automation

Use mlenvdoctor doctor --json for compact triage. Use mlenvdoctor diagnose --json - when another tool needs all raw checks.