Team Requirements Verify Workflow

This example lets a repository declare the local inference state it expects and fail fast when a teammate or CI runner has drifted.

python examples/team-requirements/verify-local-llm.py

The verifier is read-only. It runs:

inferctl config validate --json
inferctl status --json
inferctl preflight <task> --prompt-file <file> --json

It reports drift; it does not pull models, start daemons, load models, rewrite config, or repair backend state.

Requirements vs. Lockfiles

inferctl.requirements.toml is author intent: a hand-editable statement of what the project needs. A future lockfile would be a generated snapshot of a known local inference state. This example does not require inferctl lock or inferctl verify; those can graduate later if the requirements pattern proves useful.

CI Sketch

name: local-inference-preflight
on:
  workflow_dispatch:
jobs:
  verify:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - uses: actions/setup-go@v5
        with:
          go-version: stable
      - run: go install github.com/inferctl/inferctl/cmd/inferctl@latest
      - run: python examples/team-requirements/verify-local-llm.py

For hosted CI without real local model daemons, use deterministic fixture backends in the same pattern as examples/team-requirements/test.sh.