Tutorial: Chaining Workflows

Commands are designed to feed one another: deterministic checks produce facts and an OKF bundle; the bundle can be browsed, shared and re-checked; LLM-assisted commands help you decide which rules to write.

graph LR
  Q[G-REASON / AUDIT: explore] --> R[Write a rule in archai_standards.yaml]
  R --> C[CHECK-STANDARDS: verify with evidence]
  C --> B[OKF bundle]
  B --> V[OKF-VIEW: investigate]
  B --> X[CHECK-STANDARDS --from-okf: re-check elsewhere]

Explore, then verify

ai-architect G-REASON . "Which modules couple the API to the database?"

The answer is LLM output. Encode the question as a rule and get a verified answer:

rules:
  - id: api-not-db
    kind: forbidden_dependency
    from: ["myapp.api"]
    to: ["myapp.db"]

ai-architect CHECK-STANDARDS .

Check, export, investigate

ai-architect CHECK-STANDARDS . --json report.json --okf-out ../bundle
ai-architect OKF-VIEW ../bundle findings --status violation
ai-architect OKF-VIEW ../bundle F-2b20e2df65e7

Script it

CHECK-STANDARDS exits 0 (no verified violations), 1 (violations) or 2 (configuration or usage error):

#!/usr/bin/env bash
set -u
ai-architect CHECK-STANDARDS . --json report.json --okf-out bundle
case $? in
  0) echo "Standards met." ;;
  1) echo "Verified violations:"; ai-architect OKF-VIEW bundle findings --status violation --no-color; exit 1 ;;
  *) echo "Check could not run; see the output above."; exit 2 ;;
esac

Re-check somewhere else

A bundle carries the graph and the rules, so a reviewer can re-run the check without the source, or apply their own rules:

ai-architect CHECK-STANDARDS --from-okf bundle
ai-architect CHECK-STANDARDS --from-okf bundle --config reviewer-rules.yaml