parity-gate:

name: Run the GPT-5.4 / Opus 4.6 parity gate against the qa-lab mock

if: ${{ github.event.pull_request.draft != true }}

runs-on: blacksmith-8vcpu-ubuntu-2404

timeout-minutes: 20

runs-on: blacksmith-32vcpu-ubuntu-2404

timeout-minutes: 30

env:

# Fence the gate off from any real provider credentials. The qa-lab

# mock server + auth staging (PR N) should be enough to produce a

# meaningful verdict without touching a real API. If any of these

# leak into the job env, fail hard instead of silently running

# against a live provider and burning real budget.

#

# The parity pack has 11 isolated scenario workers. The Blacksmith

# runner label can still expose a small 2-vCPU VM, and concurrent

# isolated gateway workers make the short strict-agentic scenarios

# flaky, especially the approval-turn followthrough gate that expects

# a fast post-approval read within a 30s agent.wait timeout.

# The parity pack has 11 isolated scenario workers. It exercises a real

# gateway child plus mock model turns and subagents, so keep it serial in

# CI even on the larger runner. Concurrent isolated gateway workers make

# the short strict-agentic scenarios flaky, especially the approval-turn

# followthrough gate that expects a fast post-approval read within a 30s

# agent.wait timeout.

QA_PARITY_CONCURRENCY: "1"

OPENAI_API_KEY: ""

ANTHROPIC_API_KEY: ""