Model gateway settings

.env-backedloadingrequiredlocal persisted values

Credential setup

Import an environment file or complete the required fields manually. Saved values are stored only in the local Workbench runtime directory.

  1. TEST_INTELLIGENCE_LLM_GATEWAY_API_KEY: Required field is empty
  2. TEST_INTELLIGENCE_FIGMA_ACCESS_TOKEN: Required field is empty
  3. TEST_INTELLIGENCE_REGION_ATTESTATION_SIGNING_KEY: Required field is empty
Workspace-local path read by the Workbench process.
import.env

LLM gateway

Required for run/figma-export routes. The Azure OpenAI resource the runtime calls for generation and judging.

Foundry account

Azure AI Foundry account endpoints used by the test-case and visual paths. Separate from the gateway endpoint above.

For Azure Foundry/OpenAI-compatible deployments, use the v1 base URL.
Use the visual model account's matching v1 base URL when it differs from the text endpoint.

Deployment pins

Pinned deployment names. Each pin is wired to a specific role in the pipeline; do not change without an attested rollout.

Region attestation

Sovereign-region pin and attestation source. The gateway will refuse the run if its inferred region disagrees with this value.

1 — attestation comes from a sovereign source. 0 — degraded; emit warn-tagged evidence.
Required to sign region-attestation evidence artifacts. Keep this operator-managed and tenant-local.

Runtime trust

Optional enterprise TLS trust configuration used for Figma REST and image export calls.

Optional workspace-local PEM bundle path for corporate TLS interception.

Policy

Global override. Mirrors --allow-policy-blocked. When ON, the runtime will still emit artifacts even if the policy gate rejects.

Cleared after every release-gate run unless an attested rationale is filed.
export downloads a local file only