GonkaGate OpenHuman Setup

openhuman-setup configures OpenHuman to use GonkaGate as an OpenAI-compatible provider.

npx @gonkagate/openhuman-setup

Current status: the CLI writes OpenHuman's native cloud_providers provider shape, preserves unrelated TOML config, creates a timestamped backup before replacing an existing config file, and verifies the written config.

Current OpenHuman credential storage is not a safe standalone JSON write target: OpenHuman may store secrets in the OS keychain or encrypted JSON depending on runtime policy. This CLI therefore does not mutate auth-profiles.json; it prints the OpenHuman Settings -> AI step for the GonkaGate key and redacts any gp-... value from output.

Target Contract

• Package: @gonkagate/openhuman-setup
• Provider id: gonkagate
• Base URL: https://api.gonkagate.com/v1
• Transport: /v1/chat/completions
• OpenHuman provider shape: cloud_providers entry plus workload provider strings
• OpenHuman credential key: provider:gonkagate, profile default
• Recommended model: moonshotai/kimi-k2.6
• Additional validated model: qwen/qwen3-235b-a22b-instruct-2507-fp8

Safe Secret Inputs

• hidden interactive prompt
• GONKAGATE_API_KEY
• --api-key-stdin

The utility must not accept a plain --api-key flag and must never print the gp-... key.

Development

npm install
npm run ci

The CLI can be exercised against a temporary OpenHuman workspace with:

npm run build
OPENHUMAN_WORKSPACE="$(mktemp -d)" node bin/gonkagate-openhuman.js --json --yes

Implementation Notes

Current OpenHuman builds expose a native custom-provider shape: cloud_providers entries plus workload provider strings like gonkagate:moonshotai/kimi-k2.6. The implementation should write that shape, not the older api_url + api_key + model_routes direct-inference path.

Provider keys belong in OpenHuman's provider credential namespace (provider:gonkagate, profile default). Because direct standalone writes cannot safely reproduce OpenHuman's keychain/encrypted-JSON policy, the current CLI reports the manual Settings -> AI key step instead of persisting secrets. OpenHuman still requires an active app session before custom providers can run, so setup verification reports that gate separately from GonkaGate key validity.

Release Proof

Automated release proof is npm run ci plus a temp-workspace CLI smoke. Before publishing, also run one real-workspace smoke with an active OpenHuman session and a real GonkaGate key pasted through OpenHuman Settings -> AI. Current proof notes live in docs/specs/openhuman-setup-prd/release-proof.md.

The product requirements for the implementation live in docs/specs/openhuman-setup-prd/spec.md. The current execution plan lives in docs/specs/openhuman-setup-prd/implementation-plan.md.