Skip to content

Gateway allowlist request #1322

Closed @Puyre opened 2026-06-08 15:13 UTC 1 comment Updated 2026-06-10 15:20 UTC

Request to be added as a Gonka gateway operator (devshard creator allowlist)

Hi Gonka core team & community,

We're requesting inclusion of our on-chain address in devshard_escrow_params.allowed_creator_addresses so we can operate our own devshard gateway rather than routing through a third-party broker.

On-chain identity

  • Devshard creator address (requested for allowlisting): gonka15y6xg0ps5w0u4w3ttx557mf46v82ms8svcavy2
  • This account is funded above devshard_escrow_params.min_amount and has its on-chain pubkey published (secp256k1).

About the project

We're building an OpenAI-compatible inference gateway that uses Gonka as its primary backend — a drop-in replacement for closed AI APIs (developers change one base_url and keep their existing OpenAI SDK code).

Under the hood, the gateway routes requests to Gonka by default and automatically falls back to a stable provider (OpenRouter) when the network is unavailable. The goal is to let end users run their workloads on Gonka — capturing the cost savings — without being exposed to the network's current instability, since the fallback transparently covers the downtime.

Why this helps the network

  • Opens new demand that Gonka can't currently capture. A transparent fallback to stable providers makes the network usable for the large set of users who today can't rely on Gonka directly because of its uptime variance. They get Gonka-level pricing for the majority of traffic that does land on the network, instead of avoiding it entirely.
  • Reliability engineering published back. We run automatic failover and capacity-aware routing in front of the network; we're willing to publish anonymized reliability and latency data (TTFT, tokens/sec, tail latency, error/refund rate), which the ecosystem currently lacks sustained public numbers for.

Supported models (intent)

We intend to serve all three currently governance-approved models:

  • qwen/qwen3-235b-a22b-instruct-2507-fp8
  • moonshotai/kimi-k2.6
  • minimaxai/minimax-m2.7

Governance and ops commitments

  • We will respond to maintainer comments on this issue within one business day.
  • We will adjust scope, staging, or framing in this issue rather than open a new one if maintainers prefer.

Team and contact

  • Contact: andres@rogilabs.com
  • TG: @puyre

Thanks for reviewing.

Rogi AI

💬 Comments (1)

@Puyre commented 2026-06-10 15:20 UTC

Closing this — after checking with the Gonka community, I realized we don't need to run our own devshard. Getting a broker key to send requests through node4 will be enough for our case. I'll likely open a separate issue requesting a broker key (or asking for documentation on how one can be obtained).

🔄 Auto-synced from Issue #1322 every hour.