Gondolabeta
⌁ Guides

New API

Add Gondola as an upstream channel in your self-hosted New API gateway and resell any Venice model, paid per request in USDC on Base. Your users never see Gondola; they just get more models.

1

Create an API key

Sign in with your Base wallet at /keys and mint a Gondola API key. Keys start with gnd_. It authenticates every request and each call is charged in USDC against your Gondola balance.

Top up USDC on Base once at /wallet. For a key that lives inside a gateway, also consider setting the optional per-key spend cap and expiry when you mint it: they bound what the channel can ever spend, even if the key leaks.

2

Add the channel

In the New API console open Channels, then Add channel. Pick type OpenAI, not Custom: New API's own hint on the Custom type recommends the OpenAI type for relay upstreams, and the OpenAI type asks the upstream to report real token usage on streamed requests. Name the channel, set the base URL, and paste your key.

values to enter
Type:      OpenAI
Name:      Gondola
Base URL:  https://api.gondola-ai.com
Key:       gnd_xxxx
Base URL gotcha
Enter https://api.gondola-ai.com with no /v1. New API appends /v1/chat/completions and the other paths itself, so a base URL that already ends in /v1 doubles the segment and every request 404s.
3

Fetch models and go live

Use the channel form's fetch models action to fill the model list from Gondola's catalog. The endpoint is public, so it works before the first paid request:

shell
curl https://api.gondola-ai.com/v1/models

Then prune the list to the models you actually want to resell and save the channel. The same steps work in One API and most of its forks.

Prices are live, not static
Gondola is a marketplace: each request routes to the cheapest supplier with capacity, so per-model prices move. Check /models for current per-token prices and /marketplace for the live order book before hardcoding ratios in your own pricing.
4

Notes and troubleshooting

Streamed usage is real, not estimated
Gondola reports token usage in the final chunk of every streamed response, so New API's usage logs bill from upstream-reported counts rather than its own estimate.
Claude-format traffic
On an OpenAI-type channel, New API converts Anthropic-format requests to chat completions before forwarding. That works for most use. Gondola also serves the native Anthropic Messages API at /v1/messages for clients that point at it directly.
Endpoints Gondola does not serve
Legacy completions, embeddings, rerank, and audio are not served. Route only chat, image, and video models to this channel.

Watch your balance from the gateway side; successful responses also carry an x-gondola-balance header with the remaining spendable USD:

shell
curl https://api.gondola-ai.com/v1/balance -H "Authorization: Bearer gnd_xxxx"

A 402 with code insufficient_balance means top up at /wallet. A 402 with spend_limit_exceeded means the key hit its own spend cap: raise it under the API keys tab, topping up will not help.

Keep going
All setup guides →Claude Code guide →SillyTavern guide →Live prices →