nodecode-belive-aihub
BeLive AI Hub: an OpenAI-compatible model gateway as a Nodecode provider, its model picker filled from the hub's own catalog. A Nodecode rewrite of belivetech/hermes-belive-aihub by belivetech.
Get it
Tell the agent "add the belive-aihub cell", or run nodecode add nc://belive-aihub.
Set it up
Make a workspace inference key in BeLive AI Hub at https://aihub.belive.sg/.
Save it with /connect (pick BeLive AI Hub), or set
BELIVE_TOKENHUB_API_KEY in the environment; a saved key answers first. The
plugin's own advice holds: use a workspace inference key, not a console login
or management key.
Optional, and rarely needed — the section is on unless it says
"enabled": false:
{
"belive-aihub": {
// where the gateway is served; the chat lane appends /chat/completions
"base_url": "https://tokenhub-api.belive.sg/api/v1"
}
}
Use it
/models lists BeLive AI Hub's chat models, and a turn on one goes to
https://tokenhub-api.belive.sg/api/v1/chat/completions.
The hub is one endpoint in front of many upstream models, so the cell reads
the hub's own model list (<base_url>/models, with your key) and keeps only
the entries a turn can actually run on: text out, tool calls in, and a ready
chat_completions route that takes tools. The hub's exact author/slug ids
are preserved, and a model's context window is read off the same row.
When that catalog is unreachable, the picker falls back to the three models the original plugin curates, so an offline start still fills it. A successful empty catalog is an answer, not a failure: it empties the picker, as the original leaves it.
A base_url other than the hub's own — a proxy or a self-hosted relay — is
not the hub's catalog to interpret, so the curated list answers there. The
hub's staging base, https://api-stg-tokenhub.belive.sg/api/v1, is the one
exception: it is accepted, and lists staging's models.
Test
From this repository's root:
.github/test-cell.sh providers/nodecode-belive-aihub
MIT licensed — see LICENSE. Ported from the original plugin; see NOTICE.