nodecode-aihubmix
AIHubMix: 400+ OpenAI-compatible models as a Nodecode provider, with the model picker filtered to the routes that can call tools. A Nodecode rewrite of AIhubmix/hermes-provider-aihubmix by AIhubmix.
Get it
Tell the agent "add the aihubmix cell", or run nodecode add nc://aihubmix.
Set it up
Make a key at https://aihubmix.com/token. Save it with /connect (pick
AIHubMix), or set AIHUBMIX_API_KEY in the environment; a saved key answers
first.
Optional, and rarely needed — the section is on unless it says
"enabled": false:
{
"aihubmix": {
// where the OpenAI-compatible API is served; a relay's address goes here
"base_url": "https://aihubmix.com/v1"
}
}
Use it
/models lists AIHubMix's models, and a turn on one goes to
https://aihubmix.com/v1/chat/completions.
AIHubMix fronts many upstream backends and publishes no capabilities on its
plain model list, so the cell reads the capability-annotated catalog at
https://aihubmix.com/api/v1/models?types=llm and keeps only the entries
advertising tools or function_calling — either marker alone qualifies,
because requiring both would hide about a third of the agentic catalog. A
model that cannot call tools is dead weight in an agent's picker.
When that catalog is unreachable, the picker falls back to a curated list of
tool-capable flagships, so an offline start still fills it. A base_url
other than AIHubMix's own means a proxy or a self-hosted relay, whose catalog
is not AIHubMix's to interpret, so the curated list answers there too.
Test
From this repository's root:
.github/test-cell.sh providers/nodecode-aihubmix
MIT licensed — see LICENSE. Ported from the original plugin; see NOTICE.
