hermes-gondola-provider
Adds Gondola as a native model provider for Nous Research's Hermes Agent.
Gondola is a pay-as-you-go marketplace for Venice AI inference: each request is routed to the cheapest available supplier and settled in USDC on Base. One OpenAI-compatible endpoint, no subscription, no Venice account. Prompts and completions are never persisted.
What you get
- A
gondolaprovider inhermes model,hermes setup, and--provider gondola - Live model catalog from Gondola's public
GET /v1/models, filtered to chat-capable models (the raw catalog also lists image and video models) - A Gondola row in
hermes doctor's API Connectivity section - A curated fallback model list when the live fetch is unreachable
Get an API key
- Mint a key at https://gondola-ai.com/wallet?tab=keys (format:
gnd_...) - Top up USDC on Base once at https://gondola-ai.com/wallet
- Setup guides for other tools: https://gondola-ai.com/guides
Recommended install
Clone straight into the directory Hermes scans for model-provider plugins:
macOS / Linux / WSL:
mkdir -p "$HOME/.hermes/plugins/model-providers"
git clone https://github.com/Gondola-Market/hermes-gondola-provider \
"$HOME/.hermes/plugins/model-providers/gondola-provider"
Windows (PowerShell):
New-Item -ItemType Directory -Force "$env:LOCALAPPDATA\hermes\plugins\model-providers" | Out-Null
git clone https://github.com/Gondola-Market/hermes-gondola-provider `
"$env:LOCALAPPDATA\hermes\plugins\model-providers\gondola-provider"
(replace the base directory with $HERMES_HOME if you set that variable)
Alternative: hermes plugins install
hermes plugins install Gondola-Market/hermes-gondola-provider
This path is tracked by hermes plugins list/update/remove, but the installer
places model-provider plugins in a directory the provider registry does not
scan, so one manual move is required afterwards. The on-screen after-install
note (also in after-install.md) has the exact commands.
Configure
export GONDOLA_API_KEY=gnd_... # or let the installer prompt and save it
hermes model # pick Gondola, then a model
hermes -z "hello" --provider gondola -m deepseek-v4-flash
Or pin it in ~/.hermes/config.yaml:
model:
provider: gondola
default: deepseek-v4-flash
Verify
Run hermes doctor. With the plugin in place, Gondola appears under API
Connectivity; the probe is a GET /v1/models with your key as Bearer. Note
that this endpoint is public, so the row proves reachability, not key
validity: a wrong key still surfaces on the first paid request.
Models
Curated fallbacks: deepseek-v4-flash (default aux), claude-sonnet-5,
kimi-k2-7-code, qwen3-coder-480b-a35b-instruct-turbo,
venice-uncensored-1-2. The live catalog (100+ text models incl.
hermes-3-llama-3.1-405b, plus image and video via Gondola's other
endpoints) is at https://api.gondola-ai.com/v1/models with per-model
pricing, context windows, and capability flags.
Payments
The plugin authenticates with GONDOLA_API_KEY; each request is metered
against your prepaid USDC balance. Separately, Gondola's API accepts
accountless per-request x402 payments (HTTP 402 challenges, USDC on Base)
for clients that speak the protocol natively; Hermes is not one of them, so
the plugin always uses key auth.
Updating / removing
- Installed via git clone:
git pullin the plugin directory, or delete it. - Installed via
hermes plugins install:hermes plugins update gondola-provider/hermes plugins remove gondola-providermanage the flat install location only; if you moved the directory per after-install.md, manage it with git directly.
Compatibility notes
Hermes' plugin manifest has no minimum-version field, so this plugin tracks
ProviderProfile in providers/base.py on hermes-agent main. If a Hermes
update renames profile fields, reinstall the latest version of this plugin.
License
MIT