I've published an open-source Hermes provider plugin for your API — it's live at
https://github.com/thepurpose-npub/ppq-hermes-provider and I've opened a PR to include it in the official Hermes plugin catalog: https://github.com/NousResearch/hermes-agent/pull/142626
It's an extended version of the starter plugin in your repo (3-layer model picker: user-local pins → top chat models from stats.ppq.ai → static fallback, plus ZDR opt-in routing). The `fetch_models` fix from my PR #5 on your repo is included — I closed that PR since I'd moved the work to my own repo instead.
One request: the plugin reads your "Top Chat Models (Last 7 Days)" by replaying the Plotly Dash update protocol (`POST /_dash-update-component` with `output: top-models-list.children`). It works, but it's semi-fragile — any dashboard restructure (component id rename, layout change) silently breaks the parser.
Would you consider exposing a tiny machine-readable stats endpoint, e.g.: `GET https://stats.ppq.ai/api/top-models?days=7&tier=chat&limit=10`
returning something like:
````
{"generated_at": "...", "models": [
{"id": "google/gemini-2.5-flash-lite", "delta_pct": -12},
{"id": "deepseek/deepseek-v4.1-flash", "delta_pct": -27}
]}
````
Even a minimal JSON dump behind the same dashboard would do. It would let integrators (like this plugin) present your live landscape without piggybacking on the UI layer. Happy to contribute a PR to the dashboard code if that's easier on your side — or a PR to your repo if you'd like the fetch_models fix upstreamed.
Thanks for the great relay — ZDR + TEE tiers are exactly what privacy-conscious agentic users have been asking for. And credit where it's due: the 5% topup bonus you give Bitcoin Lightning and NWC users is exactly the right incentive — paying per query with sats and getting a little extra for it is how pay-per-use AI should work.
For the record, this whole effort got rolling after HERMETICVM (nprofile…u08z) called on you to open your harness integration tutorials up to community contributions:
PPQdotAI (npub16g4…kv4h) you seem to be behind w/r/t tutorials for using ppq.ai with various new harnesses like pi, omp, etc.
How about moving those to markdown, put them in a repo and accept submissions?
I'm currently working with this script to pull the latest models from your API, put them in into a compatible config file to use it without losing my mind.
I'm considering to make this into a ppq-to-$harness script as well.
h/t HERMETICVM (nprofile…u08z)