Docs say "TEE models" but the model ids are `e2ee-*` — document the prefix and capability-based discovery

[beautifulrem]

What happened

While building a TEE-based governance agent for the MetaMask Smart Accounts Kit × 1Shot API × Venice AI Dev Cook-Off, we went looking for the "TEE models" the docs and product pages describe. We expected ids like tee-* — but nothing in GET /models matches that. The actual TEE-attestation models are named with an e2ee- prefix (e.g. e2ee-gpt-oss-120b-p), and the only reliable way we found to discover them is to filter /models by the capability flag:

model_spec.capabilities.supportsTeeAttestation === true

That mapping ("TEE" in prose ⇄ e2ee-* ids + supportsTeeAttestation flag) doesn't appear anywhere we could find in the docs, so we lost a stretch of time hardcoding guessed ids before reading raw /models output.

Suggestion

1. Wherever TEE inference is described, document that the model ids carry the e2ee- prefix (and briefly why, since "E2EE" vs "TEE" naming is itself confusing).
2. Document capability-based discovery as the recommended pattern — a 5-line snippet filtering /models on supportsTeeAttestation would let clients auto-resolve a TEE model instead of pinning ids that may rotate.
3. Bonus: the /tee/attestation endpoint docs could cross-link to that discovery snippet, since attestation requests need exactly these model ids.

Our workaround (auto-resolve by capability, fall back from a configured id) is here if useful: https://github.com/beautifulrem/regent (packages/shared/src/venice.ts, resolveTeeModel; live demo https://mandate-app-murex.vercel.app).