chore: .env.example follows the NEMOTRON_ naming and says a model id can reach end of life

The owner's .env names the NVIDIA variables after the model family; the
example now matches, adds NEMOTRON_MODEL, and says how a chosen id is
checked before the live suite runs, since the first id set had reached
end of life on 2026-08-26 (410 from the endpoint).

Signed-off-by: Ayla Croft <aylacroft@proton.me>

Author: HackTuah

.env.example

@@ -20,12 +20,17 @@
 OPENROUTER_API_KEY=
 OPENROUTER_BASE_URL=https://openrouter.ai/api/v1
 
-# NVIDIA NIM: https://build.nvidia.com/ (API key from the account page)
-NVIDIA_API_KEY=
-NVIDIA_BASE_URL=https://integrate.api.nvidia.com/v1
+# NVIDIA NIM: https://build.nvidia.com/ (API key from the account page). Named after the
+# model family it is used for. Model ids there reach end of life: the id you set is checked
+# against GET /models before the live suite runs, and a 410 names the replacement to find.
+NEMOTRON_API_KEY=
+NEMOTRON_BASE_URL=https://integrate.api.nvidia.com/v1
+NEMOTRON_MODEL=
 
 # Which provider and model the live suite uses. The model id must exist on the provider's
-# current list; check it there rather than here, since free-tier ids change.
+# current list and answer a tool call and a streamed completion; check both there rather
+# than here, since free-tier ids change. The pick is recorded in the slice's NOTES.md with
+# the probe output.
 TRINITY_LIVE_PROVIDER=openrouter
 TRINITY_LIVE_MODEL=