feat: restructure the customer-facing model catalog

.env.example

@@ -144,8 +144,25 @@ GROQ_API_KEY=                  # Groq free (chat fallback + reasoning)
 # arbitrary free-model router -- that billed customers the priced model's cost
 # while an undisclosed free model actually answered. deploy-demo-box.yml's
 # "Assert model catalog prices agree..." step fails the deploy if these ever
-# drift from the catalog again. Failures still cascade via LiteLLM fallbacks
-# to the Groq routes below.
+# drift from the catalog again.
+#
+# SUPERSEDED IN PART, 2026-08-22 catalog restructure. hive-default and hive-auto
+# both moved off OpenRouter onto Groq gpt-oss models, and their routes were
+# retired in favour of route-groq-default and route-groq-auto, whose upstream
+# model is written literally in deploy/litellm/config.yaml and is owned by
+# provider_routes.provider_model through the config sync (issue #713). No new
+# env var was added for them on purpose: another os.environ indirection would
+# rebuild exactly the drift that #689 and #965 were.
+#
+#   * OPENROUTER_DEFAULT_MODEL is now UNUSED. Nothing reads it. Left in place
+#     rather than deleted only because several CI workflows still pass it
+#     through; treat it as dead and do not wire anything new to it.
+#   * OPENROUTER_AUTO_MODEL is still LIVE, but for route-doc-vlm alone, which
+#     is the only vision-capable route left after hive-auto moved to the
+#     text-only gpt-oss-120b. It must stay a multimodal slug.
+#
+# The LiteLLM chat fallbacks that used to cascade these to the Groq routes are
+# also gone; a fallback answers from a model the alias was not priced against.
 OPENROUTER_DEFAULT_MODEL=openrouter/openai/gpt-4o-mini
 OPENROUTER_AUTO_MODEL=openrouter/openai/gpt-4.1-mini
 # route-openrouter-fast-fallback is disabled (health_state='disabled') and is