fix(models): delist hy3-free and laguna-s-2.1-free — OpenCode relay dropped them (anon 401)

The OpenCode Zen relay no longer serves hy3-free (since ~2026-08-31) or laguna-s-2.1-free (new, verified 2026-09-09): both are gone from the live GET /zen/v1/models catalog and anonymous chat completions return 401 {"type":"ModelError","message":"Model <id> is not supported"} (2 probes >=60s apart, x-opencode-session header present). - hermes_cli/models_catalog_static.py: remove both slugs from the opencode-free offline floor and the opencode-zen discovery floor; document the delist dates in the catalog comment. - plugins/model-providers/opencode-free: default_aux_model moves from the dead laguna-s-2.1-free to nemotron-3.5-lightning-free (fastest surviving anonymous model). - tests: swap fixtures off the dead slugs; extend the floor-exclusion invariant to cover both. The live revalidation path already hides them when the relay is reachable; this fixes the OFFLINE floor and the aux default, which would otherwise offer/route to models that 401.
2026-09-09 02:44:40 -07:00
parent 230ca004a7
commit cc81e436ce
6 changed files with 18 additions and 14 deletions
--- a/hermes_cli/models_catalog_static.py
+++ b/hermes_cli/models_catalog_static.py
@@ -229,15 +229,17 @@ _PROVIDER_MODELS: dict[str, list[str]] = {
        "grok-4.6", "grok-4.5", "grok-build-0.1", "muse-spark-1.2", "minimax-m3", "minimax-m2.7", "minimax-m2.5",
        "glm-5.3", "glm-5.3-flash", "glm-5.2", "glm-5.1", "glm-5", "kimi-k2.7-code", "deepseek-v4-pro",
        "deepseek-v4-flash", "deepseek-v4-flash-free", "qwen3.6-plus", "qwen3.5-plus", "big-pickle", "mimo-v2.5-free",
-        "hy3-free", "laguna-s-2.1-free", "nemotron-3-ultra-free", "nemotron-3.5-lightning-free",
+        "nemotron-3-ultra-free", "nemotron-3.5-lightning-free",
        "muse-spark-1.2-contributor-free", "muse-spark-1.3-contributor-free",
    ],
    # OpenCode keyless free tier — OFFLINE FLOOR only. provider_model_ids("opencode-free")
    # revalidates live against GET /zen/v1/models and filters to the anonymous tier, so this list
    # may lag the relay (intentional). Known-delisted models are REMOVED (the offline fallback must
-    # not offer a model that 401s, e.g. x-preview-f-free).
+    # not offer a model that 401s; x-preview-f-free delisted 2026-08-26, hy3-free and
+    # laguna-s-2.1-free delisted 2026-09-09 — both dropped from live /zen/v1/models and 401
+    # "Model … is not supported" anonymously).
    "opencode-free": [
-        "deepseek-v4-flash-free", "hy3-free", "mimo-v2.5-free", "laguna-s-2.1-free",
+        "deepseek-v4-flash-free", "mimo-v2.5-free",
        "nemotron-3-ultra-free", "nemotron-3.5-lightning-free", "muse-spark-1.2-contributor-free",
        "muse-spark-1.3-contributor-free",
    ],
--- a/plugins/model-providers/opencode-free/init.py
+++ b/plugins/model-providers/opencode-free/init.py
@@ -40,9 +40,10 @@ opencode_free = OpenCodeFreeProfile(
        "X-Title": "Hermes Agent",
        "User-Agent": f"HermesAgent/{_HERMES_VERSION}",
    },
-    # laguna is the fastest non-UA-gated free model; big-pickle 429s every
+    # laguna-s-2.1-free was delisted by the relay 2026-09-09 (anon 401); the fastest
+    # surviving free model is the lightning-tier Nemotron. big-pickle 429s every
    # client except the opencode CLI's own User-Agent.
-    default_aux_model="laguna-s-2.1-free",
+    default_aux_model="nemotron-3.5-lightning-free",
 )

 register_provider(opencode_free)
--- a/tests/agent/test_opencode_session_affinity.py
+++ b/tests/agent/test_opencode_session_affinity.py
@@ -35,7 +35,7 @@ def _agent(provider, model, base_url, api_mode=None):
        ("opencode-go", "glm-5", "https://opencode.ai/zen/go/v1", None),  # chat_completions
        ("opencode-go", "gpt-5.6-luna", "https://opencode.ai/zen/go/v1", None),  # codex_responses
        ("opencode-go", "minimax-m2.7", "https://opencode.ai/zen/go/v1", "anthropic_messages"),
-        ("opencode-free", "laguna-s-2.1-free", "https://opencode.ai/zen/v1", None),
+        ("opencode-free", "nemotron-3.5-lightning-free", "https://opencode.ai/zen/v1", None),
        ("custom", "glm-5", "https://opencode.ai/zen/go/v1", None),  # URL-only detection
    ],
 )
--- a/tests/hermes_cli/test_model_validation.py
+++ b/tests/hermes_cli/test_model_validation.py
@@ -285,7 +285,7 @@ class TestCopilotNormalization:
        assert opencode_model_api_mode("opencode-zen", "x-preview-f-free") == "chat_completions"
        assert opencode_model_api_mode("opencode-zen", "opencode-zen/x-preview-f-free") == "chat_completions"
        # Other free-tier Zen models are chat/completions too.
-        assert opencode_model_api_mode("opencode-zen", "hy3-free") == "chat_completions"
+        assert opencode_model_api_mode("opencode-zen", "nemotron-3.5-lightning-free") == "chat_completions"
        assert opencode_model_api_mode("opencode-zen", "nemotron-3.5-lightning-free") == "chat_completions"
        # Hy3 on Go is chat/completions (Go endpoint table).
        assert opencode_model_api_mode("opencode-go", "hy3") == "chat_completions"
--- a/tests/hermes_cli/test_opencode_free_live_catalog.py
+++ b/tests/hermes_cli/test_opencode_free_live_catalog.py
@@ -32,12 +32,11 @@ from hermes_cli.models import (
 _STATIC_FLOOR = list(_PROVIDER_MODELS["opencode-free"])

 # The live relay's current free tier. x-preview-f-free was DELISTED 2026-08-26;
-# deepseek-v4-flash-free + mimo-v2.5-free are back on the live list.
+# hy3-free and laguna-s-2.1-free were DELISTED 2026-09-09 (gone from live
+# /models, anon 401 "Model … is not supported").
 _LIVE_FREE_MODELS = [
    "deepseek-v4-flash-free",
-    "hy3-free",
    "mimo-v2.5-free",
-    "laguna-s-2.1-free",
    "nemotron-3-ultra-free",
    "nemotron-3.5-lightning-free",
    "muse-spark-1.2-contributor-free",
@@ -220,3 +219,5 @@ class TestOpencodeFreeFollowUps:
    def test_static_floor_excludes_delisted_model(self):
        """The offline floor must not offer a model known to 401 (#95914)."""
        assert "x-preview-f-free" not in _PROVIDER_MODELS["opencode-free"]
+        assert "hy3-free" not in _PROVIDER_MODELS["opencode-free"]
+        assert "laguna-s-2.1-free" not in _PROVIDER_MODELS["opencode-free"]
--- a/tests/hermes_cli/test_opencode_zen_free_keyless.py
+++ b/tests/hermes_cli/test_opencode_zen_free_keyless.py
@@ -32,7 +32,7 @@ from hermes_cli.models import (

 class TestFreeRuntime:
    def test_zen_provider_free_model(self):
-        rt = opencode_zen_free_runtime("opencode-zen", "hy3-free")
+        rt = opencode_zen_free_runtime("opencode-zen", "nemotron-3.5-lightning-free")
        assert rt is not None
        assert rt["base_url"] == "https://opencode.ai/zen/v1"
        assert rt["api_key"] == OPENCODE_ZEN_FREE_KEYLESS_PLACEHOLDER
@@ -42,7 +42,7 @@ class TestFreeRuntime:
    def test_go_provider_heals_to_zen(self):
        # Free slugs only exist on the Zen relay; a Go selection must be
        # routed to Zen (the Go relay rejects the model outright).
-        rt = opencode_zen_free_runtime("opencode-go", "hy3-free")
+        rt = opencode_zen_free_runtime("opencode-go", "nemotron-3.5-lightning-free")
        assert rt is not None
        assert rt["base_url"] == "https://opencode.ai/zen/v1"

@@ -82,13 +82,13 @@ class TestRuntimeProviderKeylessRouting:
            return resolve_runtime_provider(requested=provider, target_model=model)

    def test_zen_free_model_resolves_keyless(self):
-        rt = self._resolve("opencode-zen", "hy3-free")
+        rt = self._resolve("opencode-zen", "nemotron-3.5-lightning-free")
        assert rt["api_key"] == OPENCODE_ZEN_FREE_KEYLESS_PLACEHOLDER
        assert rt["base_url"] == "https://opencode.ai/zen/v1"
        assert rt["api_mode"] == "chat_completions"

    def test_go_free_model_resolves_keyless_on_zen(self):
-        rt = self._resolve("opencode-go", "hy3-free")
+        rt = self._resolve("opencode-go", "nemotron-3.5-lightning-free")
        assert rt["api_key"] == OPENCODE_ZEN_FREE_KEYLESS_PLACEHOLDER
        assert rt["base_url"] == "https://opencode.ai/zen/v1"