feat(providers): add Opper as an OpenAI-compatible gateway

core/config.py

@@ -317,6 +317,7 @@ class ProvidersConfig(_Base):
     openrouter: ProviderConfig = Field(default_factory=ProviderConfig)
     forge: ProviderConfig = Field(default_factory=ProviderConfig)
     requesty: ProviderConfig = Field(default_factory=ProviderConfig)
+    opper: ProviderConfig = Field(default_factory=ProviderConfig)
     bedrock: ProviderConfig = Field(default_factory=ProviderConfig)
     anthropic: ProviderConfig = Field(default_factory=ProviderConfig)
     openai: ProviderConfig = Field(default_factory=ProviderConfig)

core/providers/registry.py

@@ -219,6 +219,23 @@ def _matches_declared_endpoint(self, api_base: str | None) -> bool:
         default_api_base="https://router.requesty.ai/v1",
         supports_prompt_caching=True,
     ),
+    ProviderSpec(
+        name="opper",
+        keywords=("opper",),
+        env_key="OPPER_API_KEY",
+        display_name="Opper",
+        backend="openai_compat",
+        is_gateway=True,
+        endpoint_class="gateway",
+        detect_by_base_keyword="opper.ai",
+        default_api_base="https://api.opper.ai/v3/compat",
+        # Opper serves Anthropic models and prices cached input separately, so
+        # the Anthropic cache-control markers are worth sending.
+        supports_prompt_caching=True,
+        # Ids are bare pool names such as ``claude-sonnet-4-6``, but a
+        # ``provider/model`` id like ``azure/gpt-5.5`` deliberately pins one
+        # route, so the prefix must survive rather than be stripped.
+    ),
     ProviderSpec(
         name="forge",
         keywords=("forge",),

tests/test_opper_provider.py

@@ -0,0 +1,78 @@
+"""Pin the Opper provider as a mirror of the Requesty provider.
+
+Opper is an EU-hosted OpenAI-compatible gateway wired on the same generic
+``openai_compat`` path as OpenRouter and Requesty. These tests assert the
+shared gateway wiring, pin the Opper-specific base URL and env var, and lock
+the one place Opper differs from the other gateways: its ids are bare pool
+names, but a ``provider/model`` id pins a single route, so the prefix must not
+be stripped the way Forge's is.
+"""
+
+from __future__ import annotations
+
+import sys
+from pathlib import Path
+
+ROOT = Path(__file__).resolve().parents[1]
+if str(ROOT) not in sys.path:
+    sys.path.insert(0, str(ROOT))
+BACKEND = ROOT / "new_ui" / "backend"
+if str(BACKEND) not in sys.path:
+    sys.path.insert(0, str(BACKEND))
+
+from core.providers.registry import find_by_name  # noqa: E402
+
+OPPER = find_by_name("opper")
+REQUESTY = find_by_name("requesty")
+FORGE = find_by_name("forge")
+
+
+# ---- registry --------------------------------------------------------------
+
+
+def test_opper_is_registered() -> None:
+    assert OPPER is not None
+    assert OPPER.name == "opper"
+    assert OPPER.display_name == "Opper"
+
+
+def test_opper_mirrors_the_generic_gateway_wiring() -> None:
+    assert OPPER is not None and REQUESTY is not None
+    assert OPPER.backend == REQUESTY.backend == "openai_compat"
+    assert OPPER.is_gateway is True
+    assert OPPER.endpoint_class == "gateway"
+    assert OPPER.is_local is False
+    assert OPPER.is_oauth is False
+
+
+def test_opper_provider_specific_endpoint() -> None:
+    assert OPPER is not None
+    assert OPPER.default_api_base == "https://api.opper.ai/v3/compat"
+    assert OPPER.env_key == "OPPER_API_KEY"
+    assert OPPER.detect_by_base_keyword == "opper.ai"
+
+
+def test_opper_does_not_borrow_another_gateways_key_prefix() -> None:
+    # OpenRouter keys start with ``sk-or-``; Opper keys carry no such marker,
+    # so the prefix heuristic must stay empty.
+    assert OPPER is not None
+    assert OPPER.detect_by_key_prefix == ""
+
+
+def test_opper_sends_anthropic_cache_control() -> None:
+    # Opper serves Anthropic models and prices cached input separately, so the
+    # cache-control markers apply, as they do for the other gateways.
+    assert OPPER is not None
+    assert OPPER.supports_prompt_caching is True
+
+
+def test_opper_keeps_a_pinned_route_prefix() -> None:
+    """Unlike Forge, a ``provider/`` prefix on an Opper id is meaningful.
+
+    Opper ids are bare pool names such as ``claude-sonnet-4-6``, where a pool
+    is every provider serving that model. ``azure/gpt-5.5`` pins one provider
+    or region instead, so stripping the prefix would silently change routing.
+    """
+    assert OPPER is not None and FORGE is not None
+    assert FORGE.strip_model_prefix is True
+    assert OPPER.strip_model_prefix is False