From 75ee313f4c9fea331afe0cf44abd71034487efbe Mon Sep 17 00:00:00 2001
From: aiapienthusiast <327629069+aiapienthusiast@users.noreply.github.com>
Date: Mon, 19 Oct 2026 19:02:38 +0400
Subject: [PATCH] feat(llm): add Cheaper Inference provider (#7088)

---
 app/agent/llm/catalog.py                    |  8 +++
 tests/test_llm_provider_cheaperinference.py | 60 +++++++++++++++++++++
 2 files changed, 68 insertions(+)
 create mode 100644 tests/test_llm_provider_cheaperinference.py

diff --git a/app/agent/llm/catalog.py b/app/agent/llm/catalog.py
index 71976a32c..73c095b96 100644
--- a/app/agent/llm/catalog.py
+++ b/app/agent/llm/catalog.py
@@ -407,6 +407,14 @@ class _ProviderCatalog:
                 api_key_hint="填写 API Route API Key，可在 https://api-route.com 获取。",
                 description="API Route OpenAI-compatible 聚合模型平台，支持刷新账户可用模型目录。",
             ),
+            openai_provider(
+                provider_id="cheaperinference",
+                name="Cheaper Inference",
+                default_base_url="https://api.cheaperinference.com/v1",
+                sort_order=57,
+                api_key_hint="填写 Cheaper Inference API Key，可在 https://cheaperinference.com/signup 获取。",
+                description="Cheaper Inference OpenAI-compatible 聚合模型平台，一个 Key 覆盖多家厂商的模型。",
+            ),
             ProviderSpec(
                 id="github-copilot",
                 name="GitHub Copilot",
diff --git a/tests/test_llm_provider_cheaperinference.py b/tests/test_llm_provider_cheaperinference.py
new file mode 100644
index 000000000..c2278a62a
--- /dev/null
+++ b/tests/test_llm_provider_cheaperinference.py
@@ -0,0 +1,60 @@
+"""Cheaper Inference 预设通过共享 OpenAI 兼容通道发现模型的边界回归。"""
+
+from unittest.mock import AsyncMock
+
+import httpx
+import pytest
+
+from app.agent.llm.provider import LLMProviderManager
+from app.foundation.singleton import Singleton
+
+
+@pytest.fixture
+def manager(monkeypatch):
+    """隔离管理器单例及目录缓存，防止测试影响其他 provider。"""
+    key = (LLMProviderManager, (), frozenset())
+    monkeypatch.setattr(Singleton, "_instances", {k: v for k, v in Singleton._instances.items() if k != key})
+    instance = LLMProviderManager()
+    monkeypatch.setattr(instance, "get_models_dev_data", AsyncMock(return_value={}))
+    return instance
+
+
+@pytest.mark.asyncio
+@pytest.mark.parametrize(
+    "base_url, expected_url",
+    [
+        (None, "https://api.cheaperinference.com/v1/models"),
+        ("https://relay.example/v1", "https://relay.example/v1/models"),
+    ],
+)
+async def test_cheaperinference_discovers_authenticated_models(manager, monkeypatch, base_url, expected_url):
+    """预设及覆盖地址都传递密钥，保留完整模型 ID，不推断缺失的上下文元数据。"""
+    requests = []
+
+    def respond(request):
+        requests.append(request)
+        return httpx.Response(
+            200,
+            json={
+                "object": "list",
+                "data": [
+                    {"id": "gpt-5.4-mini", "object": "model", "created": 0, "owned_by": "cheaperinference"},
+                    {"id": "other/model", "object": "model", "created": 0, "owned_by": "cheaperinference"},
+                ],
+            },
+        )
+
+    monkeypatch.setattr(
+        manager,
+        "_build_sdk_http_client",
+        lambda use_proxy: httpx.AsyncClient(transport=httpx.MockTransport(respond)),
+    )
+
+    models = await manager.list_models("cheaperinference", api_key="test-cheaperinference-key", base_url=base_url)
+
+    assert len(requests) == 1
+    assert requests[0].method == "GET"
+    assert str(requests[0].url) == expected_url
+    assert requests[0].headers["Authorization"] == "Bearer test-cheaperinference-key"
+    assert [model["id"] for model in models] == ["gpt-5.4-mini", "other/model"]
+    assert all(model["source"] == "provider" and model["context_tokens"] is None for model in models)