feat(providers): add Phala as an OpenAI-compatible provider

litellm/constants.py

@@ -1057,6 +1057,7 @@ openai_compatible_endpoints: Final[list[str]] = [
     "https://api.cognition.ai/v1",
     "https://inference.coralbricks.ai/v1",
     "https://api.cortecs.ai/v1",
+    "https://inference.phala.com/v1",
     "https://api.scx.ai/v1",
     "https://api.prisminference.com/v1",
     "https://api.reka.ai/v1",
@@ -1137,6 +1138,7 @@ openai_compatible_providers: Final[list[str]] = [
     "cognition",
     "coralbricks",
     "cortecs",
+    "phala",
     "scx-ai",
     "prism",
     "sail",

litellm/llms/openai_like/providers.json

@@ -192,6 +192,15 @@
     "api_base_env": "CORTECS_API_BASE",
     "supported_endpoints": ["/v1/chat/completions", "/v1/responses", "/v1/messages"]
   },
+  "phala": {
+    "base_url": "https://inference.phala.com/v1",
+    "api_key_env": "PHALA_AI_API_KEY",
+    "api_base_env": "PHALA_API_BASE",
+    "provider_info": {
+      "supports_reasoning": true
+    },
+    "supported_endpoints": ["/v1/chat/completions"]
+  },
   "pinstripes": {
     "base_url": "https://pinstripes.io/v1",
     "api_key_env": "PINSTRIPES_API_KEY",

litellm/provider_endpoints_support_backup.json

@@ -636,6 +636,24 @@
         "interactions": false
       }
     },
+    "phala": {
+      "display_name": "Phala (`phala`)",
+      "url": "https://docs.phala.com/phala-cloud/confidential-ai/confidential-model/confidential-ai-api",
+      "endpoints": {
+        "chat_completions": true,
+        "messages": false,
+        "responses": false,
+        "embeddings": false,
+        "image_generations": false,
+        "audio_transcriptions": false,
+        "audio_speech": false,
+        "moderations": false,
+        "batches": false,
+        "rerank": false,
+        "a2a": false,
+        "interactions": false
+      }
+    },
     "cortecs": {
       "display_name": "Cortecs (`cortecs`)",
       "url": "https://docs.litellm.ai/docs/providers/cortecs",

litellm/proxy/public_endpoints/provider_create_fields.json

@@ -3119,6 +3119,34 @@
     ],
     "default_model_placeholder": "gpt-3.5-turbo"
   },
+  {
+    "provider": "PHALA",
+    "provider_display_name": "Phala",
+    "litellm_provider": "phala",
+    "credential_fields": [
+      {
+        "key": "api_base",
+        "label": "API Base",
+        "placeholder": "https://inference.phala.com/v1",
+        "tooltip": null,
+        "required": false,
+        "field_type": "text",
+        "options": null,
+        "default_value": null
+      },
+      {
+        "key": "api_key",
+        "label": "API Key",
+        "placeholder": null,
+        "tooltip": null,
+        "required": true,
+        "field_type": "password",
+        "options": null,
+        "default_value": null
+      }
+    ],
+    "default_model_placeholder": "phala/z-ai/glm-5.3"
+  },
   {
     "provider": "PREDIBASE",
     "provider_display_name": "Predibase",

litellm/types/utils.py

@@ -4578,6 +4578,7 @@ class LlmProviders(str, Enum):
     COGNITION = "cognition"
     CORALBRICKS = "coralbricks"
     CORTECS = "cortecs"
+    PHALA = "phala"
     SCX_AI = "scx-ai"
     PRISM = "prism"
     OPENINFER = "openinfer"

provider_endpoints_support.json

@@ -2166,6 +2166,24 @@
         "a2a": false
       }
     },
+    "phala": {
+      "display_name": "Phala (`phala`)",
+      "url": "https://docs.phala.com/phala-cloud/confidential-ai/confidential-model/confidential-ai-api",
+      "endpoints": {
+        "chat_completions": true,
+        "messages": false,
+        "responses": false,
+        "embeddings": false,
+        "image_generations": false,
+        "audio_transcriptions": false,
+        "audio_speech": false,
+        "moderations": false,
+        "batches": false,
+        "rerank": false,
+        "a2a": false,
+        "interactions": false
+      }
+    },
     "pinstripes": {
       "display_name": "Pinstripes (`pinstripes`)",
       "url": "https://docs.litellm.ai/docs/providers/pinstripes",

tests/unit/llms/openai_like/test_phala_provider.py

@@ -0,0 +1,128 @@
+import json
+from typing import Final
+
+import pytest
+import respx
+
+import litellm
+from litellm.litellm_core_utils.get_llm_provider_logic import get_llm_provider
+
+PHALA_API_BASE: Final = "https://inference.phala.com/v1"
+PHALA_MODEL_ID: Final = "z-ai/glm-5.3"
+PHALA_MODEL: Final = f"phala/{PHALA_MODEL_ID}"
+
+
+@pytest.mark.parametrize("custom_provider", (None, "phala"))
+def test_phala_resolves_environment_credentials(monkeypatch: pytest.MonkeyPatch, custom_provider: str | None) -> None:
+    monkeypatch.setenv("PHALA_AI_API_KEY", "phala-env-key")
+
+    model, provider, api_key, api_base = get_llm_provider(
+        model=PHALA_MODEL if custom_provider is None else PHALA_MODEL_ID,
+        custom_llm_provider=custom_provider,
+    )
+
+    assert (model, provider, api_key, api_base) == (PHALA_MODEL_ID, "phala", "phala-env-key", PHALA_API_BASE)
+
+
+def test_phala_detected_from_api_base(monkeypatch: pytest.MonkeyPatch) -> None:
+    monkeypatch.setenv("PHALA_AI_API_KEY", "phala-env-key")
+
+    model, provider, api_key, api_base = get_llm_provider(model=PHALA_MODEL_ID, api_base=PHALA_API_BASE)
+
+    assert (model, provider, api_key, api_base) == (PHALA_MODEL_ID, "phala", "phala-env-key", PHALA_API_BASE)
+
+
+def test_phala_explicit_credentials_override_environment(monkeypatch: pytest.MonkeyPatch) -> None:
+    monkeypatch.setenv("PHALA_AI_API_KEY", "phala-env-key")
+    monkeypatch.setenv("PHALA_API_BASE", "https://environment.example/v1")
+
+    model, provider, api_key, api_base = get_llm_provider(
+        model=PHALA_MODEL,
+        api_base="https://explicit.example/v1",
+        api_key="phala-explicit-key",
+    )
+
+    assert (model, provider, api_key, api_base) == (
+        PHALA_MODEL_ID,
+        "phala",
+        "phala-explicit-key",
+        "https://explicit.example/v1",
+    )
+
+
+@pytest.mark.asyncio
+@pytest.mark.parametrize("is_async", (False, True))
+@pytest.mark.parametrize("api_base", (PHALA_API_BASE, "https://override.example/v1"))
+async def test_phala_chat_completion_request(monkeypatch: pytest.MonkeyPatch, is_async: bool, api_base: str) -> None:
+    monkeypatch.setenv("PHALA_AI_API_KEY", "phala-test-key")
+    monkeypatch.setenv("PHALA_API_BASE", api_base)
+    monkeypatch.setattr(litellm, "disable_aiohttp_transport", True)
+    with respx.mock() as upstream:
+        route: Final = upstream.post(f"{api_base}/chat/completions").respond(
+            200,
+            json={
+                "id": "chatcmpl_phala",
+                "object": "chat.completion",
+                "created": 1_789_550_000,
+                "model": PHALA_MODEL_ID,
+                "choices": [
+                    {
+                        "index": 0,
+                        "message": {"role": "assistant", "content": "Hello from Phala"},
+                        "finish_reason": "stop",
+                    }
+                ],
+                "usage": {"prompt_tokens": 4, "completion_tokens": 3, "total_tokens": 7},
+            },
+        )
+        response: Final = (
+            await litellm.acompletion(
+                model=PHALA_MODEL,
+                messages=[{"role": "user", "content": "Say hello"}],
+                reasoning_effort="low",
+                reasoning={"effort": "low"},
+            )
+            if is_async
+            else litellm.completion(
+                model=PHALA_MODEL,
+                messages=[{"role": "user", "content": "Say hello"}],
+                reasoning_effort="low",
+                reasoning={"effort": "low"},
+            )
+        )
+
+    request: Final = route.calls.last.request
+    assert route.call_count == 1
+    assert request.headers["authorization"] == "Bearer phala-test-key"
+    assert json.loads(request.content) == {
+        "model": PHALA_MODEL_ID,
+        "messages": [{"role": "user", "content": "Say hello"}],
+        "reasoning_effort": "low",
+        "reasoning": {"effort": "low"},
+    }
+    assert isinstance(response, litellm.ModelResponse)
+    assert response.choices[0].message.content == "Hello from Phala"
+    assert response.usage.total_tokens == 7
+
+
+@pytest.mark.asyncio
+async def test_phala_available_in_add_model() -> None:
+    from litellm.proxy.public_endpoints.public_endpoints import get_provider_fields, get_supported_providers
+
+    assert "phala" in await get_supported_providers()
+    providers: Final = await get_provider_fields()
+    phala: Final = next(provider for provider in providers if provider.litellm_provider == "phala")
+
+    assert phala.provider_display_name == "Phala"
+    assert phala.default_model_placeholder == PHALA_MODEL
+    assert {field.key: field.required for field in phala.credential_fields} == {"api_base": False, "api_key": True}
+
+
+@pytest.mark.asyncio
+async def test_phala_listed_for_chat_completions() -> None:
+    from litellm.proxy.public_endpoints.public_endpoints import get_supported_endpoints
+
+    endpoints: Final = await get_supported_endpoints()
+    chat: Final = next(endpoint for endpoint in endpoints.endpoints if endpoint.key == "chat_completions")
+
+    assert any(provider.slug == "phala" and provider.display_name == "Phala" for provider in chat.providers)