Merge b19eb8576b into 252c71c0b2

.env.example

@@ -24,6 +24,9 @@ INFISICAL_TOKEN = ""
 NOVITA_API_KEY = ""
 # INFINITY
 INFINITY_API_KEY = ""
+# Eden AI
+EDENAI_API_KEY = ""
+EDENAI_API_BASE = "https://api.edenai.run/v3"
 
 # Development Configs
 LITELLM_MASTER_KEY = "sk-1234"

README.md

@@ -307,6 +307,7 @@ For MCP OAuth, an upstream may advertise dynamic client registration but refuse
 | [Deepgram (`deepgram`)](https://docs.litellm.ai/docs/providers/deepgram) | ✅ | ✅ | ✅ |  |  | ✅ |  |  |  |  |
 | [DeepInfra (`deepinfra`)](https://docs.litellm.ai/docs/providers/deepinfra) | ✅ | ✅ | ✅ |  |  |  |  |  |  |  |
 | [Deepseek (`deepseek`)](https://docs.litellm.ai/docs/providers/deepseek) | ✅ | ✅ | ✅ |  |  |  |  |  |  |  |
+| [Eden AI (`edenai`)](https://docs.litellm.ai/docs/providers/edenai) | ✅ | ✅ |  |  |  |  |  |  |  |  |
 | [ElevenLabs (`elevenlabs`)](https://docs.litellm.ai/docs/providers/elevenlabs) | ✅ | ✅ | ✅ |  |  | ✅ | ✅ |  |  |  |
 | [Empower (`empower`)](https://docs.litellm.ai/docs/providers/empower) | ✅ | ✅ | ✅ |  |  |  |  |  |  |  |
 | [Fal AI (`fal_ai`)](https://docs.litellm.ai/docs/providers/fal_ai) | ✅ | ✅ | ✅ |  | ✅ |  |  |  |  |  |

litellm/constants.py

@@ -639,6 +639,7 @@ LITELLM_CHAT_PROVIDERS: Final = [
     "huggingface",
     "together_ai",
     "datarobot",
+    "edenai",
     "helicone",
     "openrouter",
     "cometapi",
@@ -889,6 +890,7 @@ openai_compatible_endpoints: Final[list] = [
     "https://api.inceptionlabs.ai/v1",
     "https://api.hyperbolic.xyz/v1",
     "https://ai-gateway.helicone.ai/",
+    "https://api.edenai.run/v3",
     "https://ai-gateway.vercel.sh/v1",
     "https://api.inference.wandb.ai/v1",
     "https://api.clarifai.com/v2/ext/openai/v1",
@@ -935,6 +937,7 @@ openai_compatible_providers: Final[list] = [
     "novita",
     "meta_llama",
     "publicai",  # PublicAI - JSON-configured provider
+    "edenai",  # Eden AI - JSON-configured meta-gateway provider
     "synthetic",  # Synthetic - JSON-configured provider
     "tensormesh",  # Tensormesh - JSON-configured provider
     "apertis",  # Apertis - JSON-configured provider

litellm/llms/openai_like/providers.json

@@ -115,6 +115,14 @@
       "max_completion_tokens": "max_tokens"
     }
   },
+  "edenai": {
+    "base_url": "https://api.edenai.run/v3",
+    "api_key_env": "EDENAI_API_KEY",
+    "api_base_env": "EDENAI_API_BASE",
+    "param_mappings": {
+      "max_completion_tokens": "max_tokens"
+    }
+  },
   "darkbloom": {
     "base_url": "https://api.darkbloom.dev/v1",
     "api_key_env": "DARKBLOOM_API_KEY",

litellm/types/utils.py

@@ -4051,6 +4051,7 @@ class LlmProviders(str, Enum):
     LITELLM_AGENT = "litellm_agent"
     CURSOR = "cursor"
     BEDROCK_MANTLE = "bedrock_mantle"
+    EDENAI = "edenai"
     GDC = "gdc"
 
 

provider_endpoints_support.json

@@ -885,6 +885,22 @@
         "search": true
       }
     },
+    "edenai": {
+      "display_name": "Eden AI (`edenai`)",
+      "url": "https://docs.litellm.ai/docs/providers/edenai",
+      "endpoints": {
+        "chat_completions": true,
+        "messages": true,
+        "responses": false,
+        "embeddings": false,
+        "image_generations": false,
+        "audio_transcriptions": false,
+        "audio_speech": false,
+        "moderations": false,
+        "batches": false,
+        "rerank": false
+      }
+    },
     "e2b": {
       "display_name": "E2B (`e2b`)",
       "url": "https://docs.litellm.ai/docs/providers/e2b",

tests/e2e/test_edenai_e2e.py

@@ -0,0 +1,80 @@
+"""End-to-end (live) tests for the Eden AI provider.
+
+These hit the real Eden AI API and are skipped automatically when
+EDENAI_API_KEY is not set, so CI without the secret stays green.
+Moved here from tests/test_litellm/llms/openai_like/test_edenai.py per
+review feedback: live tests belong in tests/e2e/, not the unit suite.
+"""
+
+import os
+
+import pytest
+
+import litellm
+
+
+class TestEdenAILiveIntegration:
+    """Live integration tests against the real Eden AI API.
+
+    Skipped automatically when EDENAI_API_KEY is not set, so CI without the
+    secret stays green. When the key is present (local dev, or a dedicated
+    integration runner) these validate that each endpoint flagged `true` in
+    provider_endpoints_support.json actually routes a request through to
+    api.edenai.run/v3 end-to-end. Endpoints flagged `false` (`/responses`,
+    `/embeddings`, `/image/generations`, `/audio/*`, `/moderations`,
+    `/batches`, `/rerank`) intentionally have no live test here; their
+    handlers ship in follow-up PRs and the tests come with them.
+    """
+
+    @pytest.mark.skipif(
+        not os.environ.get("EDENAI_API_KEY"), reason="EDENAI_API_KEY not set"
+    )
+    def test_edenai_live_chat_completions(self):
+        """`/v1/chat/completions` via litellm.completion() — the openai_like
+        loader's primary surface."""
+        response = litellm.completion(
+            model="edenai/openai/gpt-4o-mini",
+            messages=[
+                {
+                    "role": "user",
+                    "content": "Reply with 'PING' and nothing else.",
+                }
+            ],
+            max_tokens=8,
+        )
+
+        assert response is not None
+        assert response.choices[0].message.content
+        assert response.usage is not None
+        assert response.usage.total_tokens > 0
+
+    @pytest.mark.skipif(
+        not os.environ.get("EDENAI_API_KEY"), reason="EDENAI_API_KEY not set"
+    )
+    @pytest.mark.asyncio
+    async def test_edenai_live_anthropic_messages(self):
+        """`/v1/messages` (Anthropic-format) via litellm.anthropic_messages().
+        LiteLLM translates the Anthropic-shape request to OpenAI shape
+        internally and routes through Eden AI's openai-compatible endpoint;
+        the response is translated back to Anthropic shape before being
+        returned to the caller."""
+        response = await litellm.anthropic_messages(
+            model="edenai/anthropic/claude-opus-4-6",
+            messages=[
+                {
+                    "role": "user",
+                    "content": "Reply with 'MSG' and nothing else.",
+                }
+            ],
+            max_tokens=8,
+        )
+
+        assert response is not None
+        # anthropic_messages returns a dict (Anthropic-shape response)
+        assert response.get("type") == "message"
+        assert response.get("role") == "assistant"
+        content = response.get("content", [])
+        assert content and content[0].get("type") == "text"
+        assert content[0].get("text")
+        usage = response.get("usage", {})
+        assert usage.get("input_tokens", 0) > 0

tests/test_litellm/llms/openai_like/test_edenai.py

@@ -0,0 +1,107 @@
+"""
+Tests for Eden AI provider configuration and integration.
+
+Eden AI (https://www.edenai.co/) is an OpenAI-compatible aggregator that
+exposes 100+ underlying models from OpenAI, Anthropic, Google, Mistral and
+others through a single chat-completions endpoint.
+"""
+
+import os
+
+import pytest
+
+import litellm
+
+
+class TestEdenAIProviderConfig:
+    """Test Eden AI provider configuration"""
+
+    def test_edenai_in_provider_list(self):
+        """Test that edenai is in the provider list"""
+        from litellm import LlmProviders
+
+        assert hasattr(LlmProviders, "EDENAI")
+        assert LlmProviders.EDENAI.value == "edenai"
+        assert "edenai" in litellm.provider_list
+
+    def test_edenai_json_config_exists(self):
+        """Test that edenai is configured in providers.json"""
+        from litellm.llms.openai_like.json_loader import JSONProviderRegistry
+
+        assert JSONProviderRegistry.exists("edenai")
+
+        edenai = JSONProviderRegistry.get("edenai")
+        assert edenai is not None
+        assert edenai.base_url == "https://api.edenai.run/v3"
+        assert edenai.api_key_env == "EDENAI_API_KEY"
+        assert edenai.api_base_env == "EDENAI_API_BASE"
+        assert edenai.param_mappings.get("max_completion_tokens") == "max_tokens"
+
+    def test_edenai_provider_resolution(self):
+        """Test that provider resolution finds edenai for prefixed model ids"""
+        from litellm.litellm_core_utils.get_llm_provider_logic import get_llm_provider
+
+        model, provider, api_key, api_base = get_llm_provider(
+            model="edenai/openai/gpt-4o-mini",
+            custom_llm_provider=None,
+            api_base=None,
+            api_key=None,
+        )
+
+        # Eden AI's model ids are themselves prefixed (e.g. openai/gpt-4o-mini),
+        # so the model returned is everything after the leading "edenai/".
+        assert model == "openai/gpt-4o-mini"
+        assert provider == "edenai"
+        assert api_base == "https://api.edenai.run/v3"
+
+    def test_edenai_api_base_env_override(self, monkeypatch):
+        """Test that EDENAI_API_BASE overrides the default base URL"""
+        from litellm.litellm_core_utils.get_llm_provider_logic import get_llm_provider
+
+        monkeypatch.setenv("EDENAI_API_BASE", "https://custom-edenai.example.com/v3")
+        _, provider, _, api_base = get_llm_provider(
+            model="edenai/openai/gpt-4o-mini",
+            custom_llm_provider=None,
+            api_base=None,
+            api_key=None,
+        )
+        assert provider == "edenai"
+        assert api_base == "https://custom-edenai.example.com/v3"
+
+    def test_edenai_router_config(self):
+        """Test that edenai can be used in Router configuration"""
+        from litellm import Router
+
+        router = Router(
+            model_list=[
+                {
+                    "model_name": "edenai-gpt-4o-mini",
+                    "litellm_params": {
+                        "model": "edenai/openai/gpt-4o-mini",
+                        "api_key": "test-key",
+                    },
+                }
+            ]
+        )
+
+        assert len(router.model_list) == 1
+        assert router.model_list[0]["model_name"] == "edenai-gpt-4o-mini"
+
+
+class TestEdenAICompletionMocked:
+    """Mocked completion tests — no real API calls."""
+
+    def test_edenai_completion_mocked(self):
+        """Test that a mocked completion through the edenai/ prefix routes
+        correctly and returns the stub content unchanged."""
+        response = litellm.completion(
+            model="edenai/openai/gpt-4o-mini",
+            messages=[{"role": "user", "content": "Say PING and nothing else"}],
+            api_key="test-key",
+            mock_response="PING",
+        )
+
+        assert response is not None
+        assert hasattr(response, "choices")
+        assert len(response.choices) > 0
+        assert response.choices[0].message.content == "PING"