feat(providers): add Paxa as an OpenAI compatible speech provider

paxa/<model> now resolves to https://api.paxalabs.com/v1 with PAXA_API_KEY, and PAXA_API_BASE overrides the base. Requests go through the existing OpenAI speech path

speech() also uses the key that get_llm_provider resolved for an OpenAI compatible provider, as completion and transcription already do

README.md

@@ -398,6 +398,7 @@ Set `LITELLM_PROXY_API_BASE` and `LITELLM_PROXY_API_KEY` and every model call th
 | [OpenAI-like (`openai_like`)](https://docs.litellm.ai/docs/providers/openai_compatible) |  |  |  | ✅ |  |  |  |  |  |  |
 | [OpenRouter (`openrouter`)](https://docs.litellm.ai/docs/providers/openrouter) | ✅ | ✅ | ✅ |  |  |  |  |  |  |  |
 | [OVHCloud AI Endpoints (`ovhcloud`)](https://docs.litellm.ai/docs/providers/ovhcloud) | ✅ | ✅ | ✅ |  |  |  |  |  |  |  |
+| [Paxa (`paxa`)](https://docs.litellm.ai/docs/providers/paxa) |  |  |  |  |  |  | ✅ |  |  |  |
 | [Perplexity AI (`perplexity`)](https://docs.litellm.ai/docs/providers/perplexity) | ✅ | ✅ | ✅ |  |  |  |  |  |  |  |
 | [Petals (`petals`)](https://docs.litellm.ai/docs/providers/petals) | ✅ | ✅ | ✅ |  |  |  |  |  |  |  |
 | [Pinstripes (`pinstripes`)](https://docs.litellm.ai/docs/providers/pinstripes) | ✅ | ✅ | ✅ |  |  |  |  |  |  |  |

litellm/constants.py

@@ -1138,6 +1138,7 @@ openai_compatible_providers: Final[list[str]] = [
     "prism",
     "sail",
     "reka",
+    "paxa",
 ]
 
 OPENAI_AUDIO_TRANSCRIPTION_PROVIDERS: Final = frozenset({"openai"} | frozenset(openai_compatible_providers))

litellm/litellm_core_utils/get_llm_provider_logic.py

@@ -670,6 +670,10 @@ def _get_openai_compatible_provider_info(
     elif custom_llm_provider == "soniox":
         api_base = api_base or get_secret_str("SONIOX_API_BASE") or "https://api.soniox.com"
         dynamic_api_key = api_key or get_secret_str("SONIOX_API_KEY")
+    elif custom_llm_provider == "paxa":
+        default_paxa_base: Final = get_secret_str("PAXA_API_BASE") or "https://api.paxalabs.com/v1"
+        api_base = api_base or default_paxa_base  # rebind-ok: mirrors sibling provider branches
+        dynamic_api_key = api_key or get_secret_str("PAXA_API_KEY")  # rebind-ok: mirrors sibling provider branches
     elif custom_llm_provider == "cerebras":
         api_base = api_base or get_secret("CEREBRAS_API_BASE") or "https://api.cerebras.ai/v1"
         dynamic_api_key = api_key or get_secret_str("CEREBRAS_API_KEY")

litellm/main.py

@@ -8680,6 +8680,7 @@ def speech(
         # set API KEY
         api_key = (
             api_key
+            or dynamic_api_key
             or litellm.api_key  # for deepinfra/perplexity/anyscale we check in get_llm_provider and pass in the api key from there
             or litellm.openai_key
             or get_secret("OPENAI_API_KEY")

litellm/proxy/public_endpoints/provider_create_fields.json

@@ -3027,6 +3027,24 @@
     ],
     "default_model_placeholder": "gpt-3.5-turbo"
   },
+  {
+    "provider": "Paxa",
+    "provider_display_name": "Paxa",
+    "litellm_provider": "paxa",
+    "credential_fields": [
+      {
+        "key": "api_key",
+        "label": "API Key",
+        "placeholder": null,
+        "tooltip": null,
+        "required": true,
+        "field_type": "password",
+        "options": null,
+        "default_value": null
+      }
+    ],
+    "default_model_placeholder": "paxa/paxa-tts-flash-v2"
+  },
   {
     "provider": "Perplexity",
     "provider_display_name": "Perplexity",

litellm/types/utils.py

@@ -4440,6 +4440,7 @@ class LlmProviders(str, Enum):
     NVIDIA_NIM = "nvidia_nim"
     NVIDIA_RIVA = "nvidia_riva"
     SONIOX = "soniox"
+    PAXA = "paxa"
     CEREBRAS = "cerebras"
     NADIR = "nadir"
     AI21_CHAT = "ai21_chat"

provider_endpoints_support.json

@@ -2095,6 +2095,22 @@
         "a2a": false
       }
     },
+    "paxa": {
+      "display_name": "Paxa (`paxa`)",
+      "url": "https://docs.litellm.ai/docs/providers/paxa",
+      "endpoints": {
+        "chat_completions": false,
+        "messages": false,
+        "responses": false,
+        "embeddings": false,
+        "image_generations": false,
+        "audio_transcriptions": false,
+        "audio_speech": true,
+        "moderations": false,
+        "batches": false,
+        "rerank": false
+      }
+    },
     "perplexity": {
       "display_name": "Perplexity AI (`perplexity`)",
       "url": "https://docs.litellm.ai/docs/providers/perplexity",

tests/unit/litellm_core_utils/test_get_llm_provider_logic.py

@@ -74,6 +74,20 @@ def test_get_llm_provider_leaves_fal_ai_api_base_unset_for_global_fallback() ->
     assert explicit == "http://edge.local/fal"
 
 
+def test_get_llm_provider_reads_the_paxa_key_and_base_from_env(monkeypatch: pytest.MonkeyPatch) -> None:
+    monkeypatch.setenv("PAXA_API_KEY", "fake-paxa-key")
+    monkeypatch.delenv("PAXA_API_BASE", raising=False)
+    assert get_llm_provider(model="paxa/paxa-tts-flash-v2") == (
+        "paxa-tts-flash-v2",
+        "paxa",
+        "fake-paxa-key",
+        "https://api.paxalabs.com/v1",
+    )
+
+    monkeypatch.setenv("PAXA_API_BASE", "http://paxa.local/v1")
+    assert get_llm_provider(model="paxa/paxa-tts-flash-v2")[3] == "http://paxa.local/v1"
+
+
 def test_image_generation_fal_ai_egresses_to_global_api_base(monkeypatch: pytest.MonkeyPatch) -> None:
     monkeypatch.setattr(litellm, "api_base", "http://gateway.local/fal")
     monkeypatch.setenv("FAL_AI_API_KEY", "test")

tests/unit/test_main.py

@@ -5223,6 +5223,24 @@ def test_groq_speech_honors_base_url_alias(respx_mock: respx.MockRouter):
     assert response.content == b"mp3-bytes"
 
 
+def test_paxa_speech_uses_the_paxa_key_from_the_environment(
+    respx_mock: respx.MockRouter, monkeypatch: pytest.MonkeyPatch
+) -> None:
+    monkeypatch.setenv("PAXA_API_KEY", "fake-paxa-key")
+    monkeypatch.delenv("PAXA_API_BASE", raising=False)
+    monkeypatch.delenv("OPENAI_API_KEY", raising=False)
+    route: Final = respx_mock.post("https://api.paxalabs.com/v1/audio/speech").mock(
+        return_value=httpx.Response(200, content=b"mp3-bytes")
+    )
+
+    response: Final = litellm.speech(model="paxa/paxa-tts-flash-v2", input="hello", voice="foithong")
+
+    request: Final = route.calls.last.request
+    assert request.headers["authorization"] == "Bearer fake-paxa-key"
+    assert json.loads(request.content) == {"model": "paxa-tts-flash-v2", "input": "hello", "voice": "foithong"}
+    assert response.content == b"mp3-bytes"
+
+
 FORWARDED_CLIENT_HEADERS: Final = {"x-forwarded-for": "10.0.0.1", "x-amzn-trace-id": "Root=1-lit7694"}