streaming and multi-base-URL support

client_gen_config.md

@@ -73,4 +73,17 @@ directive:
     where: '$.paths."/v2/apps".post'
     transform: >
       $["parameters"] = [];
+
+  # Strip path-level and operation-level "servers" overrides from all paths.
+  # Without this, Autorest generates a duplicate `endpoint` parameter in
+  # GeneratedClient.__init__ (one per unique server URL), which is a Python
+  # syntax error.  Runtime multi-base-URL routing is handled by
+  # _BaseURLProxy in custom_extensions.py instead.
+  - from: openapi-document
+    where: '$.paths.*'
+    transform: >
+      delete $["servers"];
+      for (const m of ["get","post","put","patch","delete","head","options"]) {
+        if ($[m] && $[m].servers) { delete $[m].servers; }
+      }
 ```

src/pydo/_patch.py

@@ -11,6 +11,7 @@
 from azure.core.credentials import AccessToken
 
 from pydo.custom_policies import CustomHttpLoggingPolicy
+from pydo.custom_extensions import _BaseURLProxy, INFERENCE_BASE_URL
 from pydo import GeneratedClient, _version
 
 if TYPE_CHECKING:
@@ -32,13 +33,28 @@ def get_token(self, *args, **kwargs) -> AccessToken:
 class Client(GeneratedClient):  # type: ignore
     """The official DigitalOcean Python client
 
-    :param token: A valid API token.
+    :param token: A valid API token / model access key.
     :type token: str
     :keyword endpoint: Service URL. Default value is "https://api.digitalocean.com".
     :paramtype endpoint: str
+    :keyword inference_endpoint: Serverless inference URL.
+        Default value is "https://inference.do-ai.run".
+    :paramtype inference_endpoint: str
+    :keyword agent_endpoint: Agent inference URL.  Pass the per-agent
+        subdomain (e.g. ``"https://<id>.agents.do-ai.run"``).
+        Required only when using agent inference endpoints.
+    :paramtype agent_endpoint: str
     """
 
-    def __init__(self, token: str, *, timeout: int = 120, **kwargs):
+    def __init__(
+        self,
+        token: str,
+        *,
+        timeout: int = 120,
+        inference_endpoint: str = INFERENCE_BASE_URL,
+        agent_endpoint: str = "",
+        **kwargs,
+    ):
         logger = kwargs.get("logger")
         if logger is not None and kwargs.get("http_logging_policy") == "":
             kwargs["http_logging_policy"] = CustomHttpLoggingPolicy(logger=logger)
@@ -48,6 +64,41 @@ def __init__(self, token: str, *, timeout: int = 120, **kwargs):
             TokenCredentials(token), timeout=timeout, sdk_moniker=sdk_moniker, **kwargs
         )
 
+        self._setup_inference_routing(inference_endpoint, agent_endpoint)
+
+    def _setup_inference_routing(
+        self,
+        inference_endpoint: str,
+        agent_endpoint: str,
+    ) -> None:
+        """Route Inference / AgentInference operation groups to their endpoints.
+
+        * ``*Inference*`` (but not ``*AgentInference*``) → *inference_endpoint*
+        * ``*AgentInference*`` → *agent_endpoint*
+
+        Both use the same token passed to ``Client.__init__``.
+        """
+        inference_proxy = _BaseURLProxy(
+            self._client,
+            inference_endpoint,
+        )
+
+        agent_proxy = None
+        if agent_endpoint:
+            agent_proxy = _BaseURLProxy(
+                self._client,
+                agent_endpoint,
+            )
+
+        for attr in self.__dict__.values():
+            if not hasattr(attr, "_client"):
+                continue
+            class_name = type(attr).__name__
+            if class_name.startswith("AgentInference") and agent_proxy:
+                attr._client = agent_proxy
+            elif class_name.startswith("Inference"):
+                attr._client = inference_proxy
+
 
 __all__ = ["Client"]
 

src/pydo/aio/_patch.py

@@ -13,6 +13,7 @@
 
 from pydo import _version
 from pydo.custom_policies import CustomHttpLoggingPolicy
+from pydo.custom_extensions import _BaseURLProxy, INFERENCE_BASE_URL
 from pydo.aio import GeneratedClient
 
 if TYPE_CHECKING:
@@ -34,15 +35,30 @@ async def get_token(self, *args, **kwargs) -> AccessToken:
 
 
 class Client(GeneratedClient):  # type: ignore
-    """The official DigitalOcean Python client
+    """The official DigitalOcean Python client (async)
 
-    :param token: A valid API token.
+    :param token: A valid API token / model access key.
     :type token: str
     :keyword endpoint: Service URL. Default value is "https://api.digitalocean.com".
     :paramtype endpoint: str
+    :keyword inference_endpoint: Serverless inference URL.
+        Default value is "https://inference.do-ai.run".
+    :paramtype inference_endpoint: str
+    :keyword agent_endpoint: Agent inference URL.  Pass the per-agent
+        subdomain (e.g. ``"https://<id>.agents.do-ai.run"``).
+        Required only when using agent inference endpoints.
+    :paramtype agent_endpoint: str
     """
 
-    def __init__(self, token: str, *, timeout: int = 120, **kwargs):
+    def __init__(
+        self,
+        token: str,
+        *,
+        timeout: int = 120,
+        inference_endpoint: str = INFERENCE_BASE_URL,
+        agent_endpoint: str = "",
+        **kwargs,
+    ):
         logger = kwargs.get("logger")
         if logger is not None and kwargs.get("http_logging_policy") == "":
             kwargs["http_logging_policy"] = CustomHttpLoggingPolicy(logger=logger)
@@ -52,6 +68,41 @@ def __init__(self, token: str, *, timeout: int = 120, **kwargs):
             TokenCredentials(token), timeout=timeout, sdk_moniker=sdk_moniker, **kwargs
         )
 
+        self._setup_inference_routing(inference_endpoint, agent_endpoint)
+
+    def _setup_inference_routing(
+        self,
+        inference_endpoint: str,
+        agent_endpoint: str,
+    ) -> None:
+        """Route Inference / AgentInference operation groups to their endpoints.
+
+        * ``*Inference*`` (but not ``*AgentInference*``) → *inference_endpoint*
+        * ``*AgentInference*`` → *agent_endpoint*
+
+        Both use the same token passed to ``Client.__init__``.
+        """
+        inference_proxy = _BaseURLProxy(
+            self._client,
+            inference_endpoint,
+        )
+
+        agent_proxy = None
+        if agent_endpoint:
+            agent_proxy = _BaseURLProxy(
+                self._client,
+                agent_endpoint,
+            )
+
+        for attr in self.__dict__.values():
+            if not hasattr(attr, "_client"):
+                continue
+            class_name = type(attr).__name__
+            if class_name.startswith("AgentInference") and agent_proxy:
+                attr._client = agent_proxy
+            elif class_name.startswith("Inference"):
+                attr._client = inference_proxy
+
 
 # Add all objects you want publicly available to users at this package level
 __all__ = ["Client"]  # type: List[str]

src/pydo/aio/operations/_patch.py

@@ -5,16 +5,96 @@
 """Customize generated code here.
 
 Follow our quickstart for examples: https://aka.ms/azsdk/python/dpcodegen/python/customize
+
+Async mirror of ``pydo/operations/_patch.py``.  See that module for the
+full design rationale.  If no inference / agent-inference operations have
+been generated, this module exports nothing.
 """
 from typing import TYPE_CHECKING
 
 if TYPE_CHECKING:
     # pylint: disable=unused-import,ungrouped-imports
     from typing import List
 
-__all__ = (
-    []
-)  # type: List[str]  # Add all objects you want publicly available to users at this package level
+# ---------------------------------------------------------------------------
+# Serverless Inference operations (async)
+# ---------------------------------------------------------------------------
+
+try:
+    from ._operations import InferenceOperations as _GeneratedInferenceOperations
+
+    import pydo.operations._operations as _ops
+
+    _HAS_INFERENCE = True
+except ImportError:
+    _HAS_INFERENCE = False
+
+if _HAS_INFERENCE:
+    from pydo.custom_extensions import (
+        AsyncStreamingMixin,
+        install_streaming_wrappers,
+    )
+
+    class InferenceOperations(AsyncStreamingMixin, _GeneratedInferenceOperations):
+        """Async InferenceOperations with fully automatic streaming support.
+
+        Mirror of the sync version in ``pydo/operations/_patch.py``.
+        """
+
+        def __init__(self, *args, **kwargs):
+            super().__init__(*args, **kwargs)
+            install_streaming_wrappers(
+                self, _GeneratedInferenceOperations, _ops, is_async=True
+            )
+
+
+# ---------------------------------------------------------------------------
+# Agent Inference operations (async)
+# ---------------------------------------------------------------------------
+
+try:
+    from ._operations import (
+        AgentInferenceOperations as _GeneratedAgentInferenceOperations,
+    )
+
+    if not _HAS_INFERENCE:
+        import pydo.operations._operations as _ops
+
+    _HAS_AGENT_INFERENCE = True
+except ImportError:
+    _HAS_AGENT_INFERENCE = False
+
+if _HAS_AGENT_INFERENCE:
+    if not _HAS_INFERENCE:
+        from pydo.custom_extensions import (
+            AsyncStreamingMixin,
+            install_streaming_wrappers,
+        )
+
+    class AgentInferenceOperations(
+        AsyncStreamingMixin, _GeneratedAgentInferenceOperations
+    ):
+        """Async AgentInferenceOperations with fully automatic streaming support.
+
+        Mirror of the sync version in ``pydo/operations/_patch.py``.
+        """
+
+        def __init__(self, *args, **kwargs):
+            super().__init__(*args, **kwargs)
+            install_streaming_wrappers(
+                self, _GeneratedAgentInferenceOperations, _ops, is_async=True
+            )
+
+
+# ---------------------------------------------------------------------------
+# Exports
+# ---------------------------------------------------------------------------
+
+__all__ = []  # type: ignore[assignment]
+if _HAS_INFERENCE:
+    __all__.append("InferenceOperations")
+if _HAS_AGENT_INFERENCE:
+    __all__.append("AgentInferenceOperations")
 
 
 def patch_sdk():