GPTRouter 0.4.1__tar.gz → 0.4.3__tar.gz

{GPTRouter-0.4.1 → GPTRouter-0.4.3}/GPTRouter.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: GPTRouter
-Version: 0.4.1
+Version: 0.4.3
 Summary: A Python package for working with GPTRouter APIs
 Author-email: sirjanpreet@writesonic.com
 Classifier: Development Status :: 3 - Alpha

{GPTRouter-0.4.1 → GPTRouter-0.4.3}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: GPTRouter
-Version: 0.4.1
+Version: 0.4.3
 Summary: A Python package for working with GPTRouter APIs
 Author-email: sirjanpreet@writesonic.com
 Classifier: Development Status :: 3 - Alpha

{GPTRouter-0.4.1 → GPTRouter-0.4.3}/gpt_router/client.py

@@ -15,7 +15,7 @@ from gpt_router.models import (
     ChunkedGenerationResponse,
     ModelRouterEmbeddingsGenerationRequest,
     ModelRouterGenerationResponse,
-    ModelRouterEmbeddingsGenerationResponse
+    ModelRouterEmbeddingsGenerationResponse,
 )
 from gpt_router.exceptions import (
     GPTRouterApiTimeoutError,
@@ -77,9 +77,7 @@ class GPTRouterClient:
     async def agenerate_embeddings(
         self,
         *,
-        ordered_generation_requests: List[
-            ModelRouterEmbeddingsGenerationRequest
-        ],
+        ordered_generation_requests: List[ModelRouterEmbeddingsGenerationRequest],
         model_router_metadata: Optional[GPTRouterMetadata] = None,
     ) -> ModelRouterGenerationResponse:
         api_payload = {
@@ -184,6 +182,9 @@ class GPTRouterClient:
     async def astream_events(self, *, path: str, method: str, payload: dict, **kwargs):
         retry_count = kwargs.get("stream_retry_count") or 1
         stop_stream_signal = kwargs.get("stop_stream_signal")
+        switch_models_order_on_timeout = (
+            kwargs.get("switch_models_order_on_timeout") or False
+        )
         async_retrying = AsyncRetrying(
             stop=stop_after_attempt(retry_count),
             wait=wait_fixed(0.5),
@@ -252,6 +253,10 @@ class GPTRouterClient:
                 if attempt_number == retry_count:
                     logger.error("All retry attempts failed")
                     raise
+                elif switch_models_order_on_timeout:
+                    if attempt_number % 2:
+                        self._reorder_models_on_timeout(payload)
+                    logger.info(f"Retrying as per attempt #{attempt_number}")
             except httpx.HTTPStatusError as err:
                 raise Exception(
                     f"HTTP Error {err.response.status_code}: {err.response.text}"
@@ -267,6 +272,9 @@ class GPTRouterClient:
         self, *, path: str, method: str, payload: dict, **kwargs
     ) -> Generator[ChunkedGenerationResponse]:
         retry_count = kwargs.get("stream_retry_count") or 1
+        switch_models_order_on_timeout = (
+            kwargs.get("switch_models_order_on_timeout") or False
+        )
         retrying = Retrying(
             stop=stop_after_attempt(retry_count),
             wait=wait_fixed(0.5),
@@ -331,6 +339,10 @@ class GPTRouterClient:
                 if attempt_number == retry_count:
                     logger.error("All retry attempts failed")
                     raise
+                elif switch_models_order_on_timeout:
+                    if attempt_number % 2:
+                        self._reorder_models_on_timeout(payload)
+                    logger.info(f"Retrying as per attempt #{attempt_number}")
             except httpx.HTTPStatusError as err:
                 raise Exception(
                     f"HTTP Error {err.response.status_code}: {err.response.text}"
@@ -338,6 +350,30 @@ class GPTRouterClient:
             except httpx.TimeoutException as err:
                 raise TimeoutError("Request timed out") from err
 
+    def _reorder_models_on_timeout(self, payload: dict) -> None:
+        """Helper method to reorder models in payload when timeout occurs."""
+        if not payload:
+            payload = {}
+        data = payload.get("data", [])
+        try:
+            if (
+                len(data) >= 2
+                and isinstance(data[0], dict)
+                and isinstance(data[1], dict)
+            ):
+                data_0_order = data[0].get("order", None)
+                data_1_order = data[1].get("order", None)
+                if data_1_order and data_0_order:
+                    (
+                        payload["data"][0]["order"],
+                        payload["data"][1]["order"],
+                    ) = (
+                        payload["data"][1]["order"],
+                        payload["data"][0]["order"],
+                    )
+        except Exception as ex:
+            logger.exception(f"Error reordering models: {ex}")
+
     def generate(
         self,
         *,
@@ -430,7 +466,7 @@ class GPTRouterClient:
     def _extract_images_from_dict(self, image_dict: dict) -> List[dict]:
         if artifacts := image_dict.get("artifacts"):
             return artifacts
-        
+
         if image_urls := image_dict.get("imageUrls"):
             return [{"url": url} for url in image_urls]
 

{GPTRouter-0.4.1 → GPTRouter-0.4.3}/gpt_router/enums.py

@@ -56,6 +56,7 @@ class ModelsEnum(Enum):
 
     OPENAI_EMBEDDINGS_ADA_002 = "text-embedding-ada-002"
     AZURE_OPENAI_EMBEDDINGS_ADA_002 = "text-embedding-ada-002"
+    OPENAI_EMBEDDINGS_TEXT_SMALL_3 = "text-embedding-3-small"
 
 class StreamingEventType(Enum):
     UPDATE = "update"

{GPTRouter-0.4.1 → GPTRouter-0.4.3}/setup.py

@@ -5,7 +5,7 @@ with open("README.md", "r") as fh:
 
 setup(
     name="GPTRouter",
-    version="0.4.1",
+    version="0.4.3",
     description="A Python package for working with GPTRouter APIs",
     long_description=long_description,
     long_description_content_type="text/markdown",