Merge pull request #214 from MadcowD/wguss/azure

Azure

examples/providers/azure_ex.py

@@ -0,0 +1,37 @@
+import ell
+import openai
+import os
+ell.init(verbose=True, store='./logdir')
+
+# your subscription key
+subscription_key = os.getenv("AZURE_OPENAI_API_KEY")
+# Your Azure OpenAI resource https://<your resource name>.openai.azure.com/
+azure_endpoint = "https://<your resource name>.openai.azure.com/"
+# Option 2: Use a client directly
+azureClient = openai.AzureOpenAI(
+    azure_endpoint = azure_endpoint,
+    api_key = subscription_key,
+    api_version = "2024-05-01-preview",
+)
+# (Recommended) Option 1: Register all the models on your Azure resource & use your models automatically
+ell.config.register_model("<your-azure-model-deployment-name>", azureClient)
+
+@ell.simple(model="<your-azure-model-deployment-name>")
+def write_a_story(about : str):
+    return f"write me a story about {about}!"
+
+write_a_story("cats")
+
+
+# Option 2: Use a client directly
+azureClient = openai.AzureOpenAI(
+    azure_endpoint = azure_endpoint,
+    api_key = subscription_key,
+    api_version = "2024-05-01-preview",
+)
+
+@ell.simple(model="<your-azure-model-deployment-name>", client=azureClient)
+def write_a_story(about : str):
+    return f"write me a story about {about}"
+
+write_a_story("cats")

examples/providers/groq_ex.py

@@ -6,6 +6,8 @@
 
 
 ell.init(verbose=True, store='./logdir')
+
+# (Recomended) Option 1: Register all groq models.
 ell.models.groq.register() # use GROQ_API_KEY env var
 # ell.models.groq.register(api_key="gsk-") # 
 
@@ -15,7 +17,7 @@ def write_a_story(about : str):
 
 write_a_story("cats")
 
-# or use the client directly
+# Option 2: Use a client directly
 client = groq.Groq()
 
 @ell.simple(model="llama3-8b-8192", temperature=0.1, client=client)

src/ell/configurator.py

@@ -1,4 +1,4 @@
-from functools import wraps
+from functools import lru_cache, wraps
 from typing import Dict, Any, Optional, Tuple, Union, Type
 import openai
 import logging
@@ -133,7 +133,10 @@ def get_provider_for(self, client: Union[Type[Any], Any]) -> Optional[Provider]:
         """
 
         client_type = type(client) if not isinstance(client, type) else client
-        return self.providers.get(client_type)
+        for provider_type, provider in self.providers.items():
+            if issubclass(client_type, provider_type) or client_type == provider_type:
+                return provider
+        return None
 
 # Single* instance
 # XXX: Make a singleton

src/ell/lmp/complex.py

@@ -17,7 +17,6 @@ def complex(model: str, client: Optional[Any] = None, tools: Optional[List[Calla
     default_client_from_decorator = client
     default_model_from_decorator = model
     default_api_params_from_decorator = api_params
-
     def parameterized_lm_decorator(
         prompt: LMP,
     ) -> Callable[..., Union[List[Message], Message]]:
@@ -66,7 +65,7 @@ def model_call(
 
             if should_log: model_usage_logger_post_start(n)
             with model_usage_logger_post_intermediate(n) as _logger:
-                (result, final_api_params, metadata) = provider.call(ell_call, origin_id=_invocation_origin, logger=_logger)
+                (result, final_api_params, metadata) = provider.call(ell_call, origin_id=_invocation_origin, logger=_logger if should_log else None)
                 if isinstance(result, list) and len(result) == 1:
                     result = result[0]
 

src/ell/lmp/simple.py

@@ -11,7 +11,7 @@ def simple(model: str, client: Optional[Any] = None,  exempt_from_tracking=False
 
     def convert_multimodal_response_to_lstr(response):
         return [x.content[0].text for x in response] if isinstance(response, list) else response.content[0].text
-    return complex(model, client,  exempt_from_tracking, **api_params, post_callback=convert_multimodal_response_to_lstr)
+    return complex(model, client,  exempt_from_tracking=exempt_from_tracking, **api_params, post_callback=convert_multimodal_response_to_lstr)