Python: fixes and release (#855)

* Fix lint errors
* Bump version number
* Add retry loop to integration tests
* Enable all integration tests: some tests will fail if OpenAI/Azure are
throttling.

Author: dluc

Diff for: python/pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "semantic-kernel"
-version = "0.2.6.dev"
+version = "0.2.7.dev"
 description = ""
 authors = ["Microsoft <[email protected]>"]
 readme = "pip/README.md"

Diff for: python/semantic_kernel/connectors/ai/hugging_face/services/hf_text_completion.py

@@ -3,7 +3,6 @@
 from logging import Logger
 from typing import Optional
 
-
 from semantic_kernel.connectors.ai.ai_exception import AIException
 from semantic_kernel.connectors.ai.complete_request_settings import (
     CompleteRequestSettings,
@@ -48,11 +47,13 @@ def __init__(
         self._log = log if log is not None else NullLogger()
 
         try:
-            import transformers
             import torch
+            import transformers
         except (ImportError, ModuleNotFoundError):
-            raise ImportError("Please ensure that torch and transformers are installed to use HuggingFaceTextCompletion")
-        
+            raise ImportError(
+                "Please ensure that torch and transformers are installed to use HuggingFaceTextCompletion"
+            )
+
         self.device = (
             "cuda:" + device if device >= 0 and torch.cuda.is_available() else "cpu"
         )
@@ -75,6 +76,7 @@ async def complete_async(
         """
         try:
             import transformers
+
             generation_config = transformers.GenerationConfig(
                 temperature=request_settings.temperature,
                 top_p=request_settings.top_p,

Diff for: python/semantic_kernel/connectors/ai/hugging_face/services/hf_text_embedding.py

@@ -2,6 +2,7 @@
 
 from logging import Logger
 from typing import List, Optional
+
 from numpy import array, ndarray
 
 from semantic_kernel.connectors.ai.ai_exception import AIException
@@ -37,11 +38,13 @@ def __init__(
         self._log = log if log is not None else NullLogger()
 
         try:
-            import torch
             import sentence_transformers
-        except (ImportError):
-            raise ImportError("Please ensure that torch and sentence-transformers are installed to use HuggingFaceTextEmbedding")
-        
+            import torch
+        except ImportError:
+            raise ImportError(
+                "Please ensure that torch and sentence-transformers are installed to use HuggingFaceTextEmbedding"
+            )
+
         self.device = (
             "cuda:" + device if device >= 0 and torch.cuda.is_available() else "cpu"
         )

Diff for: python/semantic_kernel/kernel.py

@@ -49,6 +49,7 @@
 
 T = TypeVar("T")
 
+
 class Kernel(KernelBase, KernelExtensions):
     _log: Logger
     _skill_collection: SkillCollectionBase

Diff for: python/tests/integration/completions/e2e_text_completion.py

@@ -1,5 +1,25 @@
+import logging
+import time
+
 import semantic_kernel as sk
 
+logging.basicConfig(level=logging.DEBUG)
+logger = logging.getLogger()
+
+
+async def retry(func, retries=15, delay=1):
+    for i in range(retries):
+        try:
+            result = str(await func())
+            if "Error" in result:
+                raise ValueError(result)
+            return result
+        except Exception as e:
+            logger.error(f"Retry {i + 1}: {e}")
+            if i == retries - 1:  # Last retry
+                raise
+            time.sleep(delay)
+
 
 async def summarize_function_test(kernel: sk.Kernel):
     # Define semantic function using SK prompt template language
@@ -33,68 +53,80 @@ async def summarize_function_test(kernel: sk.Kernel):
     print()
 
     # Summarize input string and print
-    summary = await kernel.run_async(tldr_function, input_str=text_to_summarize)
-
+    summary = await retry(
+        lambda: kernel.run_async(tldr_function, input_str=text_to_summarize)
+    )
     output = str(summary).strip()
     print(f"Summary using input string: '{output}'")
-    assert len(output.split(" ")) == 5
+    assert "humans" in output or "Humans" in output or "preserve" in output
+    assert len(output) < 100
 
     # Summarize input as context variable and print
     context_vars = sk.ContextVariables(text_to_summarize)
-    summary = await kernel.run_async(tldr_function, input_vars=context_vars)
-
+    summary = await retry(
+        lambda: kernel.run_async(tldr_function, input_vars=context_vars)
+    )
     output = str(summary).strip()
     print(f"Summary using context variables: '{output}'")
-    assert len(output.split(" ")) == 5
+    assert "humans" in output or "Humans" in output or "preserve" in output
+    assert len(output) < 100
 
     # Summarize input context and print
     context = kernel.create_new_context()
     context["input"] = text_to_summarize
-    summary = await kernel.run_async(tldr_function, input_context=context)
-
+    summary = await retry(
+        lambda: kernel.run_async(tldr_function, input_context=context)
+    )
     output = str(summary).strip()
     print(f"Summary using input context: '{output}'")
-    assert len(output.split(" ")) == 5
+    assert "humans" in output or "Humans" in output or "preserve" in output
+    assert len(output) < 100
 
     # Summarize input context with additional variables and print
     context = kernel.create_new_context()
     context["input"] = text_to_summarize
     context_vars = sk.ContextVariables("4) All birds are robots.")
-    summary = await kernel.run_async(
-        tldr_function, input_context=context, input_vars=context_vars
+    summary = await retry(
+        lambda: kernel.run_async(
+            tldr_function, input_context=context, input_vars=context_vars
+        )
     )
-
     output = str(summary).strip()
     print(f"Summary using context and additional variables: '{output}'")
-    assert len(output.split(" ")) == 5
+    assert "humans" in output or "Humans" in output or "preserve" in output
+    assert len(output) < 100
 
     # Summarize input context with additional input string and print
     context = kernel.create_new_context()
     context["input"] = text_to_summarize
-    summary = await kernel.run_async(
-        tldr_function, input_context=context, input_str="4) All birds are robots."
+    summary = await retry(
+        lambda: kernel.run_async(
+            tldr_function, input_context=context, input_str="4) All birds are robots."
+        )
     )
-
     output = str(summary).strip()
     print(f"Summary using context and additional string: '{output}'")
-    assert len(output.split(" ")) == 5
+    assert "humans" in output or "Humans" in output or "preserve" in output
+    assert len(output) < 100
 
     # Summarize input context with additional variables and string and print
     context = kernel.create_new_context()
     context["input"] = text_to_summarize
     context_vars = sk.ContextVariables(variables={"input2": "4) All birds are robots."})
-    summary = await kernel.run_async(
-        tldr_function,
-        input_context=context,
-        input_vars=context_vars,
-        input_str="new text",
+    summary = await retry(
+        lambda: kernel.run_async(
+            tldr_function,
+            input_context=context,
+            input_vars=context_vars,
+            input_str="new text",
+        )
     )
-
     output = str(summary).strip()
     print(
         f"Summary using context, additional variables, and additional string: '{output}'"
     )
-    assert len(output.split(" ")) == 5
+    assert "humans" in output or "Humans" in output or "preserve" in output
+    assert len(output) < 100
 
 
 async def simple_summarization(kernel: sk.Kernel):

Diff for: python/tests/integration/completions/test_azure_oai_chat_service.py

@@ -11,10 +11,6 @@
 
 
 @pytest.mark.asyncio
-@pytest.mark.xfail(
-    raises=AssertionError,
-    reason="Azure OpenAI may throttle requests, preventing this test from passing",
-)
 async def test_azure_chat_completion_with_skills():
     kernel = sk.Kernel()
 
@@ -27,6 +23,10 @@ async def test_azure_chat_completion_with_skills():
         deployment_name, api_key, endpoint = sk.azure_openai_settings_from_dot_env()
         deployment_name = "gpt-4"
 
+    print("* Service: Azure OpenAI Chat Completion")
+    print(f"* Endpoint: {endpoint}")
+    print(f"* Deployment: {deployment_name}")
+
     # Configure LLM service
     kernel.add_chat_service(
         "chat_completion",

Diff for: python/tests/integration/completions/test_azure_oai_text_service.py

@@ -23,6 +23,10 @@ async def test_azure_text_completion_with_skills():
         deployment_name, api_key, endpoint = sk.azure_openai_settings_from_dot_env()
         deployment_name = "text-davinci-003"
 
+    print("* Service: Azure OpenAI Text Completion")
+    print(f"* Endpoint: {endpoint}")
+    print(f"* Deployment: {deployment_name}")
+
     # Configure LLM service
     kernel.add_text_completion_service(
         "text_completion",

Diff for: python/tests/integration/completions/test_oai_chat_service.py

@@ -11,10 +11,6 @@
 
 
 @pytest.mark.asyncio
-@pytest.mark.xfail(
-    raises=AssertionError,
-    reason="OpenAI may throttle requests, preventing this test from passing",
-)
 async def test_oai_chat_service_with_skills():
     kernel = sk.Kernel()
 
@@ -25,6 +21,10 @@ async def test_oai_chat_service_with_skills():
         # Load credentials from .env file
         api_key, org_id = sk.openai_settings_from_dot_env()
 
+    print("* Service: OpenAI Chat Completion")
+    print("* Endpoint: OpenAI")
+    print("* Model: gpt-3.5-turbo")
+
     kernel.add_chat_service(
         "chat-gpt", sk_oai.OpenAIChatCompletion("gpt-3.5-turbo", api_key, org_id)
     )

Diff for: python/tests/integration/completions/test_oai_text_service.py

@@ -11,10 +11,6 @@
 
 
 @pytest.mark.asyncio
-@pytest.mark.xfail(
-    raises=AssertionError,
-    reason="OpenAI may throttle requests, preventing this test from passing",
-)
 async def test_oai_text_completion_with_skills():
     kernel = sk.Kernel()
 
@@ -25,6 +21,10 @@ async def test_oai_text_completion_with_skills():
         # Load credentials from .env file
         api_key, org_id = sk.openai_settings_from_dot_env()
 
+    print("* Service: OpenAI Text Completion")
+    print("* Endpoint: OpenAI")
+    print("* Model: text-davinci-003")
+
     kernel.add_chat_service(
         "davinci-003", sk_oai.OpenAITextCompletion("text-davinci-003", api_key, org_id)
     )

Diff for: python/tests/integration/embeddings/test_oai_embedding_service.py

@@ -11,7 +11,6 @@
 
 
 @pytest.mark.asyncio
-# @pytest.mark.xfail(raises=AssertionError, reason="OpenAI may throttle requests, preventing this test from passing")
 async def test_oai_embedding_service_with_memories():
     kernel = sk.Kernel()