Python: Document generator agent framework demo

python/samples/demos/document_generator/GENERATED_DOCUMENT.md

@@ -0,0 +1,58 @@
+### Understanding Semantic Kernel AI Connectors
+
+AI Connectors in Semantic Kernel are components that facilitate communication between the Kernel's core functionalities and various AI services. They abstract the intricate details of service-specific protocols, allowing developers to seamlessly interact with AI services for tasks like text generation, chat interactions, and more.
+
+### Using AI Connectors in Semantic Kernel
+
+Developers utilize AI connectors to connect their applications to different AI services efficiently. The connectors manage the requests and responses, providing a streamlined way to leverage the power of these AI services without needing to handle the specific communication protocols each service requires.
+
+### Creating Custom AI Connectors in Semantic Kernel
+
+To create a custom AI connector in Semantic Kernel, one must extend the base classes provided, such as `ChatCompletionClientBase` and `AIServiceClientBase`. Below is a guide and example for implementing a mock AI connector:
+
+#### Step-by-Step Walkthrough
+
+1. **Understand the Base Classes**: The foundational classes `ChatCompletionClientBase` and `AIServiceClientBase` provide necessary methods and structures for creating chat-based AI connectors.
+
+2. **Implementing the Connector**: Here's a mock implementation example illustrating how to implement a connector without real service dependencies, ensuring compatibility with Pydantic's expectations within the framework:
+
+```python
+from semantic_kernel.connectors.ai.chat_completion_client_base import ChatCompletionClientBase
+
+class MockAIChatCompletionService(ChatCompletionClientBase):
+    def __init__(self, ai_model_id: str):
+        super().__init__(ai_model_id=ai_model_id)
+
+    async def _inner_get_chat_message_contents(self, chat_history, settings):
+        # Mock implementation: returns dummy chat message content for demonstration.
+        return [{"role": "assistant", "content": "Mock response based on your history."}]
+
+    def service_url(self):
+        return "http://mock-ai-service.com"
+```
+
+### Usage Example
+
+The following example demonstrates how to integrate and use the `MockAIChatCompletionService` in an application:
+
+```python
+import asyncio
+from semantic_kernel.contents.chat_history import ChatHistory
+from semantic_kernel.connectors.ai.prompt_execution_settings import PromptExecutionSettings
+
+async def main():
+    chat_history = ChatHistory(messages=[{"role": "user", "content": "Hello"}])
+    settings = PromptExecutionSettings(model="mock-model")
+
+    service = MockAIChatCompletionService(ai_model_id="mock-model")
+
+    response = await service.get_chat_message_contents(chat_history, settings)
+    print(response)
+
+# Run the main function
+asyncio.run(main())
+```
+
+### Conclusion
+
+By following the revised guide and understanding the base class functionalities, developers can effectively create custom connectors within Semantic Kernel. This structured approach enhances integration with various AI services while ensuring alignment with the framework's architectural expectations. Custom connectors offer flexibility, allowing developers to adjust implementations to meet specific service needs, such as additional logging, authentication, or modifications tailored to specific protocols. This guide provides a strong foundation upon which more complex and service-specific extensions can be built, promoting robust and scalable AI service integration.

python/samples/demos/document_generator/README.md

@@ -0,0 +1,100 @@
+# Document Generator
+
+This sample app demostrates how to create technical documents for a codebase using AI. More specifically, it uses the agent framework offered by **Semantic Kernel** to ochestrate multiple agents to create a technical document.
+
+This sample app also provides telemtry to monitor the agents, making it easier to oberve the inner workings of the agents.
+
+To learn more about agents, please refer to this introduction [video](https://learn.microsoft.com/en-us/shows/generative-ai-for-beginners/ai-agents-generative-ai-for-beginners).
+To learn more about the Semantic Kernel Agent Framework, please refer to the [Semantic Kernel documentation](https://learn.microsoft.com/en-us/semantic-kernel/frameworks/agent/agent-architecture?pivots=programming-language-python).
+
+> Note: This sample app cannot guarantee to generate a perfect technical document each time due to the stochastic nature of the AI model. Please a version of the document generated by the app in [GENERATED_DOCUMENT.md](GENERATED_DOCUMENT.md).
+
+## Desgin
+
+### Tools/PLugins
+
+- **Code Execution Plugin**: This plugin offers a sandbox environment to execute Python snippets. It returns the output of the program or errors if any.
+- **Repository File Plugin**: This plugin allows the AI to retrieve files from the Semantic Kernel repository.
+- **User Input Plugin**: This plugin allows the AI to present content to the user and receive feedback.
+
+### Agents
+
+- **Content Creation Agent**: This agent is responsible for creating the content of the document. This agent has access to the **Repository File Plugin** to read source files it deems necessary for reference.
+- **Code Validation Agent**: This agent is responsible for validating the code snippets in the document. This agent has access to the **Code Execution Plugin** to execute the code snippets.
+- **User Agent**: This agent is responsible for interacting with the user. This agent has access to the **User Input Plugin** to present content to the user and receive feedback.
+
+### Agent Selection Strategy
+
+### Termination Strategy
+
+## Prerequisites
+
+1. Azure OpenAI
+2. Azure Application Insights
+
+## Additional packages
+
+- `AICodeSandbox` - for executing AI generated code in a sandbox environment
+
+    ```bash
+    pip install AICodeSandbox
+    ```
+
+    > You must also have `docker` installed and running on your machine. Simply run docker. Images will be pulled during runtime if not already present. Containers will be created and destroyed during code execution.
+
+## Running the app
+
+### Step 1: Set up the environment
+
+Make sure you have the following environment variables set:
+
+```env
+OPENAI_CHAT_MODEL_ID=<model-id>
+OPENAI_API_KEY=<your-key>
+```
+
+### Step 2: Run the app
+
+```bash
+python ./main.py
+```
+
+Expected output:
+
+```bash
+==== ContentCreationAgent just responded ====
+==== CodeValidationAgent just responded ====
+==== ContentCreationAgent just responded ====
+...
+```
+
+## Customization
+
+Since this is a sample app that demonstrates the creation of a technical document on Semantic Kernel AI connectors, you can customize the app to suit your needs. You can try different tasks, add more agents, tune existing agents, change the agent selection strategy, or modify the termination strategy.
+
+- To try a different task, modify the `TASK` prompt in `main.py`.
+- To add more agents, create a new agent under `agents/` and add it to the `agents` list in `main.py`.
+- To tune existing agents, modify the `INSTRUCTION` prompt in the agent's source code.
+- To change the agent selection strategy, modify `custom_selection_strategy.py`.
+- To change the termination strategy, modify `custom_termination_strategy.py`.
+
+## Optional: Monitoring the agents
+
+When you see the final document generated by the app, what you see is actually the creation of multiple agents working together. You may wonder, how did the agents work together to create the document? What was the sequence of actions taken by the agents? How did the agents interact with each other? To answer these questions, you need to **observe** the agents.
+
+Semantic Kernel by default instruments all the LLM calls. However, for agents there is no default instrumentation. This sample app shows how one can extend the Semantic Kernel agent to add instrumentation.
+
+> There are currently no standards on what information needs to be captured for agents as the concept of agents is still relatively new. At the time of writing, the Semantic Convention for agents is still in the draft stage: <https://github.com/open-telemetry/semantic-conventions/issues/1732>
+
+To monitor the agents, set the following environment variables:
+
+```env
+AZURE_APP_INSIGHTS_CONNECTION_STRING=<your-connection-string>
+
+SEMANTICKERNEL_EXPERIMENTAL_GENAI_ENABLE_OTEL_DIAGNOSTICS=true
+SEMANTICKERNEL_EXPERIMENTAL_GENAI_ENABLE_OTEL_DIAGNOSTICS_SENSITIVE=true
+```
+
+Follow this guide to inspect the telemetry data: <https://learn.microsoft.com/en-us/semantic-kernel/concepts/enterprise-readiness/observability/telemetry-with-app-insights?tabs=Powershell&pivots=programming-language-python#inspect-telemetry-data>
+
+Or follow this guide to visualize the telemetry data on Azure AI Foundry: <https://learn.microsoft.com/en-us/semantic-kernel/concepts/enterprise-readiness/observability/telemetry-with-azure-ai-foundry-tracing#visualize-traces-on-azure-ai-foundry-tracing-ui-1>

python/samples/demos/document_generator/agents/code_validation_agent.py

@@ -0,0 +1,57 @@
+# Copyright (c) Microsoft. All rights reserved.
+
+import sys
+from collections.abc import AsyncIterable
+
+if sys.version_info >= (3, 12):
+    from typing import override  # pragma: no cover
+else:
+    from typing_extensions import override  # pragma: no cover
+
+from samples.demos.document_generator.agents.custom_agent_base import CustomAgentBase
+from samples.demos.document_generator.plugins.code_execution_plugin import CodeExecutionPlugin
+from semantic_kernel.connectors.ai.function_choice_behavior import FunctionChoiceBehavior
+from semantic_kernel.contents.chat_history import ChatHistory
+from semantic_kernel.contents.chat_message_content import ChatMessageContent
+
+INSTRUCTION = """
+You are a code validation agent in a collaborative document creation chat.
+
+Your task is to validate Python code in the latest document draft and summarize any errors.
+Follow the instructions in the document to assemble the code snippets into a single Python script.
+If the snippets in the document are from multiple scripts, you need to modify them to work together as a single script.
+Execute the code to validate it. If there are errors, summarize the error messages.
+
+Do not try to fix the errors.
+"""
+
+DESCRIPTION = """
+Select me to validate the Python code in the latest document draft.
+"""
+
+
+class CodeValidationAgent(CustomAgentBase):
+    def __init__(self):
+        kernel = self._create_kernel()
+        kernel.add_plugin(plugin=CodeExecutionPlugin(), plugin_name="CodeExecutionPlugin")
+
+        settings = kernel.get_prompt_execution_settings_from_service_id(service_id=CustomAgentBase.SERVICE_ID)
+        settings.function_choice_behavior = FunctionChoiceBehavior.Auto(maximum_auto_invoke_attempts=1)
+
+        super().__init__(
+            kernel=kernel,
+            execution_settings=settings,
+            name="CodeValidationAgent",
+            instructions=INSTRUCTION.strip(),
+            description=DESCRIPTION.strip(),
+        )
+
+    @override
+    async def invoke(self, history: ChatHistory) -> AsyncIterable[ChatMessageContent]:
+        cloned_history = history.model_copy(deep=True)
+        cloned_history.add_user_message_str(
+            "Now validate the Python code in the latest document draft and summarize any errors."
+        )
+
+        async for response_message in super().invoke(cloned_history):
+            yield response_message

python/samples/demos/document_generator/agents/content_creation_agent.py

@@ -0,0 +1,54 @@
+# Copyright (c) Microsoft. All rights reserved.
+
+import sys
+from collections.abc import AsyncIterable
+
+if sys.version_info >= (3, 12):
+    from typing import override  # pragma: no cover
+else:
+    from typing_extensions import override  # pragma: no cover
+
+from samples.demos.document_generator.agents.custom_agent_base import CustomAgentBase
+from samples.demos.document_generator.plugins.repo_file_plugin import RepoFilePlugin
+from semantic_kernel.connectors.ai.function_choice_behavior import FunctionChoiceBehavior
+from semantic_kernel.contents.chat_history import ChatHistory
+from semantic_kernel.contents.chat_message_content import ChatMessageContent
+
+INSTRUCTION = """
+You are part of a chat with multiple agents focused on creating technical content.
+
+Your task is to generate informative and engaging technical content,
+including code snippets to explain concepts or demonstrate features.
+Incorporate feedback by providing the updated full content with changes.
+"""
+
+DESCRIPTION = """
+Select me to generate new content or to revise existing content.
+"""
+
+
+class ContentCreationAgent(CustomAgentBase):
+    def __init__(self):
+        kernel = self._create_kernel()
+        kernel.add_plugin(plugin=RepoFilePlugin(), plugin_name="RepoFilePlugin")
+
+        settings = kernel.get_prompt_execution_settings_from_service_id(service_id=CustomAgentBase.SERVICE_ID)
+        settings.function_choice_behavior = FunctionChoiceBehavior.Auto()
+
+        super().__init__(
+            kernel=kernel,
+            execution_settings=settings,
+            name="ContentCreationAgent",
+            instructions=INSTRUCTION.strip(),
+            description=DESCRIPTION.strip(),
+        )
+
+    @override
+    async def invoke(self, history: ChatHistory) -> AsyncIterable[ChatMessageContent]:
+        cloned_history = history.model_copy(deep=True)
+        cloned_history.add_user_message_str(
+            "Now generate new content or revise existing content to incorporate feedback."
+        )
+
+        async for response_message in super().invoke(cloned_history):
+            yield response_message

python/samples/demos/document_generator/agents/custom_agent_base.py

@@ -0,0 +1,56 @@
+# Copyright (c) Microsoft. All rights reserved.
+
+import sys
+from abc import ABC
+from collections.abc import AsyncIterable
+from typing import ClassVar
+
+from opentelemetry import trace
+
+from semantic_kernel.connectors.ai.open_ai.services.open_ai_chat_completion import OpenAIChatCompletion
+
+if sys.version_info >= (3, 12):
+    from typing import override  # pragma: no cover
+else:
+    from typing_extensions import override  # pragma: no cover
+
+from semantic_kernel.agents.chat_completion.chat_completion_agent import ChatCompletionAgent
+from semantic_kernel.contents.chat_history import ChatHistory
+from semantic_kernel.contents.chat_message_content import ChatMessageContent
+from semantic_kernel.kernel import Kernel
+
+
+class CustomAgentBase(ChatCompletionAgent, ABC):
+    SERVICE_ID: ClassVar[str] = "chat_completion"
+
+    def _create_kernel(self) -> Kernel:
+        kernel = Kernel()
+        kernel.add_service(OpenAIChatCompletion(service_id=self.SERVICE_ID))
+
+        return kernel
+
+    @override
+    async def invoke(self, history: ChatHistory) -> AsyncIterable[ChatMessageContent]:
+        # Since the history contains internal messages from other agents,
+        # we will do our best to filter out those. Unfortunately, there will
+        # be a side effect of losing the context of the conversation internal
+        # to the agent when the conversation is handed back to the agent, i.e.
+        # previous function call results.
+        filtered_chat_history = ChatHistory()
+        for message in history:
+            content = message.content
+            # We don't want to add messages whose text content is empty.
+            # Those messages are likely messages from function calls and function results.
+            if content:
+                filtered_chat_history.add_message(message)
+
+        tracer = trace.get_tracer(__name__)
+        response_messages: list[ChatMessageContent] = []
+        with tracer.start_as_current_span(self.name):
+            # Cache the messages within the span such that subsequent spans
+            # that process the message stream don't become children of this span
+            async for response_message in super().invoke(filtered_chat_history):
+                response_messages.append(response_message)
+
+        for response_message in response_messages:
+            yield response_message

python/samples/demos/document_generator/agents/user_agent.py

@@ -0,0 +1,55 @@
+# Copyright (c) Microsoft. All rights reserved.
+
+import sys
+from collections.abc import AsyncIterable
+
+if sys.version_info >= (3, 12):
+    from typing import override  # pragma: no cover
+else:
+    from typing_extensions import override  # pragma: no cover
+
+from samples.demos.document_generator.agents.custom_agent_base import CustomAgentBase
+from samples.demos.document_generator.plugins.user_plugin import UserPlugin
+from semantic_kernel.connectors.ai.function_choice_behavior import FunctionChoiceBehavior
+from semantic_kernel.contents.chat_history import ChatHistory
+from semantic_kernel.contents.chat_message_content import ChatMessageContent
+
+INSTRUCTION = """
+You are part of a chat with multiple agents working on a document.
+
+Your task is to summarize the user's feedback on the latest draft from the author agent.
+Present the draft to the user and summarize their feedback.
+
+Do not try to address the user's feedback in this chat.
+"""
+
+DESCRIPTION = """
+Select me if you want to ask the user to review the latest draft for publication.
+"""
+
+
+class UserAgent(CustomAgentBase):
+    def __init__(self):
+        kernel = self._create_kernel()
+        kernel.add_plugin(plugin=UserPlugin(), plugin_name="UserPlugin")
+
+        settings = kernel.get_prompt_execution_settings_from_service_id(service_id=CustomAgentBase.SERVICE_ID)
+        settings.function_choice_behavior = FunctionChoiceBehavior.Auto(maximum_auto_invoke_attempts=1)
+
+        super().__init__(
+            kernel=kernel,
+            execution_settings=settings,
+            name="UserAgent",
+            instructions=INSTRUCTION.strip(),
+            description=DESCRIPTION.strip(),
+        )
+
+    @override
+    async def invoke(self, history: ChatHistory) -> AsyncIterable[ChatMessageContent]:
+        cloned_history = history.model_copy(deep=True)
+        cloned_history.add_user_message_str(
+            "Now present the latest draft to the user for feedback and summarize their feedback."
+        )
+
+        async for response_message in super().invoke(cloned_history):
+            yield response_message