Python: Complete durableagent package

[larohra]

Motivation and Context

Description

This PR adds the remainder of the changes needed for getting the durabletask package to work in agent-framework.

This pull request refactors and simplifies the Azure Functions agent integration, focusing on removing redundant models, improving agent session handling, and clarifying type usage. The changes enhance maintainability and consistency by delegating responsibilities to shared libraries and updating internal APIs.

Removal of redundant models and code simplification

• Deleted the entire AgentSessionId and DurableAgentThread implementation from agent_framework_azurefunctions/_models.py, delegating session ID management to the shared durable task library instead python/packages/azurefunctions/agent_framework_azurefunctions/_models.pyL1-L201
• Refactored imports in agent_framework_azurefunctions/_app.py and _orchestration.py to use session ID and agent types from the shared durable task library, eliminating local definitions and reducing duplication.

Agent proxy and executor improvements

• Updated the get_agent method in AgentFunctionApp to instantiate DurableAIAgent with an AzureFunctionsAgentExecutor, ensuring better separation of orchestration context and agent execution. Also clarified the return type to DurableAIAgent[AgentTask].

Session ID and entity management consistency

• Changed session ID handling to consistently use entity_name and key properties when constructing EntityId objects, rather than relying on a local conversion method. This affects both HTTP trigger and tool invocation flows.

API response formatting and payload handling

• Simplified response construction by extracting message text from agent responses and updating the success result payload to use the new format and field names, making API responses more consistent.

Durable orchestration response handling

• Updated orchestration code to use shared load_agent_response and ensure_response_format functions, improving reliability and centralizing response format enforcement.

Contribution Checklist

• The code builds clean without any errors or warnings
• The PR follows the Contribution Guidelines
• All unit tests pass, and I have added new tests where possible
• Is this a breaking change? If yes, add "[BREAKING]" prefix to the title of the PR.

[Copilot]

Pull request overview

This PR completes the durabletask package implementation by refactoring the Azure Functions agent integration to use shared durable task library components. The main changes remove redundant models from Azure Functions, consolidate session ID management, and improve the separation of concerns between orchestration context and agent execution.

Key Changes:

• Removes duplicate model implementations (AgentSessionId, DurableAgentThread) from Azure Functions package, delegating to shared durabletask library
• Adds comprehensive test coverage for the durabletask package (worker, client, executors, models, orchestration context)
• Introduces executor pattern with ClientAgentExecutor and OrchestrationAgentExecutor for different execution contexts
• Updates Azure Functions to use shared components and improves API consistency
• Adds new durabletask sample code demonstrating single agent, orchestration chaining, and multi-agent concurrency patterns

Reviewed changes

Copilot reviewed 47 out of 48 changed files in this pull request and generated no comments.

Show a summary per file

File	Description
python/packages/durabletask/agent_framework_durabletask/_models.py	Added AgentSessionId and DurableAgentThread with correlation_id as required field
python/packages/durabletask/agent_framework_durabletask/_client.py	New client wrapper for external interactions with durable agents
python/packages/durabletask/agent_framework_durabletask/_worker.py	New worker wrapper for agent registration as durable entities
python/packages/durabletask/agent_framework_durabletask/_executors.py	Executor implementations for client and orchestration contexts
python/packages/durabletask/agent_framework_durabletask/_shim.py	Agent protocol implementation delegating to executors
python/packages/durabletask/agent_framework_durabletask/_orchestration_context.py	Orchestration context wrapper for agent access
python/packages/durabletask/agent_framework_durabletask/_response_utils.py	Shared response handling utilities
python/packages/durabletask/tests/*.py	Comprehensive unit tests for all new components
python/packages/azurefunctions/agent_framework_azurefunctions/_orchestration.py	Refactored to use shared executor pattern and utilities
python/packages/azurefunctions/agent_framework_azurefunctions/_app.py	Updated to use shared models and improved session ID handling
python/packages/azurefunctions/agent_framework_azurefunctions/_models.py	Removed (all content moved to durabletask package)
python/samples/getting_started/durabletask/*	New samples demonstrating durabletask agent patterns
python/samples/getting_started/azure_functions/*.py	Minor fixes for type hints and removed redundant null checks

[Copilot]

Pull request overview

Copilot reviewed 47 out of 48 changed files in this pull request and generated no new comments.

[larohra]

@andystaples just to confirm, if a CompositeTask throws an exception, does the durable orchestrator handle it gracefully or do we need to do extra checks here?

Context: We're wrapping the task to call an entity in a CompositeTask so that we can strongly type the response back to AgentRunResponse to the caller.

[andystaples]

This throw will cause the entire orchestration to fail without hitting user code, preventing users from try/catching the yield.
Instead you need to set _is_complete and _exception like you are doing on line 71-77, but since you'll need to construct a TaskFailedError, you will probably need a change like this
microsoft/durabletask-python#99
Reviewing this with the Python folks from Durable

[larohra]

Given that the required changes in Durabletask are in a PR, I'm going to merge this PR and follow-up with the required changes once the new version of durabletask is ready in production.

[Copilot]

Pull request overview

Copilot reviewed 45 out of 46 changed files in this pull request and generated no new comments.

[moonbox3]

Python Test Coverage Report •

File	Stmts	Miss	Cover	Missing
packages/azurefunctions/agent_framework_azurefunctions
_app.py	368	82	77%	202–203, 208–209, 320–321, 429, 437–438, 458–460, 466–468, 474–476, 509–510, 570–571, 708, 711, 720–722, 724–726, 728, 730, 741, 743–746, 748, 750–751, 753, 760–761, 763–764, 766–767, 769, 773, 783–785, 787–788, 790–792, 799, 801–802, 804, 825, 830, 842, 917, 927, 934–936, 981, 995, 1006–1008, 1010–1013, 1038, 1045, 1047, 1050
_orchestration.py	50	4	92%	106–107, 111, 126
TOTAL	15971	2454	84%

Python Unit Test Overview

Tests	Skipped	Failures	Errors	Time
2615	148 💤	0 ❌	0 🔥	1m 2s ⏱️

[Copilot]

Pull request overview

Copilot reviewed 45 out of 46 changed files in this pull request and generated no new comments.