Python: Add declarative workflow runtime

[moonbox3]

Motivation and Context

Adds a graph-based runtime for executing declarative YAML workflows, enabling multi-agent orchestration without writing Python code.

Features

• Workflow Graph Builder: Transforms YAML workflow definitions into executable directed graphs
• Action Executors: SetValue, SendActivity, Foreach, If/Switch, RepeatUntil, GotoAction, InvokeAzureAgent, and more
• Expression Evaluation: PowerFx-style expressions with namespace support (Local., System., Workflow.)
• Human-in-the-Loop: Question, WaitForInput, and RequestExternalInput actions for interactive workflows
• Schema Validation: Required field checking, duplicate ID detection, and self-referencing goto detection

Samples

• simple_workflow: Basic variables, conditionals, and loops
• conditional_workflow: Nested branching based on user input
• human_in_loop: Interactive workflows requesting user input
• function_tools: Invoking Python functions from workflows
• customer_support, marketing, deep_research, student_teacher: Multi-agent patterns

Description

• Closes Python runtime for Declarative Workflows (Epic) #2727

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.

[markwallace-microsoft]

Python Test Coverage Report •

File	Stmts	Miss	Cover	Missing
packages/core/agent_framework/_workflows
_edge.py	233	31	86%	36, 58–59, 61, 142, 195, 213–214, 216, 464–465, 470, 474–479, 481–483, 601, 715, 722–724, 771, 799, 825, 879, 893
_edge_runner.py	168	16	90%	53, 58, 71, 117, 142–143, 147, 154, 203, 214, 219, 241, 367–368, 372, 413
_workflow_builder.py	278	36	87%	259, 600, 699, 706–707, 808, 811, 816, 818, 825, 828–832, 834, 896, 971, 974, 1034–1035, 1178, 1192–1199, 1201, 1204, 1206–1208, 1216
packages/core/agent_framework/openai
_chat_client.py	233	43	81%	110–111, 115, 133, 143–144, 146, 157, 171, 173–176, 185, 197, 233, 264, 269–283, 290–292, 367, 374, 376, 378, 398–399, 405, 421
TOTAL	16609	2563	84%

Python Unit Test Overview

Tests	Skipped	Failures	Errors	Time
2741	130 💤	0 ❌	0 🔥	1m 1s ⏱️

[Copilot]

Copilot encountered an error and was unable to review this pull request. You can try again by re-requesting a review.

[eavanvalkenburg]

can't we tighten this? AgentProtocol and AgentExecutor?

[eavanvalkenburg]

I think we could just add a create_agent_from_dict method to the agent factory, so that we don't have to yaml dump first?

[eavanvalkenburg]

Suggested change

	from agent_framework_declarative import WorkflowFactory
	from agent_framework.declarative import WorkflowFactory