Python: Fix tool normalization and provider sample consolidation

[eavanvalkenburg]

Summary

This PR addresses tool-normalization regressions and consolidates provider samples.

Core/runtime fixes

• Restores callable/single-tool normalization so direct callable and single-tool inputs work consistently.
• Changes tool-choice behavior so unset values remain unset (providers use service defaults), while explicit values are preserved.

Sample consolidation and provider coverage

• Consolidates duplicate samples/02-agents/chat_client/* built-in samples into built_in_chat_clients.py.
• Expands built-in client coverage to include OpenAI, Azure OpenAI (including Foundry project endpoint), Anthropic, Ollama, and Bedrock.
• Keeps custom_chat_client.py and adds a local custom options type in that sample.

Bedrock promotion and provider-sample move

• Adds Bedrock to agent-framework-core[all].
• Adds Bedrock lazy import surface under agent_framework.amazon.
• Moves Bedrock sample content into samples/02-agents/providers/amazon/ and updates related docs.

Context-provider sample updates

• Updates context provider samples to use AzureOpenAIResponsesClient with Foundry project_endpoint + deployment_name setup where applicable.

Fixes #3577
Fixes #3946
Fixes #3945
Fixes #3936

[markwallace-microsoft]

Python Test Coverage Report •

File	Stmts	Miss	Cover	Missing
packages/anthropic/agent_framework_anthropic
_chat_client.py	366	149	59%	422, 451, 483, 485, 500, 522–525, 534, 536, 573–577, 579, 581–582, 584, 589–590, 592, 625–626, 635, 637–638, 643, 660–661, 705, 724, 730, 739, 741, 745–746, 789–791, 793, 806–807, 814–816, 820–822, 826–829, 840, 842, 864, 874, 896–902, 909–910, 918–919, 927–930, 937–938, 944–945, 951–952, 958, 966–968, 972, 979–980, 986–987, 993–994, 1000, 1008–1011, 1018–1019, 1038, 1045–1046, 1065, 1087, 1089, 1098–1099, 1105, 1127–1128, 1134–1135, 1144–1154, 1161–1167, 1174–1180, 1187–1196, 1203–1206
packages/azure-ai/agent_framework_azure_ai
_agent_provider.py	122	2	98%	245, 253
_chat_client.py	478	75	84%	388–389, 391, 574, 579–580, 582–583, 586, 589, 591, 596, 857–858, 860, 863, 866, 869–874, 877, 879, 887, 899–901, 905, 908–909, 917–920, 930, 938–941, 943–944, 946–947, 954, 962–963, 971–972, 977–978, 982–989, 994, 997, 1005, 1011, 1019–1021, 1024, 1046–1047, 1180, 1208, 1223, 1339, 1461
_client.py	275	37	86%	379, 381, 424, 432–444, 457, 517, 532–537, 580, 615, 617, 650, 858, 861, 864–865, 867–870, 913
_project_provider.py	117	4	96%	209, 303, 343, 376
packages/core/agent_framework
_agents.py	310	39	87%	476, 841, 877, 954–957, 1014–1016, 1130, 1143, 1156, 1162, 1198, 1200, 1209–1214, 1219, 1221, 1227–1228, 1235, 1237–1238, 1246–1247, 1250–1252, 1260–1261, 1263, 1268, 1270
_clients.py	71	3	95%	295, 488, 490
_tools.py	819	95	88%	165–166, 303, 305, 323–325, 332, 350, 364, 371, 378, 394, 396, 403, 440, 465, 469, 486–488, 535–537, 600, 622, 680–692, 727, 738–749, 771–773, 778, 782, 796–798, 837, 906, 916, 926, 982, 1013, 1032, 1294, 1351, 1371, 1442–1446, 1568, 1572, 1622, 1624, 1640, 1642, 1707, 1734, 1791, 1859, 2042–2043, 2070, 2078, 2091, 2101–2102, 2137, 2193, 2225
_types.py	998	92	90%	49, 58–59, 113, 118, 137, 139, 143, 147, 149, 151, 153, 171, 175, 201, 223, 228, 233, 237, 263, 267, 613–614, 985, 1047, 1064, 1082, 1087, 1105, 1115, 1132–1133, 1135, 1153–1154, 1156, 1163–1164, 1166, 1201, 1212–1213, 1215, 1253, 1480, 1532, 1623–1628, 1650, 1655, 1821, 1833, 2076, 2085, 2106, 2201, 2426, 2633, 2703, 2715, 2722, 2733, 2936–2938, 2941–2943, 2947, 2952, 2956, 3040–3042, 3071, 3125, 3129–3131, 3133, 3144–3145, 3148–3152, 3158
exceptions.py	50	0	100%
observability.py	617	83	86%	346, 348–350, 353–355, 360–361, 367–368, 374–375, 382, 384–386, 389–391, 396–397, 403–404, 410–411, 418, 687, 690, 698–699, 702–705, 707, 710–712, 715–716, 744, 746, 757–759, 761–763, 767, 775, 876, 878, 1027, 1029, 1033–1038, 1040, 1043–1047, 1049, 1158–1159, 1161, 1218–1219, 1354, 1408–1409, 1525–1527, 1586, 1756, 1910, 1912
packages/core/agent_framework/openai
_assistant_provider.py	107	9	91%	163, 289, 355, 470–475
_assistants_client.py	273	34	87%	408, 410, 412, 415, 419–420, 423, 426, 431–432, 434, 437–439, 444, 455, 480, 482, 484, 486, 488, 493, 496, 499, 503, 514, 599, 684, 713, 750–753, 823
_chat_client.py	269	23	91%	202, 232–233, 237, 355, 362, 443–450, 452–455, 465, 550, 585, 601, 641
_responses_client.py	613	79	87%	291–294, 298–299, 302–303, 309–310, 315, 328–334, 355, 363, 386, 549, 552, 607, 611, 613, 615, 617, 693, 703, 708, 751, 829, 846, 859, 915, 1008, 1013, 1017–1019, 1023–1024, 1047, 1116, 1138–1139, 1154–1155, 1173–1174, 1305–1306, 1322, 1324, 1403–1411, 1506, 1561, 1576, 1612–1613, 1615–1617, 1631–1633, 1643–1644, 1650, 1665
TOTAL	20563	3238	84%

Python Unit Test Overview

Tests	Skipped	Failures	Errors	Time
3955	225 💤	0 ❌	0 🔥	1m 9s ⏱️

[Copilot]

Pull request overview

This PR restores callable/single-tool normalization and unset tool-choice handling, consolidates chat client samples, migrates Bedrock to the agent_framework.amazon namespace, and updates context provider samples to use AzureOpenAIResponsesClient with Foundry project endpoint setup.

Changes:

• Restored tool normalization to accept single tools, callables, and sequences via unified normalize_tools() function, fixing issues #3936, #3946, and #3577
• Changed validate_tool_mode() to return None when tool_choice is unset (instead of {'mode': 'none'}), allowing clients to omit the parameter
• Migrated Bedrock to lazy-import agent_framework.amazon namespace, promoted in core[all], and relocated sample to providers/amazon/
• Consolidated chat client samples into single built_in_chat_clients.py with factory pattern supporting OpenAI, Azure, Anthropic, Ollama, and Bedrock
• Updated all context provider samples (Redis, Mem0, Azure AI Search, simple) to use AzureOpenAIResponsesClient with Foundry project endpoints

Reviewed changes

Copilot reviewed 45 out of 47 changed files in this pull request and generated 2 comments.

Show a summary per file

File	Description
python/packages/core/agent_framework/_tools.py	Added ToolTypes and normalize_tools() to handle single tools, callables, and sequences uniformly; supports tools passed via kwargs
python/packages/core/agent_framework/_types.py	Updated validate_tool_mode() to return None when unset; updated tool type annotations to use ToolTypes
python/packages/core/agent_framework/_agents.py	Updated Agent tool parameter types to use ToolTypes; simplified tool normalization logic
python/packages/core/agent_framework/_clients.py	Updated as_agent() tool parameter types to use ToolTypes
python/packages/core/agent_framework/openai/_chat_client.py	Updated _prepare_tools_for_openai() to use normalize_tools()
python/packages/core/agent_framework/openai/_assistants_client.py	Updated to use normalize_tools(); added null check for tool_mode before setting tool_choice
python/packages/core/agent_framework/openai/_assistant_provider.py	Updated tool type annotations to use ToolTypes; filter out MCP tools before passing to assistant API
python/packages/core/agent_framework/amazon/__init__.py	New lazy-import module for Bedrock package via agent_framework.amazon namespace
python/packages/core/pyproject.toml	Added agent-framework-bedrock to all extra
python/uv.lock	Promoted Bedrock in dependency resolution
python/packages/bedrock/agent_framework_bedrock/_chat_client.py	Updated import path in docstring to agent_framework.amazon
python/packages/bedrock/README.md	Updated sample reference to new location under providers/amazon/
python/packages/bedrock/AGENTS.md	Updated import path examples to agent_framework.amazon
python/packages/bedrock/samples/bedrock_sample.py	Deleted (moved to samples/02-agents/providers/amazon/)
python/samples/02-agents/providers/amazon/bedrock_chat_client.py	New consolidated Bedrock sample with tool usage
python/samples/02-agents/providers/amazon/README.md	New README documenting Bedrock sample and environment variables
python/samples/02-agents/providers/README.md	Added Bedrock provider entry to overview
python/samples/02-agents/chat_client/built_in_chat_clients.py	New consolidated sample supporting 11 client types via factory pattern
python/samples/02-agents/chat_client/README.md	Updated to document consolidated sample approach
python/samples/02-agents/chat_client/*.py (deleted)	Removed individual client samples (openai_, azure_, azure_ai_*) in favor of consolidated approach
python/samples/02-agents/context_providers/simple_context_provider.py	Updated to use AzureOpenAIResponsesClient with Foundry endpoints; improved state management
python/samples/02-agents/context_providers/redis/*.py	Updated all Redis samples to use AzureOpenAIResponsesClient and added source_id parameter
python/samples/02-agents/context_providers/redis/README.md	Updated environment variable documentation for Foundry setup
python/samples/02-agents/context_providers/mem0/*.py	Added source_id parameter to all Mem0ContextProvider instances
python/samples/02-agents/context_providers/azure_ai_search/*.py	Added source_id parameter to AzureAISearchContextProvider (fixes #3945)
python/samples/02-agents/response_stream.py	Fixed ChatResponse construction to use messages list instead of text/role
python/samples/02-agents/typed_options.py	Moved model_id to client constructor instead of options dict
python/samples/01-get-started/README.md	Updated to use Azure Foundry environment variables
python/samples/01-get-started/01_hello_agent.py	Added note about XML tags in samples
python/packages/core/tests/openai/test_openai_chat_client.py	Added test for single FunctionTool normalization
python/packages/core/tests/openai/test_openai_assistants_client.py	Added tests for tool_choice handling and single tool normalization
python/packages/core/tests/core/test_types.py	Updated test for validate_tool_mode(None) to expect None instead of {'mode': 'none'}
python/packages/core/tests/core/test_function_invocation_logic.py	Added test for tools passed via kwargs