Python: [BREAKING] Python: Provider-leading client design & OpenAI package extraction

[eavanvalkenburg]

Motivation and Context

Make agent-framework-core lightweight by extracting OpenAI clients into their own package, and improve discoverability by using provider-leading naming conventions. See ADR-0020 for full rationale.

Description

New agent-framework-openai package:

• OpenAIChatClient (Responses API, was OpenAIResponsesClient)
• OpenAIChatCompletionClient (Chat Completions API, was OpenAIChatClient)
• OpenAIEmbeddingClient
• Raw clients do all work (settings, client creation); public clients just combine layers

Core made lighter:

• Removed openai, azure-identity, azure-ai-projects, packaging from core deps
• agent_framework.openai and agent_framework.azure are lazy-loading gateways

New Foundry clients in azure-ai:

• FoundryChatClient / RawFoundryChatClient — Responses API via Foundry project
• FoundryAgent / RawFoundryAgent — connect to pre-configured PromptAgents/HostedAgents
• RawFoundryAgentChatClient — extension point for custom client middleware

Parameter unification:

• model_id / deployment_name / model_deployment_name → model
• OPENAI_RESPONSES_MODEL_ID / OPENAI_CHAT_MODEL_ID → OPENAI_MODEL
• FOUNDRY_MODEL, FOUNDRY_PROJECT_ENDPOINT for Foundry clients

Class hierarchy cleanup:

• OpenAIBase / OpenAIConfigMixin removed from non-deprecated MRO
• _ensure_client / _initialize_client removed (client set directly in Raw init)
• Deprecated: AzureOpenAI* clients, AzureAIClient, AzureAIAgentClient, OpenAIAssistantsClient
• All deprecated Azure OpenAI classes in single _deprecated_azure_openai.py for clean deletion

Samples:

• azure_openai/ + azure_ai/ + azure_ai_agent/ merged into azure/
• CRUD-based Azure AI samples removed (no longer supported)
• New FoundryAgent and FoundryChatClient samples added
• Default client for concept samples: FoundryChatClient
• as_agent() replaced with Agent(client=...) pattern
• Always AzureCliCredential, never DefaultAzureCredential

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? Yes — client renames, parameter renames, package extraction

[Copilot]

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

[markwallace-microsoft]

Python Test Coverage Report •

File	Stmts	Miss	Cover	Missing
packages/azure-ai-search/agent_framework_azure_ai_search
_context_provider.py	336	10	97%	110–111, 284, 335–336, 459–460, 638–639, 686
packages/azure-ai/agent_framework_azure_ai
_agent_provider.py	114	1	99%	270
_chat_client.py	495	78	84%	465–466, 468, 658, 663–664, 666–667, 670, 673, 675, 680, 941–942, 944, 947, 950, 953–958, 961, 963, 971, 983–985, 989, 992–993, 1001–1004, 1014, 1022–1025, 1027–1028, 1030–1031, 1038, 1046–1047, 1055–1068, 1073, 1076, 1084, 1090, 1098–1100, 1103, 1123–1124, 1257, 1284, 1299, 1415, 1465, 1471
_client.py	403	49	87%	214, 394, 396, 444, 454–458, 460–466, 479, 540, 555–560, 602, 618–619, 645–646, 664, 695, 697, 718–719, 722, 767, 770, 772, 863, 894, 936, 1145, 1148, 1151–1152, 1154–1157
_deprecated_azure_openai.py	332	62	81%	152–159, 161–162, 164–165, 167, 170–183, 185, 187–193, 195, 197–198, 200, 309, 328, 333, 338, 344, 350, 354, 395, 401–404, 542, 547, 558, 612–613, 622, 853, 863, 869, 873, 892
_entra_id_authentication.py	22	0	100%
_foundry_agent.py	44	19	56%	157, 159, 161–163, 165–168, 173, 175–178, 183, 185–186, 188, 194
_foundry_agent_client.py	100	30	70%	159–160, 164–166, 171–174, 215, 237–238, 250–252, 254–255, 257–263, 265–266, 268, 270, 274–275
_foundry_chat_client.py	121	49	59%	126, 135–137, 139, 141–142, 146–148, 153–155, 160–162, 164, 170, 174–177, 201, 203–206, 211, 213–216, 221, 223–224, 226, 232, 254, 315, 317, 405, 407, 409–410, 412–415, 508
_foundry_memory_provider.py	91	0	100%
_project_provider.py	122	5	95%	157, 213, 312, 352, 385
packages/azure-cosmos/agent_framework_azure_cosmos
_history_provider.py	125	5	96%	162–164, 261, 268
packages/core/agent_framework
_types.py	1074	88	91%	58, 67–68, 122, 127, 146, 148, 152, 156, 158, 160, 162, 180, 184, 210, 232, 237, 242, 246, 276, 686–687, 846–847, 1232, 1304, 1339, 1359, 1369, 1413, 1545–1547, 1737, 1828–1833, 1858, 2030, 2042, 2306, 2331, 2339, 2434, 2665, 2871, 2940, 2951, 2953–2957, 2959, 2962–2970, 2980, 3185–3187, 3190–3192, 3196, 3201, 3205, 3289–3291, 3320, 3374, 3397–3401, 3407, 3491
packages/core/agent_framework/_workflows
_events.py	149	21	85%	90–91, 224, 228, 230, 262, 337, 352, 367, 382, 394–396, 408–410, 412–413, 415–416, 420
packages/foundry_local/agent_framework_foundry_local
_foundry_local_client.py	42	0	100%
packages/openai/agent_framework_openai
_assistant_provider.py	109	11	89%	162, 290, 356, 471–478
_assistants_client.py	333	36	89%	428, 430, 432, 435, 439–440, 443, 446, 451–452, 454, 457–459, 464, 475, 500, 502, 504, 506, 508, 513, 516, 519, 523, 534, 729, 815, 818, 837, 848, 885–888, 958
_chat_client.py	859	131	84%	296, 298–299, 337, 355, 408–411, 415–416, 421–422, 432–433, 440, 455–461, 482, 490, 513, 610, 709, 768, 770, 772, 774, 840, 854, 934, 944, 949, 992, 1071, 1088, 1101, 1166, 1259, 1264, 1268–1270, 1274–1275, 1341, 1370, 1376, 1386, 1392, 1397, 1403, 1408–1409, 1470, 1492–1493, 1508–1509, 1527–1528, 1569–1572, 1734, 1789, 1791, 1871–1879, 2001, 2056, 2071, 2091–2101, 2114, 2125–2129, 2143, 2157–2168, 2177, 2209–2212, 2220–2221, 2223–2225, 2239–2241, 2251–2252, 2258, 2273
_chat_completion_client.py	367	33	91%	205, 207–208, 264, 314, 395–396, 400, 525, 532, 608–615, 617–620, 630, 708, 710, 727, 748, 776, 789, 813, 833, 873, 948
_embedding_client.py	92	10	89%	93, 95–96, 133, 135, 146, 156, 191, 292, 300
_exceptions.py	33	0	100%
_shared.py	138	22	84%	64, 70–73, 165, 167–168, 184, 186, 188, 195, 197, 210, 291, 296, 315, 374–377, 379
packages/orchestrations/agent_framework_orchestrations
_sequential.py	88	11	87%	74, 172, 183, 189, 220, 222–223, 225, 230, 247, 251
packages/purview/agent_framework_purview
_client.py	128	2	98%	74–75
_middleware.py	117	0	100%
TOTAL	27759	3412	87%

Python Unit Test Overview

Tests	Skipped	Failures	Errors	Time
5369	20 💤	0 ❌	0 🔥	1m 27s ⏱️