unique_toolkit 2026.42.0.dev5__py3-none-any.whl → 2026.42.0.dev7__py3-none-any.whl

unique_toolkit/agentic/loop_runner/_responses_iteration_handler_utils.py

@@ -1,6 +1,9 @@
 import logging
+from collections.abc import Sequence
 from typing import Unpack
 
+from openai.types.responses import ToolParam, response_create_params
+
 from unique_toolkit.agentic.loop_runner._responses_stream_handler_utils import (
     responses_stream_response,
 )
@@ -9,12 +12,44 @@ from unique_toolkit.agentic.loop_runner.base import (
 )
 from unique_toolkit.language_model.schemas import (
     LanguageModelTokenUsage,
+    LanguageModelToolDescription,
     ResponsesLanguageModelStreamResponse,
 )
 
 _LOGGER = logging.getLogger(__name__)
 
 
+def _is_forced_function_choice(
+    tool_choice: response_create_params.ToolChoice,
+) -> bool:
+    return isinstance(tool_choice, dict) and tool_choice.get("type") == "function"
+
+
+def _is_hosted_tool(tool: LanguageModelToolDescription | ToolParam) -> bool:
+    """Return whether OpenAI executes the tool."""
+    if isinstance(tool, LanguageModelToolDescription):
+        return False
+    return tool.get("type") != "function"
+
+
+def _tools_for_forced_choice(
+    tools: Sequence[LanguageModelToolDescription | ToolParam] | None,
+    tool_choice: response_create_params.ToolChoice,
+) -> list[LanguageModelToolDescription | ToolParam] | None:
+    """Drop hosted tools from requests that force a function tool."""
+    if tools is None or not _is_forced_function_choice(tool_choice):
+        return list(tools) if tools is not None else None
+
+    remaining = [tool for tool in tools if not _is_hosted_tool(tool)]
+    dropped = len(tools) - len(remaining)
+    if dropped > 0:
+        _LOGGER.info(
+            "Dropped %d hosted tool(s) from the forced function request.",
+            dropped,
+        )
+    return remaining
+
+
 async def handle_responses_last_iteration(
     **kwargs: Unpack[_ResponsesLoopIterationRunnerKwargs],
 ) -> ResponsesLanguageModelStreamResponse:
@@ -47,9 +82,16 @@ async def handle_responses_forced_tools_iteration(
     responses: list[ResponsesLanguageModelStreamResponse] = []
 
     for opt in tool_choices:
-        responses.append(
-            await responses_stream_response(loop_runner_kwargs=kwargs, tool_choice=opt)
-        )
+        tools = _tools_for_forced_choice(kwargs.get("tools"), opt)
+        if tools is None:
+            response = await responses_stream_response(
+                loop_runner_kwargs=kwargs, tool_choice=opt
+            )
+        else:
+            response = await responses_stream_response(
+                loop_runner_kwargs=kwargs, tool_choice=opt, tools=tools
+            )
+        responses.append(response)
 
     # Merge responses and refs:
     tool_calls = []

unique_toolkit/language_model/infos.py

@@ -1825,7 +1825,9 @@ class LanguageModelInfo(BaseModel):
                     ),
                     info_cutoff_at=date(2026, 1, 1),
                     published_at=date(2026, 2, 17),
-                    supported_reasoning_efforts=[],
+                    # Effort includes max. xhigh is not accepted.
+                    default_options={"reasoning_effort": "medium"},
+                    supported_reasoning_efforts=["low", "medium", "high", "max"],
                 )
             case (
                 LanguageModelName.ANTHROPIC_CLAUDE_SONNET_5
@@ -1853,7 +1855,15 @@ class LanguageModelInfo(BaseModel):
                     temperature_bounds=TemperatureBounds(
                         min_temperature=1.0, max_temperature=1.0
                     ),
-                    supported_reasoning_efforts=[],
+                    # Effort includes xhigh and max.
+                    default_options={"reasoning_effort": "medium"},
+                    supported_reasoning_efforts=[
+                        "low",
+                        "medium",
+                        "high",
+                        "xhigh",
+                        "max",
+                    ],
                 )
             case (
                 LanguageModelName.ANTHROPIC_CLAUDE_SONNET_5_5
@@ -1879,7 +1889,15 @@ class LanguageModelInfo(BaseModel):
                     ),
                     info_cutoff_at=date(2026, 6, 1),
                     published_at=date(2026, 9, 28),
-                    supported_reasoning_efforts=[],
+                    # Effort includes xhigh and max.
+                    default_options={"reasoning_effort": "medium"},
+                    supported_reasoning_efforts=[
+                        "low",
+                        "medium",
+                        "high",
+                        "xhigh",
+                        "max",
+                    ],
                     supports_forced_tool_choice=False,
                 )
             case LanguageModelName.ANTHROPIC_CLAUDE_OPUS_4:
@@ -1943,7 +1961,9 @@ class LanguageModelInfo(BaseModel):
                     ),
                     info_cutoff_at=date(2025, 8, 1),
                     published_at=date(2025, 11, 13),
-                    supported_reasoning_efforts=[],
+                    # https://platform.claude.com/docs/en/build-with-claude/effort
+                    default_options={"reasoning_effort": "medium"},
+                    supported_reasoning_efforts=["low", "medium", "high"],
                 )
             case (
                 LanguageModelName.ANTHROPIC_CLAUDE_OPUS_4_6
@@ -1967,7 +1987,9 @@ class LanguageModelInfo(BaseModel):
                     ),
                     info_cutoff_at=date(2025, 8, 1),
                     published_at=date(2026, 2, 5),
-                    supported_reasoning_efforts=[],
+                    # Effort includes max. xhigh is not accepted.
+                    default_options={"reasoning_effort": "medium"},
+                    supported_reasoning_efforts=["low", "medium", "high", "max"],
                 )
             case (
                 LanguageModelName.ANTHROPIC_CLAUDE_OPUS_4_7
@@ -1992,7 +2014,15 @@ class LanguageModelInfo(BaseModel):
                     ),
                     info_cutoff_at=date(2026, 1, 1),
                     published_at=date(2026, 4, 16),
-                    supported_reasoning_efforts=[],
+                    # Effort includes xhigh and max.
+                    default_options={"reasoning_effort": "medium"},
+                    supported_reasoning_efforts=[
+                        "low",
+                        "medium",
+                        "high",
+                        "xhigh",
+                        "max",
+                    ],
                 )
             case (
                 LanguageModelName.ANTHROPIC_CLAUDE_OPUS_4_8
@@ -2017,7 +2047,15 @@ class LanguageModelInfo(BaseModel):
                     ),
                     info_cutoff_at=date(2026, 1, 1),
                     published_at=date(2026, 5, 28),
-                    supported_reasoning_efforts=[],
+                    # Effort includes xhigh and max.
+                    default_options={"reasoning_effort": "medium"},
+                    supported_reasoning_efforts=[
+                        "low",
+                        "medium",
+                        "high",
+                        "xhigh",
+                        "max",
+                    ],
                 )
             case (
                 LanguageModelName.ANTHROPIC_CLAUDE_OPUS_5
@@ -2042,7 +2080,15 @@ class LanguageModelInfo(BaseModel):
                     ),
                     info_cutoff_at=date(2026, 5, 1),
                     published_at=date(2026, 7, 24),
-                    supported_reasoning_efforts=[],
+                    # Effort includes xhigh and max.
+                    default_options={"reasoning_effort": "medium"},
+                    supported_reasoning_efforts=[
+                        "low",
+                        "medium",
+                        "high",
+                        "xhigh",
+                        "max",
+                    ],
                 )
             case (
                 LanguageModelName.ANTHROPIC_CLAUDE_OPUS_5_5
@@ -2068,7 +2114,15 @@ class LanguageModelInfo(BaseModel):
                     ),
                     info_cutoff_at=date(2026, 6, 1),
                     published_at=date(2026, 9, 22),
-                    supported_reasoning_efforts=[],
+                    # Effort includes xhigh and max. API default is medium.
+                    default_options={"reasoning_effort": "medium"},
+                    supported_reasoning_efforts=[
+                        "low",
+                        "medium",
+                        "high",
+                        "xhigh",
+                        "max",
+                    ],
                     supports_forced_tool_choice=False,
                 )
             case (
@@ -2094,7 +2148,15 @@ class LanguageModelInfo(BaseModel):
                     ),
                     info_cutoff_at=date(2026, 1, 1),
                     published_at=date(2026, 7, 1),
-                    supported_reasoning_efforts=[],
+                    # Effort includes xhigh and max.
+                    default_options={"reasoning_effort": "medium"},
+                    supported_reasoning_efforts=[
+                        "low",
+                        "medium",
+                        "high",
+                        "xhigh",
+                        "max",
+                    ],
                 )
             case LanguageModelName.GEMINI_2_0_FLASH:
                 return cls(
@@ -2283,7 +2345,9 @@ class LanguageModelInfo(BaseModel):
                     ),
                     info_cutoff_at=date(2025, 1, day=1),
                     published_at=date(2026, 1, 19),
-                    supported_reasoning_efforts=[],
+                    # thinkingLevel is low, medium, or high.
+                    default_options={"reasoning_effort": "medium"},
+                    supported_reasoning_efforts=["low", "medium", "high"],
                 )
             case LanguageModelName.GEMINI_3_6_FLASH:
                 return cls(
@@ -2305,6 +2369,7 @@ class LanguageModelInfo(BaseModel):
                     ),
                     info_cutoff_at=date(2026, 3, day=1),
                     published_at=date(2026, 7, 21),
+                    default_options={"reasoning_effort": "medium"},
                     supported_reasoning_efforts=["minimal", "low", "medium", "high"],
                 )
             case (
@@ -2335,6 +2400,7 @@ class LanguageModelInfo(BaseModel):
                     published_at=date(2026, 9, 2),
                     # minimal is rejected; default thinking level is medium when unset.
                     # https://ai.google.dev/gemini-api/docs/models/gemini-3.8-flash
+                    default_options={"reasoning_effort": "medium"},
                     supported_reasoning_efforts=["low", "medium", "high"],
                 )
             case LanguageModelName.GEMINI_3_FLASH_PREVIEW:
@@ -2356,7 +2422,9 @@ class LanguageModelInfo(BaseModel):
                     ),
                     info_cutoff_at=date(2025, 1, day=1),
                     published_at=date(2025, 12, 17),
-                    supported_reasoning_efforts=[],
+                    # thinkingLevel includes minimal. Default when unset is high.
+                    default_options={"reasoning_effort": "medium"},
+                    supported_reasoning_efforts=["minimal", "low", "medium", "high"],
                 )
             case LanguageModelName.GEMINI_3_PRO_PREVIEW:
                 return cls(
@@ -2377,7 +2445,9 @@ class LanguageModelInfo(BaseModel):
                     ),
                     info_cutoff_at=date(2025, 1, day=1),
                     published_at=date(2025, 11, 13),
-                    supported_reasoning_efforts=[],
+                    # thinkingLevel is low or high.
+                    default_options={"reasoning_effort": "low"},
+                    supported_reasoning_efforts=["low", "high"],
                 )
             case LanguageModelName.GROK_4_1_FAST_NON_REASONING:
                 return cls(
@@ -2446,6 +2516,7 @@ class LanguageModelInfo(BaseModel):
                     ),
                     info_cutoff_at=date(2026, 2, day=1),
                     published_at=date(2026, 7, 16),
+                    default_options={"reasoning_effort": "medium"},
                     supported_reasoning_efforts=["low", "medium", "high"],
                 )
             case LanguageModelName.GROK_4_6:
@@ -3313,7 +3384,9 @@ class LanguageModelInfo(BaseModel):
                         token_limit_input=1_000_000, token_limit_output=384_000
                     ),
                     published_at=date(2026, 1, 1),
-                    supported_reasoning_efforts=[],
+                    # reasoning_effort is high or max.
+                    default_options={"reasoning_effort": "high"},
+                    supported_reasoning_efforts=["high", "max"],
                 )
             case LanguageModelName.LITELLM_DEEPSEEK_V4_1_FLASH:
                 return cls(
@@ -3337,7 +3410,9 @@ class LanguageModelInfo(BaseModel):
                     token_limits=LanguageModelTokenLimits(
                         token_limit_input=1_000_000, token_limit_output=384_000
                     ),
-                    supported_reasoning_efforts=[],
+                    # Same V4 effort scale as Pro: high or max.
+                    default_options={"reasoning_effort": "high"},
+                    supported_reasoning_efforts=["high", "max"],
                 )
             case LanguageModelName.LITELLM_GLM_5_1:
                 return cls(
@@ -3394,7 +3469,9 @@ class LanguageModelInfo(BaseModel):
                         token_limit_input=1_000_000, token_limit_output=1_000_000
                     ),
                     published_at=date(2026, 7, 31),  # HF repo creation date
-                    supported_reasoning_efforts=[],
+                    # Phoeniqs accepts low, high, and max.
+                    default_options={"reasoning_effort": "low"},
+                    supported_reasoning_efforts=["low", "high", "max"],
                 )
             case LanguageModelName.LITELLM_PHOENIQS_GEMMA_4_31B:
                 return cls(
@@ -3433,7 +3510,9 @@ class LanguageModelInfo(BaseModel):
                         2026, 6, 16
                     ),  # HF repo creation date — unlike the
                     # Together AI entry above, not an unsourced placeholder.
-                    supported_reasoning_efforts=[],
+                    # Phoeniqs accepts low and high.
+                    default_options={"reasoning_effort": "low"},
+                    supported_reasoning_efforts=["low", "high"],
                 )
             case LanguageModelName.LITELLM_PHOENIQS_GLM_5_3_FLASH:
                 return cls(
@@ -3516,7 +3595,9 @@ class LanguageModelInfo(BaseModel):
                         token_limit_output=128_000,
                     ),
                     published_at=date(2026, 7, 27),
-                    supported_reasoning_efforts=[],
+                    # reasoning_effort is low, high, or max.
+                    default_options={"reasoning_effort": "low"},
+                    supported_reasoning_efforts=["low", "high", "max"],
                 )
             case LanguageModelName.LITELLM_QWEN_3:
                 return cls(

{unique_toolkit-2026.42.0.dev5.dist-info → unique_toolkit-2026.42.0.dev7.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.3
 Name: unique-toolkit
-Version: 2026.42.0.dev5
+Version: 2026.42.0.dev7
 Summary: 
 Author: Cedric Klinkert, Martin Fadler, Sadique Sheik, Fabian Schläpfer, Pascal Hauri, Ejiroghene Oki-Peter, Rami Azouz, Ricardo Visini, Sanzio Monti
 Author-email: Cedric Klinkert <cedric.klinkert@unique.ch>, Martin Fadler <martin.fadler@unique.ch>, Sadique Sheik <sadique@unique.ch>, Fabian Schläpfer <fabian@unique.ch>, Pascal Hauri <pascal@unique.ch>, Ejiroghene Oki-Peter <oki-peter.ejiroghene@unique.ai>, Rami Azouz <rami.ext@unique.ch>, Ricardo Visini <ricardo.visini@unique.ch>, Sanzio Monti <sanzio.monti@unique.ai>

{unique_toolkit-2026.42.0.dev5.dist-info → unique_toolkit-2026.42.0.dev7.dist-info}/RECORD

@@ -120,7 +120,7 @@ unique_toolkit/agentic/history_manager/settings.py,sha256=Lrx3a8yXeU0MXrWlV6RZSH
 unique_toolkit/agentic/history_manager/utils.py,sha256=DiBip8xtah3I9-G-_Ul3OjlYKunrXR-5RJtzREHWV_0,6579
 unique_toolkit/agentic/loop_runner/__init__.py,sha256=bhwOIloPn2MZyCw60FkHy330twCCRovg3bgye9OjZRE,1496
 unique_toolkit/agentic/loop_runner/_iteration_handler_utils.py,sha256=y3tjTAtpkyL_Y7279_3v0fGaq9FUYN3UquwdIB3ZhgI,3831
-unique_toolkit/agentic/loop_runner/_responses_iteration_handler_utils.py,sha256=DuQRsToV0KLs7YcFjJVnEmrryhlX5q4isztx2jUxZcI,2064
+unique_toolkit/agentic/loop_runner/_responses_iteration_handler_utils.py,sha256=SbfV5KCZ-3QiMRYjaDDUnY0JuBtfBCSBofF6sxTOugI,3615
 unique_toolkit/agentic/loop_runner/_responses_stream_handler_utils.py,sha256=6HXpSPGUp096a175-NvXqU2Gw7e8Msjo60CP4wGFOoM,2340
 unique_toolkit/agentic/loop_runner/_stream_handler_utils.py,sha256=ERk-qS8MG8FM2F2AEzaub_XsuZQD1tsD8-li692LKp0,1708
 unique_toolkit/agentic/loop_runner/base.py,sha256=OddFwxT7zOOLd2MlAef63BNhzH-NNET8RAsfaMcsHQE,3307
@@ -397,7 +397,7 @@ unique_toolkit/language_model/builder.py,sha256=KwBzsD6-fvQnB4p2I7M-5Tk6j4tmQ973
 unique_toolkit/language_model/constants.py,sha256=B-topqW0r83dkC_25DeQfnPk3n53qzIHUCBS7YJ0-1U,119
 unique_toolkit/language_model/default_language_model.py,sha256=mcyuFs_V6tp52C8xJSHiEHO4kFGj6xed1gvozvCRD7k,1181
 unique_toolkit/language_model/functions.py,sha256=W6JC4LUh-LZ9waDgph--xVmKeQ7Ls6NRAqDx9DwQdeY,26691
-unique_toolkit/language_model/infos.py,sha256=y4J53RwG-K5QjRY9EeYQ55ioX-JPZTINBZaenrVAIOo,174566
+unique_toolkit/language_model/infos.py,sha256=O5NBD-3z_8GvifPjxWXGihx-RrUKKhi1wFMojdb7NdA,178580
 unique_toolkit/language_model/invocation_stats.py,sha256=CQbpUz394IsVd8eNTB1gPSMt5f0UXmkN1amNsqtAW6U,5857
 unique_toolkit/language_model/model_costs.py,sha256=ShuDAzMew1fUkJ3tzPgU5Zqjp6ivUZaG97BFh93tEn4,6358
 unique_toolkit/language_model/prompt.py,sha256=JSawaLjQg3VR-E2fK8engFyJnNdk21zaO8pPIodzN4Q,3991
@@ -426,6 +426,6 @@ unique_toolkit/short_term_memory/service.py,sha256=2JMBEBn6GoyK-a0xvdcWpPft-3C74
 unique_toolkit/smart_rules/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 unique_toolkit/smart_rules/compile.py,sha256=vRe0aMJWuVIK-VsqomgxQImzke7MDmwgSxxJfYCa7zw,1386
 unique_toolkit/test_utilities/events.py,sha256=FLrhMh0A_mQutAUjoD0FdMcVQ4Gq31r35OVGzRjvLro,6760
-unique_toolkit-2026.42.0.dev5.dist-info/WHEEL,sha256=eh7sammvW2TypMMMGKgsM83HyA_3qQ5Lgg3ynoecH3M,79
-unique_toolkit-2026.42.0.dev5.dist-info/METADATA,sha256=zQfpXjEiup6e3ByDDnz0aKemSh-10qr0ujqFVkHJDWE,7821
-unique_toolkit-2026.42.0.dev5.dist-info/RECORD,,
+unique_toolkit-2026.42.0.dev7.dist-info/WHEEL,sha256=eh7sammvW2TypMMMGKgsM83HyA_3qQ5Lgg3ynoecH3M,79
+unique_toolkit-2026.42.0.dev7.dist-info/METADATA,sha256=lerw6mdP9Cp9f4mxf3nMFsJEZg9R_iak69i2o-uZcwU,7821
+unique_toolkit-2026.42.0.dev7.dist-info/RECORD,,