unique_toolkit 1.46.4__tar.gz → 1.46.6__tar.gz

{unique_toolkit-1.46.4 → unique_toolkit-1.46.6}/CHANGELOG.md

@@ -5,6 +5,15 @@ All notable changes to this project will be documented in this file.
 The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/), 
 and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
+## [1.46.6] - 2026-02-12
+- Add `litellm:grok-4-1-fast-non-reasoning` and `litellm:grok-4-1-fast-non-reasoning` to `info.py`
+
+## [1.46.5] - 2026-02-11
+- Add `DEFAULT_LANGUAGE_MODEL` environment-based resolution with fallback support for invalid or missing values
+- Allow default model configuration via both enum value strings and enum member names
+- Export `DEFAULT_LANGUAGE_MODEL` and `DEFAULT_GPT_4o` from `unique_toolkit.language_model` for easier imports
+- Keep `unique_toolkit._common.default_language_model.DEFAULT_GPT_4o` backward compatible via re-export
+- Add unit tests for default language model resolution behavior
 
 ## [1.46.4] - 2026-02-10
 - Fix: Add `union_mode="left_to_right"` to `encoder_name` field so Pydantic tries `EncoderName` enum before `str`, enabling bundled tokenizers when UI sends string values

{unique_toolkit-1.46.4 → unique_toolkit-1.46.6}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: unique_toolkit
-Version: 1.46.4
+Version: 1.46.6
 Summary: 
 License: Proprietary
 Author: Cedric Klinkert
@@ -125,6 +125,15 @@ All notable changes to this project will be documented in this file.
 The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/), 
 and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
+## [1.46.6] - 2026-02-12
+- Add `litellm:grok-4-1-fast-non-reasoning` and `litellm:grok-4-1-fast-non-reasoning` to `info.py`
+
+## [1.46.5] - 2026-02-11
+- Add `DEFAULT_LANGUAGE_MODEL` environment-based resolution with fallback support for invalid or missing values
+- Allow default model configuration via both enum value strings and enum member names
+- Export `DEFAULT_LANGUAGE_MODEL` and `DEFAULT_GPT_4o` from `unique_toolkit.language_model` for easier imports
+- Keep `unique_toolkit._common.default_language_model.DEFAULT_GPT_4o` backward compatible via re-export
+- Add unit tests for default language model resolution behavior
 
 ## [1.46.4] - 2026-02-10
 - Fix: Add `union_mode="left_to_right"` to `encoder_name` field so Pydantic tries `EncoderName` enum before `str`, enabling bundled tokenizers when UI sends string values

{unique_toolkit-1.46.4 → unique_toolkit-1.46.6}/pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "unique_toolkit"
-version = "1.46.4"
+version = "1.46.6"
 description = ""
 authors = [
     "Cedric Klinkert <cedric.klinkert@unique.ch>",
@@ -11,6 +11,7 @@ authors = [
     "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>"
 ]
 readme = ["README.md", "CHANGELOG.md"]
 license = "Proprietary"

unique_toolkit-1.46.6/unique_toolkit/_common/default_language_model.py

@@ -0,0 +1,15 @@
+import warnings
+
+from unique_toolkit.language_model.default_language_model import (
+    DEFAULT_LANGUAGE_MODEL as DEFAULT_LANGUAGE_MODEL,
+)
+from unique_toolkit.language_model.default_language_model import (
+    DEFAULT_GPT_4o as DEFAULT_GPT_4o,
+)
+
+warnings.warn(
+    "unique_toolkit._common.default_language_model is deprecated. "
+    "Import DEFAULT_GPT_4o from unique_toolkit.language_model instead.",
+    DeprecationWarning,
+    stacklevel=2,
+)

{unique_toolkit-1.46.4 → unique_toolkit-1.46.6}/unique_toolkit/language_model/__init__.py

@@ -1,4 +1,10 @@
 from .constants import DOMAIN_NAME as DOMAIN_NAME
+from .default_language_model import (
+    DEFAULT_LANGUAGE_MODEL as DEFAULT_LANGUAGE_MODEL,
+)
+from .default_language_model import (
+    DEFAULT_GPT_4o as DEFAULT_GPT_4o,
+)
 from .infos import LanguageModel as LanguageModel
 from .infos import LanguageModelName as LanguageModelName
 from .prompt import (

unique_toolkit-1.46.6/unique_toolkit/language_model/default_language_model.py

@@ -0,0 +1,37 @@
+import logging
+import os
+
+from unique_toolkit.language_model.infos import LanguageModelName
+
+_LOGGER = logging.getLogger(__name__)
+_ENV_VAR = "DEFAULT_LANGUAGE_MODEL"
+DEFAULT_GPT_4o = LanguageModelName.AZURE_GPT_4o_2024_1120
+_LANGUAGE_MODELS_BY_VALUE = {model.value: model for model in LanguageModelName}
+
+
+def resolve_default_language_model() -> LanguageModelName:
+    """Resolve language model name from env var with safe fallback."""
+    configured_default = os.getenv(_ENV_VAR)
+    if not configured_default:
+        return DEFAULT_GPT_4o
+
+    # Accept enum values (e.g. "litellm:openai-gpt-5") and member names
+    # (e.g. "LITELLM_OPENAI_GPT_5").
+    resolved_by_value = _LANGUAGE_MODELS_BY_VALUE.get(configured_default)
+    if resolved_by_value is not None:
+        return resolved_by_value
+
+    resolved_by_name = LanguageModelName.__members__.get(configured_default)
+    if resolved_by_name is not None:
+        return resolved_by_name
+
+    _LOGGER.warning(
+        "Invalid %s=%r. Falling back to %s.",
+        _ENV_VAR,
+        configured_default,
+        DEFAULT_GPT_4o.value,
+    )
+    return DEFAULT_GPT_4o
+
+
+DEFAULT_LANGUAGE_MODEL = resolve_default_language_model()

{unique_toolkit-1.46.4 → unique_toolkit-1.46.6}/unique_toolkit/language_model/infos.py

@@ -71,6 +71,8 @@ class LanguageModelName(StrEnum):
     GEMINI_2_5_PRO_PREVIEW_0605 = "litellm:gemini-2-5-pro-preview-06-05"
     GEMINI_3_FLASH_PREVIEW = "litellm:gemini-3-flash-preview"
     GEMINI_3_PRO_PREVIEW = "litellm:gemini-3-pro-preview"
+    GROK_4_1_FAST_NON_REASONING = "litellm:grok-4-1-fast-non-reasoning"
+    GROK_4_1_FAST_REASONING = "litellm:grok-4-1-fast-reasoning"
     LITELLM_OPENAI_GPT_5 = "litellm:openai-gpt-5"
     LITELLM_OPENAI_GPT_5_MINI = "litellm:openai-gpt-5-mini"
     LITELLM_OPENAI_GPT_5_NANO = "litellm:openai-gpt-5-nano"
@@ -1407,6 +1409,47 @@ class LanguageModelInfo(BaseModel):
                     info_cutoff_at=date(2025, 1, day=1),
                     published_at=date(2025, 11, 13),
                 )
+            case LanguageModelName.GROK_4_1_FAST_NON_REASONING:
+                return cls(
+                    name=model_name,
+                    capabilities=[
+                        ModelCapabilities.FUNCTION_CALLING,
+                        ModelCapabilities.STREAMING,
+                        ModelCapabilities.STRUCTURED_OUTPUT,
+                    ],
+                    provider=LanguageModelProvider.LITELLM,
+                    version="grok-4-1-fast-non-reasoning",
+                    encoder_name=EncoderName.O200K_BASE,  # TODO: Update encoder with grok tokenizer
+                    token_limits=LanguageModelTokenLimits(
+                        # Context window is 2_000_000, we leave 200_000 tokens as buffer due to tokenizer mismatch
+                        # Assign 90% for input and 10% for output
+                        token_limit_input=int(1_800_000 * 0.9),
+                        token_limit_output=int(1_800_000 * 0.1),
+                    ),
+                    info_cutoff_at=date(2024, 11, day=4),
+                    published_at=date(2025, 11, 19),
+                )
+            case LanguageModelName.GROK_4_1_FAST_REASONING:
+                return cls(
+                    name=model_name,
+                    capabilities=[
+                        ModelCapabilities.FUNCTION_CALLING,
+                        ModelCapabilities.STREAMING,
+                        ModelCapabilities.STRUCTURED_OUTPUT,
+                        ModelCapabilities.REASONING,
+                    ],
+                    provider=LanguageModelProvider.LITELLM,
+                    version="grok-4-1-fast-reasoning",
+                    encoder_name=EncoderName.O200K_BASE,  # TODO: Update encoder with grok tokenizer
+                    token_limits=LanguageModelTokenLimits(
+                        # Context window is 2_000_000, we leave 200_000 tokens as buffer due to tokenizer mismatch
+                        # Assign 90% for input and 10% for output
+                        token_limit_input=int(1_800_000 * 0.9),
+                        token_limit_output=int(1_800_000 * 0.1),
+                    ),
+                    info_cutoff_at=date(2024, 11, day=4),
+                    published_at=date(2025, 11, 19),
+                )
             case LanguageModelName.LITELLM_OPENAI_GPT_5:
                 return cls(
                     name=model_name,

unique_toolkit-1.46.4/unique_toolkit/_common/default_language_model.py

@@ -1,12 +0,0 @@
-import warnings
-
-from unique_toolkit.language_model.infos import LanguageModelName
-
-warnings.warn(
-    "unique_toolkit._common.default_language_model is deprecated. "
-    "Import DEFAULT_GPT_4o from unique_toolkit.language_model instead.",
-    DeprecationWarning,
-    stacklevel=2,
-)
-
-DEFAULT_GPT_4o = LanguageModelName.AZURE_GPT_4o_2024_1120

unique_toolkit-1.46.4/unique_toolkit/language_model/default_language_model.py

@@ -1,3 +0,0 @@
-from unique_toolkit.language_model.infos import LanguageModelName
-
-DEFAULT_GPT_4o = LanguageModelName.AZURE_GPT_4o_2024_1120