unique_toolkit 1.46.3__tar.gz → 1.46.5__tar.gz

{unique_toolkit-1.46.3 → unique_toolkit-1.46.5}/CHANGELOG.md

@@ -6,6 +6,16 @@ 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).
 
 
+## [1.46.5] - 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
+
 ## [1.46.3] - 2026-02-09
 ### Fixed
 - Fix hallucination check crash when responses contain invalid source indices (e.g., from code blocks with array indexing)
@@ -14,6 +24,9 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 ### Removed
 - Remove unused `regex` dependency (no longer needed after refactoring)
 
+## [1.46.2] - 2026-02-09
+- Add `litellm:anthropic-claude-opus-4-6` and `AZURE_MODEL_ROUTER_2025_1118` to `info.py`
+
 ## [1.46.1] - 2026-02-06
 - Add model-agnostic token counting via `LanguageModelInfo.get_encoder()`
 - Add bundled Qwen/DeepSeek tokenizers for accurate token counting

{unique_toolkit-1.46.3 → unique_toolkit-1.46.5}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: unique_toolkit
-Version: 1.46.3
+Version: 1.46.5
 Summary: 
 License: Proprietary
 Author: Cedric Klinkert
@@ -126,6 +126,16 @@ 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).
 
 
+## [1.46.5] - 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
+
 ## [1.46.3] - 2026-02-09
 ### Fixed
 - Fix hallucination check crash when responses contain invalid source indices (e.g., from code blocks with array indexing)
@@ -134,6 +144,9 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 ### Removed
 - Remove unused `regex` dependency (no longer needed after refactoring)
 
+## [1.46.2] - 2026-02-09
+- Add `litellm:anthropic-claude-opus-4-6` and `AZURE_MODEL_ROUTER_2025_1118` to `info.py`
+
 ## [1.46.1] - 2026-02-06
 - Add model-agnostic token counting via `LanguageModelInfo.get_encoder()`
 - Add bundled Qwen/DeepSeek tokenizers for accurate token counting

{unique_toolkit-1.46.3 → unique_toolkit-1.46.5}/pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "unique_toolkit"
-version = "1.46.3"
+version = "1.46.5"
 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-1.46.5/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-1.46.3 → unique_toolkit-1.46.5}/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-1.46.5/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-1.46.3 → unique_toolkit-1.46.5}/unique_toolkit/language_model/infos.py

@@ -220,7 +220,7 @@ class LanguageModelInfo(BaseModel):
     version: str = Field(title="Model Version", default="")
 
     encoder_name: EncoderName | Annotated[str, Field(title="Custom Encoder Name")] = (
-        EncoderName.CL100K_BASE
+        Field(default=EncoderName.CL100K_BASE, union_mode="left_to_right")
     )
 
     # TODO: Discuss if this is a sensible defaut

unique_toolkit-1.46.3/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-1.46.3/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