unique-toolkit 2026.24.0.dev5__tar.gz → 2026.24.0.dev7__tar.gz

{unique_toolkit-2026.24.0.dev5 → unique_toolkit-2026.24.0.dev7}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.3
 Name: unique-toolkit
-Version: 2026.24.0.dev5
+Version: 2026.24.0.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-2026.24.0.dev5 → unique_toolkit-2026.24.0.dev7}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "unique_toolkit"
-version = "2026.24.0.dev5"
+version = "2026.24.0.dev7"
 description = ""
 readme = "README.md"
 requires-python = ">=3.12"

unique_toolkit-2026.24.0.dev7/unique_toolkit/__init__.py

@@ -0,0 +1,149 @@
+from importlib import import_module
+from typing import TYPE_CHECKING, Any
+
+if TYPE_CHECKING:
+    from unique_toolkit.chat import ChatService as ChatService
+    from unique_toolkit.content import ContentService as ContentService
+    from unique_toolkit.data_extraction import (
+        StructuredOutputDataExtractor as StructuredOutputDataExtractor,
+    )
+    from unique_toolkit.data_extraction import (
+        StructuredOutputDataExtractorConfig as StructuredOutputDataExtractorConfig,
+    )
+    from unique_toolkit.embedding import EmbeddingService as EmbeddingService
+    from unique_toolkit.framework_utilities.openai.client import (
+        get_async_openai_client as get_async_openai_client,
+    )
+    from unique_toolkit.framework_utilities.openai.client import (
+        get_openai_client as get_openai_client,
+    )
+    from unique_toolkit.language_model import (
+        LanguageModelMessages as LanguageModelMessages,
+    )
+    from unique_toolkit.language_model import LanguageModelName as LanguageModelName
+    from unique_toolkit.language_model import (
+        LanguageModelService as LanguageModelService,
+    )
+    from unique_toolkit.language_model import (
+        LanguageModelToolDescription as LanguageModelToolDescription,
+    )
+    from unique_toolkit.services.knowledge_base import (
+        KnowledgeBaseService as KnowledgeBaseService,
+    )
+    from unique_toolkit.short_term_memory import (
+        ShortTermMemoryService as ShortTermMemoryService,
+    )
+
+_EXPORTS: dict[str, tuple[str, str]] = {
+    "ChatService": ("unique_toolkit.chat", "ChatService"),
+    "ContentService": ("unique_toolkit.content", "ContentService"),
+    "EmbeddingService": ("unique_toolkit.embedding", "EmbeddingService"),
+    "get_openai_client": (
+        "unique_toolkit.framework_utilities.openai.client",
+        "get_openai_client",
+    ),
+    "get_async_openai_client": (
+        "unique_toolkit.framework_utilities.openai.client",
+        "get_async_openai_client",
+    ),
+    "KnowledgeBaseService": (
+        "unique_toolkit.services.knowledge_base",
+        "KnowledgeBaseService",
+    ),
+    "LanguageModelMessages": (
+        "unique_toolkit.language_model",
+        "LanguageModelMessages",
+    ),
+    "LanguageModelName": ("unique_toolkit.language_model", "LanguageModelName"),
+    "LanguageModelService": ("unique_toolkit.language_model", "LanguageModelService"),
+    "LanguageModelToolDescription": (
+        "unique_toolkit.language_model",
+        "LanguageModelToolDescription",
+    ),
+    "ShortTermMemoryService": (
+        "unique_toolkit.short_term_memory",
+        "ShortTermMemoryService",
+    ),
+    "StructuredOutputDataExtractor": (
+        "unique_toolkit.data_extraction",
+        "StructuredOutputDataExtractor",
+    ),
+    "StructuredOutputDataExtractorConfig": (
+        "unique_toolkit.data_extraction",
+        "StructuredOutputDataExtractorConfig",
+    ),
+}
+
+# Conditionally import config_checker (requires pydantic-settings)
+_CONFIG_CHECKER_AVAILABLE: bool = False
+try:
+    from unique_toolkit._common.config_checker import (
+        ConfigDiffer,  # noqa: F401, I001
+        ConfigEntry,  # noqa: F401, I001
+        ConfigExporter,  # noqa: F401, I001
+        ConfigRegistry,  # noqa: F401, I001
+        ConfigValidator,  # noqa: F401, I001
+        DefaultChangeReport,  # noqa: F401, I001
+        ExportManifest,  # noqa: F401, I001
+        ValidationReport,  # noqa: F401, I001
+        register_config,  # noqa: F401, I001
+    )
+
+    _CONFIG_CHECKER_AVAILABLE = True  # pyright: ignore[reportConstantRedefinition]
+except ImportError:
+    pass
+
+# Conditionally import langchain utilities if langchain is installed
+_LANGCHAIN_AVAILABLE: bool = False
+try:
+    from unique_toolkit.framework_utilities.langchain.client import get_langchain_client  # noqa: F401, I001
+
+    _LANGCHAIN_AVAILABLE = True  # pyright: ignore[reportConstantRedefinition]
+except ImportError:
+    pass
+
+# You can add other classes you frequently use here as well
+
+__all__ = [
+    "ChatService",
+    "ContentService",
+    "EmbeddingService",
+    "get_openai_client",
+    "get_async_openai_client",
+    "KnowledgeBaseService",
+    "LanguageModelMessages",
+    "LanguageModelName",
+    "LanguageModelService",
+    "LanguageModelToolDescription",
+    "ShortTermMemoryService",
+    "StructuredOutputDataExtractor",
+    "StructuredOutputDataExtractorConfig",
+]
+
+# Add config_checker exports if available
+if _CONFIG_CHECKER_AVAILABLE:
+    __all__.extend(
+        [
+            "ConfigDiffer",
+            "ConfigEntry",
+            "ConfigExporter",
+            "ConfigRegistry",
+            "ConfigValidator",
+            "DefaultChangeReport",
+            "ExportManifest",
+            "ValidationReport",
+            "register_config",
+        ]
+    )
+# Add langchain-specific exports if available
+if _LANGCHAIN_AVAILABLE:
+    __all__.append("get_langchain_client")
+
+
+def __getattr__(name: str) -> Any:
+    export = _EXPORTS.get(name)
+    if export is None:
+        raise AttributeError(f"module {__name__!r} has no attribute {name!r}")
+
+    module_name, attribute_name = export
+    return getattr(import_module(module_name), attribute_name)

{unique_toolkit-2026.24.0.dev5 → unique_toolkit-2026.24.0.dev7}/unique_toolkit/app/schemas.py

@@ -4,10 +4,17 @@ import json
 from enum import StrEnum
 from logging import getLogger
 from pathlib import Path
-from typing import Any, Generic, Optional, TypeVar, override
+from typing import Annotated, Any, Generic, Optional, TypeVar, override
 
 from humps import camelize
-from pydantic import BaseModel, ConfigDict, Field, JsonValue, field_validator
+from pydantic import (
+    BaseModel,
+    BeforeValidator,
+    ConfigDict,
+    Field,
+    JsonValue,
+    field_validator,
+)
 from pydantic_settings import BaseSettings
 from typing_extensions import deprecated
 
@@ -16,6 +23,8 @@ from unique_toolkit.app.chat_event_filter_options_settings import (
     CHAT_EVENT_FILTER_OPTIONS_SETTINGS,
 )
 from unique_toolkit.app.unique_settings import UniqueChatEventFilterOptions
+from unique_toolkit.language_model.default_language_model import DEFAULT_LANGUAGE_MODEL
+from unique_toolkit.language_model.infos import LanguageModelInfo
 from unique_toolkit.smart_rules.compile import UniqueQL, parse_uniqueql
 
 FilterOptionsT = TypeVar("FilterOptionsT", bound=BaseSettings)
@@ -29,6 +38,15 @@ model_config = ConfigDict(
 _logger = getLogger(__name__)
 
 
+def _validate_language_model_info(v: Any) -> Any:
+    from unique_toolkit._common.validators import validate_and_init_language_model_info
+
+    return validate_and_init_language_model_info(v)
+
+
+LMI = Annotated[LanguageModelInfo, BeforeValidator(_validate_language_model_info)]
+
+
 class EventName(StrEnum):
     EXTERNAL_MODULE_CHOSEN = "unique.chat.external-module.chosen"
     USER_MESSAGE_CREATED = "unique.chat.user-message.created"
@@ -247,6 +265,10 @@ class ChatEventPayload(BaseModel):
     user_metadata: dict[str, Any] | None = Field(
         default_factory=dict,
     )
+    model_choice: LMI = Field(
+        default_factory=lambda: LanguageModelInfo.from_name(DEFAULT_LANGUAGE_MODEL),
+        description="The model choice the user has chosen to be used.",
+    )
     tool_choices: list[str] = Field(
         default_factory=list,
         description="A list containing the tool names the user has chosen to be activated.",
@@ -305,6 +327,10 @@ class ChatEventPayload(BaseModel):
         if value:
             return parse_uniqueql(value)
 
+    @property
+    def has_model_choice_override(self) -> bool:
+        return "model_choice" in self.model_fields_set
+
 
 @deprecated("""Use `ChatEventPayload` instead.
             This class will be removed in the next major version.""")

{unique_toolkit-2026.24.0.dev5 → unique_toolkit-2026.24.0.dev7}/unique_toolkit/content/__init__.py

@@ -1,3 +1,5 @@
+from typing import TYPE_CHECKING, Any
+
 from .constants import DOMAIN_NAME as DOMAIN_NAME
 from .schemas import (
     Content as Content,
@@ -23,7 +25,6 @@ from .schemas import (
 from .schemas import (
     ContentUploadInput as ContentUploadInput,
 )
-from .service import ContentService as ContentService
 from .utils import (
     count_tokens as count_tokens,
 )
@@ -36,3 +37,31 @@ from .utils import (
 from .utils import (
     sort_content_chunks as sort_content_chunks,
 )
+
+if TYPE_CHECKING:
+    from .service import ContentService as ContentService
+
+__all__ = [
+    "Content",
+    "ContentChunk",
+    "ContentMetadata",
+    "ContentReference",
+    "ContentRerankerConfig",
+    "ContentSearchResult",
+    "ContentSearchType",
+    "ContentService",
+    "ContentUploadInput",
+    "DOMAIN_NAME",
+    "count_tokens",
+    "merge_content_chunks",
+    "pick_content_chunks_for_token_window",
+    "sort_content_chunks",
+]
+
+
+def __getattr__(name: str) -> Any:
+    if name == "ContentService":
+        from .service import ContentService
+
+        return ContentService
+    raise AttributeError(f"module {__name__!r} has no attribute {name!r}")

{unique_toolkit-2026.24.0.dev5 → unique_toolkit-2026.24.0.dev7}/unique_toolkit/language_model/__init__.py

@@ -1,3 +1,5 @@
+from typing import TYPE_CHECKING, Any
+
 from .constants import DOMAIN_NAME
 from .default_language_model import DEFAULT_LANGUAGE_MODEL, DEFAULT_GPT_4o
 from .functions import (
@@ -28,9 +30,6 @@ from .schemas import (
     LanguageModelToolParameters,
     LanguageModelUserMessage,
 )
-from .service import (
-    LanguageModelService,
-)
 from .stream_transform import (
     NormalizationTransform,
     ReferenceInjectionTransform,
@@ -42,6 +41,19 @@ from .utils import (
     find_last_json_object,
 )
 
+if TYPE_CHECKING:
+    from .service import LanguageModelService as LanguageModelService
+
+
+def __getattr__(name: str) -> Any:
+    if name == "LanguageModelService":
+        from .service import LanguageModelService
+
+        return LanguageModelService
+
+    raise AttributeError(f"module {__name__!r} has no attribute {name!r}")
+
+
 __all__ = [
     "LanguageModel",
     "LanguageModelName",

unique_toolkit-2026.24.0.dev7/unique_toolkit/monitoring/gunicorn.py

@@ -0,0 +1,24 @@
+from __future__ import annotations
+
+import os
+
+
+def child_exit(server, worker) -> None:
+    """Gunicorn child_exit hook: clean up dead worker's per-PID metric files.
+
+    Wire this up in gunicorn.conf.py:
+
+        from unique_toolkit.monitoring.gunicorn import child_exit  # noqa: F401
+
+    Gunicorn calls any function named child_exit(server, worker) in the
+    config file whenever a worker process exits. Without this, dead workers'
+    .db files accumulate in PROMETHEUS_MULTIPROC_DIR and inflate aggregated
+    metric values on every future scrape.
+
+    No-ops when PROMETHEUS_MULTIPROC_DIR is not set (single-process mode).
+    """
+    if not os.environ.get("PROMETHEUS_MULTIPROC_DIR"):
+        return
+    from prometheus_client import multiprocess  # pyright: ignore[reportMissingImports]
+
+    multiprocess.mark_process_dead(worker.pid)

{unique_toolkit-2026.24.0.dev5 → unique_toolkit-2026.24.0.dev7}/unique_toolkit/monitoring/middleware.py

@@ -11,22 +11,23 @@ from prometheus_client import (  # pyright: ignore[reportMissingImports]
 from .registry import REGISTRY
 
 _http_requests_total = Counter(
-    "http_requests_total",
+    "python_http_requests_total",
     "Total HTTP requests",
     ["method", "path", "status_code"],
     registry=REGISTRY,
 )
 _http_request_duration_seconds = Histogram(
-    "http_request_duration_seconds",
+    "python_http_request_duration_seconds",
     "HTTP request duration",
     ["method", "path"],
     registry=REGISTRY,
 )
 _http_requests_in_progress = Gauge(
-    "http_requests_in_progress",
+    "python_http_requests_in_progress",
     "In-flight HTTP requests",
     ["method"],
     registry=REGISTRY,
+    multiprocess_mode="livesum",
 )
 
 _DEFAULT_EXCLUDED_PATHS = frozenset({"/health", "/metrics", "/"})

{unique_toolkit-2026.24.0.dev5 → unique_toolkit-2026.24.0.dev7}/unique_toolkit/monitoring/registry.py

@@ -2,9 +2,11 @@ from __future__ import annotations
 
 import asyncio
 import functools
+import os
 import time
+import warnings
 from contextlib import contextmanager
-from typing import Any, Callable
+from typing import Any, Callable, Literal
 
 from prometheus_client import (  # pyright: ignore[reportMissingImports]
     CollectorRegistry,
@@ -18,7 +20,36 @@ REGISTRY = CollectorRegistry()
 
 
 def get_metrics() -> bytes:
-    """Returns all registered metrics in Prometheus text format."""
+    """Returns all registered metrics in Prometheus text format.
+
+    In multiprocess mode (PROMETHEUS_MULTIPROC_DIR is set), aggregates
+    per-PID .db files from all Gunicorn workers via MultiProcessCollector.
+    In single-process mode, returns metrics from the in-process REGISTRY.
+
+    NOTE: PROMETHEUS_MULTIPROC_DIR must be set in the shell/entrypoint
+    *before* Python starts. Setting it from Python after prometheus_client
+    has been imported has no effect — metrics will silently bypass per-PID
+    files and this function will emit a warning.
+    """
+    if os.environ.get("PROMETHEUS_MULTIPROC_DIR"):
+        from prometheus_client import (  # pyright: ignore[reportMissingImports]
+            multiprocess,
+            values,
+        )
+
+        if not getattr(values.ValueClass, "_multiprocess", False):
+            warnings.warn(
+                "PROMETHEUS_MULTIPROC_DIR is set but prometheus_client was imported "
+                "before the env var — metrics will silently bypass per-PID files. "
+                "Set PROMETHEUS_MULTIPROC_DIR in the shell/entrypoint before launching Python.",
+                RuntimeWarning,
+                stacklevel=2,
+            )
+
+        registry = CollectorRegistry()
+        multiprocess.MultiProcessCollector(registry)
+        return generate_latest(registry)
+
     return generate_latest(REGISTRY)
 
 
@@ -43,8 +74,31 @@ class MetricNamespace:
     def counter(self, name: str, description: str, labels: list[str]) -> Counter:
         return Counter(f"{self._prefix}_{name}", description, labels, registry=REGISTRY)
 
-    def gauge(self, name: str, description: str, labels: list[str]) -> Gauge:
-        return Gauge(f"{self._prefix}_{name}", description, labels, registry=REGISTRY)
+    def gauge(
+        self,
+        name: str,
+        description: str,
+        labels: list[str],
+        multiprocess_mode: Literal[
+            "all",
+            "liveall",
+            "min",
+            "livemin",
+            "max",
+            "livemax",
+            "sum",
+            "livesum",
+            "mostrecent",
+            "livemostrecent",
+        ] = "all",
+    ) -> Gauge:
+        return Gauge(
+            f"{self._prefix}_{name}",
+            description,
+            labels,
+            registry=REGISTRY,
+            multiprocess_mode=multiprocess_mode,
+        )
 
 
 @contextmanager

unique_toolkit-2026.24.0.dev5/unique_toolkit/__init__.py

@@ -1,85 +0,0 @@
-# Re-export commonly used classes for easier imports
-from unique_toolkit.chat import ChatService
-from unique_toolkit.content import ContentService
-from unique_toolkit.data_extraction import (
-    StructuredOutputDataExtractor,
-    StructuredOutputDataExtractorConfig,
-)
-from unique_toolkit.embedding import EmbeddingService
-from unique_toolkit.framework_utilities.openai.client import (
-    get_async_openai_client,
-    get_openai_client,
-)
-from unique_toolkit.language_model import (
-    LanguageModelMessages,
-    LanguageModelName,
-    LanguageModelService,
-    LanguageModelToolDescription,
-)
-from unique_toolkit.services.knowledge_base import KnowledgeBaseService
-from unique_toolkit.short_term_memory import ShortTermMemoryService
-
-# Conditionally import config_checker (requires pydantic-settings)
-_CONFIG_CHECKER_AVAILABLE: bool = False
-try:
-    from unique_toolkit._common.config_checker import (
-        ConfigDiffer,  # noqa: F401, I001
-        ConfigEntry,  # noqa: F401, I001
-        ConfigExporter,  # noqa: F401, I001
-        ConfigRegistry,  # noqa: F401, I001
-        ConfigValidator,  # noqa: F401, I001
-        DefaultChangeReport,  # noqa: F401, I001
-        ExportManifest,  # noqa: F401, I001
-        ValidationReport,  # noqa: F401, I001
-        register_config,  # noqa: F401, I001
-    )
-
-    _CONFIG_CHECKER_AVAILABLE = True  # pyright: ignore[reportConstantRedefinition]
-except ImportError:
-    pass
-
-# Conditionally import langchain utilities if langchain is installed
-_LANGCHAIN_AVAILABLE: bool = False
-try:
-    from unique_toolkit.framework_utilities.langchain.client import get_langchain_client  # noqa: F401, I001
-
-    _LANGCHAIN_AVAILABLE = True  # pyright: ignore[reportConstantRedefinition]
-except ImportError:
-    pass
-
-# You can add other classes you frequently use here as well
-
-__all__ = [
-    "ChatService",
-    "ContentService",
-    "EmbeddingService",
-    "get_openai_client",
-    "get_async_openai_client",
-    "KnowledgeBaseService",
-    "LanguageModelMessages",
-    "LanguageModelName",
-    "LanguageModelService",
-    "LanguageModelToolDescription",
-    "ShortTermMemoryService",
-    "StructuredOutputDataExtractor",
-    "StructuredOutputDataExtractorConfig",
-]
-
-# Add config_checker exports if available
-if _CONFIG_CHECKER_AVAILABLE:
-    __all__.extend(
-        [
-            "ConfigDiffer",
-            "ConfigEntry",
-            "ConfigExporter",
-            "ConfigRegistry",
-            "ConfigValidator",
-            "DefaultChangeReport",
-            "ExportManifest",
-            "ValidationReport",
-            "register_config",
-        ]
-    )
-# Add langchain-specific exports if available
-if _LANGCHAIN_AVAILABLE:
-    __all__.append("get_langchain_client")