model-library 0.1.5__tar.gz → 0.1.6__tar.gz

{model_library-0.1.5 → model_library-0.1.6}/Makefile

@@ -16,8 +16,10 @@ help:
 	@echo "  make examples <model> Run all examples with specified model"
 	@echo "  make browse_models    Interactively browse models and their configurations"
 
+PYTHON_VERSION ?= 3.11
+
 install:
-	uv venv
+	uv venv --python $(PYTHON_VERSION)
 	uv sync --dev
 	@echo "🎉 Done! Run 'source .venv/bin/activate' to activate the environment locally."
 

{model_library-0.1.5 → model_library-0.1.6}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: model-library
-Version: 0.1.5
+Version: 0.1.6
 Summary: Model Library for vals.ai 
 Author-email: "Vals AI, Inc." <contact@vals.ai>
 License: MIT

{model_library-0.1.5 → model_library-0.1.6}/model_library/base/base.py

@@ -6,7 +6,6 @@ from abc import ABC, abstractmethod
 from collections.abc import Awaitable
 from pprint import pformat
 from typing import (
-    TYPE_CHECKING,
     Any,
     Callable,
     Literal,
@@ -43,9 +42,6 @@ from model_library.exceptions import (
 )
 from model_library.utils import truncate_str
 
-if TYPE_CHECKING:
-    from model_library.providers.openai import OpenAIModel
-
 PydanticT = TypeVar("PydanticT", bound=BaseModel)
 
 
@@ -66,7 +62,7 @@ class LLMConfig(BaseModel):
     top_p: float | None = None
     top_k: int | None = None
     reasoning: bool = False
-    reasoning_effort: str | None = None
+    reasoning_effort: str | bool | None = None
     supports_images: bool = False
     supports_files: bool = False
     supports_videos: bool = False
@@ -110,7 +106,7 @@ class LLM(ABC):
         self.top_k: int | None = config.top_k
 
         self.reasoning: bool = config.reasoning
-        self.reasoning_effort: str | None = config.reasoning_effort
+        self.reasoning_effort: str | bool | None = config.reasoning_effort
 
         self.supports_files: bool = config.supports_files
         self.supports_videos: bool = config.supports_videos
@@ -120,7 +116,7 @@ class LLM(ABC):
         self.supports_tools: bool = config.supports_tools
 
         self.native: bool = config.native
-        self.delegate: "OpenAIModel | None" = None
+        self.delegate: "LLM | None" = None
         self.batch: LLMBatchMixin | None = None
 
         if config.provider_config:
@@ -198,11 +194,14 @@ class LLM(ABC):
         input: Sequence[InputItem],
         *,
         tools: list[ToolDefinition] = [],
+        query_logger: logging.Logger,
         **kwargs: object,
     ) -> QueryResult:
         if not self.delegate:
             raise Exception("Delegate not set")
-        return await self.delegate._query_impl(input, tools=tools, **kwargs)  # pyright: ignore[reportPrivateUsage]
+        return await self.delegate._query_impl(  # pyright: ignore[reportPrivateUsage]
+            input, tools=tools, query_logger=query_logger, **kwargs
+        )
 
     async def query(
         self,
@@ -213,6 +212,7 @@ class LLM(ABC):
         # for backwards compatibility
         files: list[FileInput] = [],
         images: list[FileInput] = [],
+        query_logger: logging.Logger | None = None,
         **kwargs: object,
     ) -> QueryResult:
         """
@@ -256,15 +256,18 @@ class LLM(ABC):
         input = [*history, *input]
 
         # unique logger for the query
-        query_id = uuid.uuid4().hex[:14]
-        query_logger = self.logger.getChild(f"query={query_id}")
+        if not query_logger:
+            query_id = uuid.uuid4().hex[:14]
+            query_logger = self.logger.getChild(f"query={query_id}")
 
         query_logger.info(
             "Query started:\n" + item_info + tool_info + f"--- kwargs: {short_kwargs}\n"
         )
 
         async def query_func() -> QueryResult:
-            return await self._query_impl(input, tools=tools, **kwargs)
+            return await self._query_impl(
+                input, tools=tools, query_logger=query_logger, **kwargs
+            )
 
         async def timed_query() -> tuple[QueryResult, float]:
             return await LLM.timer_wrapper(query_func)
@@ -361,7 +364,8 @@ class LLM(ABC):
         input: Sequence[InputItem],
         *,
         tools: list[ToolDefinition],
-        **kwargs: object,  # TODO: pass in query logger
+        query_logger: logging.Logger,
+        **kwargs: object,
     ) -> QueryResult:
         """
         Query the model with input

{model_library-0.1.5 → model_library-0.1.6}/model_library/base/delegate_only.py

@@ -1,4 +1,5 @@
 import io
+import logging
 from typing import Any, Literal, Sequence
 
 from typing_extensions import override
@@ -48,11 +49,14 @@ class DelegateOnly(LLM):
         input: Sequence[InputItem],
         *,
         tools: list[ToolDefinition],
+        query_logger: logging.Logger,
         **kwargs: object,
     ) -> QueryResult:
         assert self.delegate
 
-        return await self.delegate_query(input, tools=tools, **kwargs)
+        return await self.delegate_query(
+            input, tools=tools, query_logger=query_logger, **kwargs
+        )
 
     @override
     async def parse_input(

{model_library-0.1.5 → model_library-0.1.6}/model_library/config/all_models.json

@@ -1,4 +1,94 @@
 {
+    "minimax/MiniMax-M2.1": {
+        "company": "MiniMax",
+        "label": "MiniMax-M2.1",
+        "description": null,
+        "release_date": "2025-12-23",
+        "open_source": true,
+        "documentation_url": "https://platform.minimax.io/docs",
+        "properties": {
+            "context_window": 204800,
+            "max_tokens": 131000,
+            "training_cutoff": null,
+            "reasoning_model": true
+        },
+        "supports": {
+            "images": false,
+            "files": false,
+            "temperature": true,
+            "tools": true
+        },
+        "metadata": {
+            "deprecated": false,
+            "available_for_everyone": true,
+            "available_as_evaluator": false,
+            "ignored_for_cost": false
+        },
+        "provider_properties": {},
+        "costs_per_million_token": {
+            "input": 0.3,
+            "output": 1.2,
+            "cache": {
+                "read": 0.03,
+                "write": 0.375,
+                "write_markup": 1.0
+            }
+        },
+        "alternative_keys": [],
+        "default_parameters": {
+            "temperature": 1.0,
+            "top_p": 0.95
+        },
+        "provider_endpoint": "MiniMax-M2.1",
+        "provider_name": "minimax",
+        "full_key": "minimax/MiniMax-M2.1",
+        "slug": "minimax_MiniMax-M2.1"
+    },
+    "zai/glm-4.7": {
+        "company": "zAI",
+        "label": "GLM 4.7",
+        "description": "Latest model from ZAI",
+        "release_date": "2025-12-22",
+        "open_source": true,
+        "documentation_url": "https://docs.z.ai/",
+        "properties": {
+            "context_window": 200000,
+            "max_tokens": 128000,
+            "training_cutoff": null,
+            "reasoning_model": true
+        },
+        "supports": {
+            "images": false,
+            "files": false,
+            "temperature": true,
+            "tools": true
+        },
+        "metadata": {
+            "deprecated": false,
+            "available_for_everyone": true,
+            "available_as_evaluator": false,
+            "ignored_for_cost": false
+        },
+        "provider_properties": {},
+        "costs_per_million_token": {
+            "input": 0.6,
+            "output": 2.2,
+            "cache": {
+                "read": 0.11,
+                "read_discount": 1.0,
+                "write_markup": 1.0
+            }
+        },
+        "alternative_keys": [],
+        "default_parameters": {
+            "temperature": 1.0,
+            "top_p": 1.0
+        },
+        "provider_endpoint": "glm-4.7",
+        "provider_name": "zai",
+        "full_key": "zai/glm-4.7",
+        "slug": "zai_glm-4.7"
+    },
     "google/gemini-3-flash-preview": {
         "company": "Google",
         "label": "Gemini 3 Flash (12/25)",
@@ -504,7 +594,8 @@
             }
         ],
         "default_parameters": {
-            "temperature": 1.0
+            "temperature": 1.0,
+            "reasoning_effort": "none"
         },
         "provider_endpoint": "deepseek-v3p2",
         "provider_name": "fireworks",

{model_library-0.1.5 → model_library-0.1.6}/model_library/config/fireworks_models.yaml

@@ -150,6 +150,8 @@ deepseek-models:
       context_window: 160_000
       max_tokens: 20_480
       reasoning_model: false
+    default_parameters:
+      reasoning_effort: "none"
     costs_per_million_token:
       input: 0.56
       output: 1.68

{model_library-0.1.5 → model_library-0.1.6}/model_library/config/minimax_models.yaml

@@ -16,6 +16,24 @@ base-config:
 
 minimax-m2-models:
 
+  minimax/MiniMax-M2.1:
+    label: MiniMax-M2.1
+    release_date: 2025-12-23
+    properties:
+      context_window: 204_800
+      max_tokens: 131_000
+      reasoning_model: true
+      training_cutoff: null
+    default_parameters:
+      temperature: 1.0
+      top_p: 0.95
+    costs_per_million_token:
+      input: 0.30
+      output: 1.20
+      cache:
+        read: 0.03
+        write: 0.375
+
   minimax/MiniMax-M2:
     label: MiniMax-M2
     description: MiniMax-M2 is a cost-efficient open-source model optimized for agentic applications and coding in particular.

{model_library-0.1.5 → model_library-0.1.6}/model_library/config/zai_models.yaml

@@ -18,6 +18,20 @@ base-config:
       write_markup: 1
 
 zai-models:
+  zai/glm-4.7:
+    label: GLM 4.7
+    description: "Latest model from ZAI"
+    release_date: 2025-12-22
+    properties:
+      context_window: 200_000
+      max_tokens: 128_000
+    costs_per_million_token:
+      input: 0.6
+      output: 2.2
+      cache:
+        read: 0.11
+    default_parameters:
+      temperature: 1
   zai/glm-4.5:
     label: GLM 4.5
     description: "z.AI old model"

{model_library-0.1.5 → model_library-0.1.6}/model_library/providers/ai21labs.py

@@ -1,4 +1,5 @@
 import io
+import logging
 from typing import Any, Literal, Sequence
 
 from ai21 import AsyncAI21Client
@@ -137,6 +138,7 @@ class AI21LabsModel(LLM):
         input: Sequence[InputItem],
         *,
         tools: list[ToolDefinition],
+        query_logger: logging.Logger,
         **kwargs: object,
     ) -> QueryResult:
         messages: list[ChatMessage] = []

{model_library-0.1.5 → model_library-0.1.6}/model_library/providers/amazon.py

@@ -3,6 +3,7 @@ import asyncio
 import base64
 import io
 import json
+import logging
 from typing import Any, Literal, Sequence, cast
 
 import boto3
@@ -337,6 +338,7 @@ class AmazonModel(LLM):
         input: Sequence[InputItem],
         *,
         tools: list[ToolDefinition],
+        query_logger: logging.Logger,
         **kwargs: object,
     ) -> QueryResult:
         body = await self.build_body(input, tools=tools, **kwargs)

{model_library-0.1.5 → model_library-0.1.6}/model_library/providers/anthropic.py

@@ -1,4 +1,5 @@
 import io
+import logging
 from typing import Any, Literal, Sequence, cast
 
 from anthropic import AsyncAnthropic
@@ -249,6 +250,8 @@ class AnthropicModel(LLM):
 
     @override
     def get_client(self) -> AsyncAnthropic:
+        if self._delegate_client:
+            return self._delegate_client
         if not AnthropicModel._client:
             headers: dict[str, str] = {}
             AnthropicModel._client = AsyncAnthropic(
@@ -262,16 +265,20 @@ class AnthropicModel(LLM):
     def __init__(
         self,
         model_name: str,
-        provider: Literal["anthropic"] = "anthropic",
+        provider: str = "anthropic",
         *,
         config: LLMConfig | None = None,
+        custom_client: AsyncAnthropic | None = None,
     ):
         super().__init__(model_name, provider, config=config)
 
+        # allow custom client to act as delegate (native)
+        self._delegate_client: AsyncAnthropic | None = custom_client
+
         # https://docs.anthropic.com/en/api/openai-sdk
-        self.delegate: OpenAIModel | None = (
+        self.delegate = (
             None
-            if self.native
+            if self.native or custom_client
             else OpenAIModel(
                 model_name=self.model_name,
                 provider=provider,
@@ -285,7 +292,10 @@ class AnthropicModel(LLM):
         )
 
         # Initialize batch support if enabled
-        self.supports_batch: bool = self.supports_batch and self.native
+        # Disable batch when using custom_client (similar to OpenAI)
+        self.supports_batch: bool = (
+            self.supports_batch and self.native and not custom_client
+        )
         self.batch: LLMBatchMixin | None = (
             AnthropicBatchMixin(self) if self.supports_batch else None
         )
@@ -555,20 +565,36 @@ class AnthropicModel(LLM):
         input: Sequence[InputItem],
         *,
         tools: list[ToolDefinition],
+        query_logger: logging.Logger,
         **kwargs: object,
     ) -> QueryResult:
         if self.delegate:
-            return await self.delegate_query(input, tools=tools, **kwargs)
+            return await self.delegate_query(
+                input, tools=tools, query_logger=query_logger, **kwargs
+            )
 
         body = await self.create_body(input, tools=tools, **kwargs)
 
-        betas = ["files-api-2025-04-14", "interleaved-thinking-2025-05-14"]
-        if "sonnet-4-5" in self.model_name:
-            betas.append("context-1m-2025-08-07")
+        client = self.get_client()
 
-        async with self.get_client().beta.messages.stream(
-            **body,
-            betas=betas,
+        # only send betas for the official Anthropic endpoint
+        is_anthropic_endpoint = self._delegate_client is None
+        if not is_anthropic_endpoint:
+            client_base_url = getattr(client, "_base_url", None) or getattr(
+                client, "base_url", None
+            )
+            if client_base_url:
+                is_anthropic_endpoint = "api.anthropic.com" in str(client_base_url)
+
+        stream_kwargs = {**body}
+        if is_anthropic_endpoint:
+            betas = ["files-api-2025-04-14", "interleaved-thinking-2025-05-14"]
+            if "sonnet-4-5" in self.model_name:
+                betas.append("context-1m-2025-08-07")
+            stream_kwargs["betas"] = betas
+
+        async with client.beta.messages.stream(
+            **stream_kwargs,
         ) as stream:  # pyright: ignore[reportAny]
             message = await stream.get_final_message()
         self.logger.info(f"Anthropic Response finished: {message.id}")

{model_library-0.1.5 → model_library-0.1.6}/model_library/providers/google/google.py

@@ -1,5 +1,6 @@
 import base64
 import io
+import logging
 from typing import Any, Literal, Sequence, cast
 
 from google.genai import Client
@@ -54,6 +55,11 @@ from model_library.exceptions import (
 from model_library.providers.google.batch import GoogleBatchMixin
 from model_library.register_models import register_provider
 from model_library.utils import normalize_tool_result
+import uuid
+
+
+def generate_tool_call_id(tool_name: str) -> str:
+    return str(tool_name + "_" + str(uuid.uuid4()))
 
 
 class GoogleConfig(ProviderConfig):
@@ -328,6 +334,7 @@ class GoogleModel(LLM):
         input: Sequence[InputItem],
         *,
         tools: list[ToolDefinition],
+        query_logger: logging.Logger,
         **kwargs: object,
     ) -> QueryResult:
         body: dict[str, Any] = await self.create_body(input, tools=tools, **kwargs)
@@ -357,9 +364,10 @@ class GoogleModel(LLM):
 
                         call_args = part.function_call.args or {}
                         tool_calls.append(
-                            # weirdly, id is not required
+                            # Weirdly, id is not required. If not provided, we generate one.
                             ToolCall(
-                                id=part.function_call.id or "",
+                                id=part.function_call.id
+                                or generate_tool_call_id(part.function_call.name),
                                 name=part.function_call.name,
                                 args=call_args,
                             )

{model_library-0.1.5 → model_library-0.1.6}/model_library/providers/minimax.py

@@ -1,13 +1,12 @@
 from typing import Literal
 
 from model_library import model_library_settings
-from model_library.base import (
-    DelegateOnly,
-    LLMConfig,
-)
-from model_library.providers.openai import OpenAIModel
+from model_library.base import DelegateOnly, LLMConfig
+from model_library.providers.anthropic import AnthropicModel
 from model_library.register_models import register_provider
-from model_library.utils import create_openai_client_with_defaults
+from model_library.utils import default_httpx_client
+
+from anthropic import AsyncAnthropic
 
 
 @register_provider("minimax")
@@ -21,13 +20,14 @@ class MinimaxModel(DelegateOnly):
     ):
         super().__init__(model_name, provider, config=config)
 
-        self.delegate = OpenAIModel(
+        self.delegate = AnthropicModel(
             model_name=self.model_name,
             provider=self.provider,
             config=config,
-            custom_client=create_openai_client_with_defaults(
+            custom_client=AsyncAnthropic(
                 api_key=model_library_settings.MINIMAX_API_KEY,
-                base_url="https://api.minimax.io/v1",
+                base_url="https://api.minimax.io/anthropic",
+                http_client=default_httpx_client(),
+                max_retries=1,
             ),
-            use_completions=True,
         )

{model_library-0.1.5 → model_library-0.1.6}/model_library/providers/mistral.py

@@ -1,4 +1,5 @@
 import io
+import logging
 import time
 from collections.abc import Sequence
 from typing import Any, Literal
@@ -171,6 +172,7 @@ class MistralModel(LLM):
         input: Sequence[InputItem],
         *,
         tools: list[ToolDefinition],
+        query_logger: logging.Logger,
         **kwargs: object,
     ) -> QueryResult:
         # mistral supports max 8 images, merge extra images into the 8th image

{model_library-0.1.5 → model_library-0.1.6}/model_library/providers/openai.py

@@ -2,6 +2,7 @@ from __future__ import annotations
 
 import io
 import json
+import logging
 from typing import Any, Literal, Sequence, cast
 
 from openai import APIConnectionError, AsyncOpenAI
@@ -505,8 +506,11 @@ class OpenAIModel(LLM):
         if self.reasoning:
             del body["max_tokens"]
             body["max_completion_tokens"] = self.max_tokens
-            if self.reasoning_effort:
-                body["reasoning_effort"] = self.reasoning_effort
+
+        # some model endpoints (like `fireworks/deepseek-v3p2`)
+        # require explicitly setting reasoning effort to disable thinking
+        if self.reasoning_effort is not None:
+            body["reasoning_effort"] = self.reasoning_effort
 
         if self.supports_temperature:
             if self.temperature is not None:
@@ -701,8 +705,8 @@ class OpenAIModel(LLM):
 
         if self.reasoning:
             body["reasoning"] = {"summary": "auto"}
-            if self.reasoning_effort:
-                body["reasoning"]["effort"] = self.reasoning_effort
+            if self.reasoning_effort is not None:
+                body["reasoning"]["effort"] = self.reasoning_effort  # type: ignore[reportArgumentType]
 
         if self.supports_temperature:
             if self.temperature is not None:
@@ -722,6 +726,7 @@ class OpenAIModel(LLM):
         input: Sequence[InputItem],
         *,
         tools: list[ToolDefinition],
+        query_logger: logging.Logger,
         **kwargs: object,
     ) -> QueryResult:
         if self.use_completions:

{model_library-0.1.5 → model_library-0.1.6}/model_library/providers/vals.py

@@ -2,6 +2,7 @@ from __future__ import annotations
 
 import io
 import json
+import logging
 import random
 import re
 import time
@@ -271,6 +272,7 @@ class DummyAIModel(LLM):
         input: Sequence[InputItem],
         *,
         tools: list[ToolDefinition],
+        query_logger: logging.Logger,
         **kwargs: object,
     ) -> QueryResult:
         body = await self.create_body(input, tools=tools, **kwargs)

{model_library-0.1.5 → model_library-0.1.6}/model_library/providers/xai.py

@@ -1,5 +1,6 @@
 import asyncio
 import io
+import logging
 from typing import Any, Literal, Sequence, cast
 
 import grpc
@@ -78,7 +79,7 @@ class XAIModel(LLM):
         super().__init__(model_name, provider, config=config)
 
         # https://docs.x.ai/docs/guides/migration
-        self.delegate: OpenAIModel | None = (
+        self.delegate = (
             None
             if self.native
             else OpenAIModel(
@@ -225,13 +226,16 @@ class XAIModel(LLM):
         input: Sequence[InputItem],
         *,
         tools: list[ToolDefinition],
+        query_logger: logging.Logger,
         **kwargs: object,
     ) -> QueryResult:
         if self.reasoning_effort:
             kwargs["reasoning_effort"] = self.reasoning_effort
 
         if self.delegate:
-            return await self.delegate_query(input, tools=tools, **kwargs)
+            return await self.delegate_query(
+                input, tools=tools, query_logger=query_logger, **kwargs
+            )
 
         messages: Sequence[Message] = []
         if "system_prompt" in kwargs:

{model_library-0.1.5 → model_library-0.1.6}/model_library/register_models.py

@@ -169,7 +169,7 @@ class DefaultParameters(BaseModel):
     temperature: float | None = None
     top_p: float | None = None
     top_k: int | None = None
-    reasoning_effort: str | None = None
+    reasoning_effort: str | bool | None = None
 
 
 class RawModelConfig(BaseModel):

{model_library-0.1.5 → model_library-0.1.6}/model_library.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: model-library
-Version: 0.1.5
+Version: 0.1.6
 Summary: Model Library for vals.ai 
 Author-email: "Vals AI, Inc." <contact@vals.ai>
 License: MIT

{model_library-0.1.5 → model_library-0.1.6}/model_library.egg-info/SOURCES.txt

@@ -114,6 +114,7 @@ tests/unit/test_context_window.py
 tests/unit/test_deep_research.py
 tests/unit/test_perplexity_provider.py
 tests/unit/test_prompt_caching.py
+tests/unit/test_query_logger.py
 tests/unit/test_registry.py
 tests/unit/test_result_metadata.py
 tests/unit/test_retry.py

{model_library-0.1.5 → model_library-0.1.6}/tests/unit/providers/test_fireworks_provider.py

@@ -32,6 +32,7 @@ async def test_fireworks_delegate_query_passes_kwargs():
         result = await model._query_impl(
             [TextInput(text="hello")],
             tools=[],
+            query_logger=MagicMock(),
             extra="value",
         )
 

{model_library-0.1.5 → model_library-0.1.6}/tests/unit/test_deep_research.py

@@ -1,4 +1,4 @@
-from unittest.mock import AsyncMock, Mock
+from unittest.mock import AsyncMock, MagicMock, Mock
 
 import pytest
 
@@ -110,7 +110,9 @@ async def test_query_impl_parses_deep_research_response():
 
     web_search_tool = Mock()
     web_search_tool.body = {"type": "web_search"}
-    result = await model._query_impl([], tools=[web_search_tool], stream=False)  # pyright: ignore[reportPrivateUsage]
+    result = await model._query_impl(
+        [], tools=[web_search_tool], stream=False, query_logger=MagicMock()
+    )  # pyright: ignore[reportPrivateUsage]
 
     assert result.output_text == "Report [1]"
     assert result.reasoning == "Reasoning"

{model_library-0.1.5 → model_library-0.1.6}/tests/unit/test_perplexity_provider.py

@@ -60,7 +60,9 @@ async def test_perplexity_delegate_query():
         query_result = QueryResult(output_text="ok")
         model.delegate._query_impl = AsyncMock(return_value=query_result)  # type: ignore[method-assign]
 
-        result = await model._query_impl([TextInput(text="hello")], tools=[])
+        result = await model._query_impl(
+            [TextInput(text="hello")], tools=[], query_logger=MagicMock()
+        )
 
         delegate_impl = model.delegate._query_impl  # type: ignore[attr-defined]
         delegate_impl.assert_awaited_once()

{model_library-0.1.5 → model_library-0.1.6}/tests/unit/test_prompt_caching.py

@@ -1,4 +1,5 @@
 from types import TracebackType
+from unittest.mock import MagicMock
 
 import pytest
 
@@ -152,7 +153,9 @@ async def test_anthropic_query_maps_cache_usage():
     m = AnthropicModel("claude-haiku-4-5-20251001", config=LLMConfig())
     m.get_client = lambda: _DummyClient()
 
-    res = await m._query_impl([TextInput(text="hi")], tools=[])
+    res = await m._query_impl(
+        [TextInput(text="hi")], tools=[], query_logger=MagicMock()
+    )
     assert res.metadata.in_tokens == 10
     assert res.metadata.out_tokens == 2
     assert res.metadata.cache_read_tokens == 456

model_library-0.1.6/tests/unit/test_query_logger.py

@@ -0,0 +1,21 @@
+"""Unit tests for query_logger parameter handling in LLM.query()"""
+
+from unittest.mock import AsyncMock, MagicMock
+
+from model_library.base import LLM, QueryResult
+
+
+async def test_query_uses_provided_query_logger(mock_llm: LLM):
+    """
+    Test that when query_logger is passed as a kwarg, it is used for logging
+    instead of generating a fresh logger.
+    """
+    custom_logger = MagicMock()
+
+    query_impl_mock = AsyncMock(return_value=QueryResult(output_text="success"))
+    mock_llm._query_impl = query_impl_mock  # pyright: ignore[reportPrivateUsage]
+
+    await mock_llm.query("Test input", query_logger=custom_logger)
+
+    # Verify the custom logger was used for logging
+    assert custom_logger.info.call_count