unique_toolkit 0.8.7__tar.gz → 0.8.9__tar.gz

{unique_toolkit-0.8.7 → unique_toolkit-0.8.9}/CHANGELOG.md

@@ -5,6 +5,12 @@ 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).
 
+## [0.8.9] - 2025-08-15
+- Reduce input token limits for `ANTHROPIC_CLAUDE_3_7_SONNET_THINKING`, `ANTHROPIC_CLAUDE_3_7_SONNET`, `ANTHROPIC_CLAUDE_OPUS_4` and `ANTHROPIC_CLAUDE_SONNET_4` to 180_000 from 200_000
+
+## [0.8.8] - 2025-08-11
+- Make chat service openai stream response openai compatible 
+- Make `ChatMessage` openai compatible
 
 ## [0.8.7] - 2025-08-11
 - Make chat service openai compatible

{unique_toolkit-0.8.7 → unique_toolkit-0.8.9}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: unique_toolkit
-Version: 0.8.7
+Version: 0.8.9
 Summary: 
 License: Proprietary
 Author: Martin Fadler
@@ -114,6 +114,12 @@ 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).
 
+## [0.8.9] - 2025-08-15
+- Reduce input token limits for `ANTHROPIC_CLAUDE_3_7_SONNET_THINKING`, `ANTHROPIC_CLAUDE_3_7_SONNET`, `ANTHROPIC_CLAUDE_OPUS_4` and `ANTHROPIC_CLAUDE_SONNET_4` to 180_000 from 200_000
+
+## [0.8.8] - 2025-08-11
+- Make chat service openai stream response openai compatible 
+- Make `ChatMessage` openai compatible
 
 ## [0.8.7] - 2025-08-11
 - Make chat service openai compatible

{unique_toolkit-0.8.7 → unique_toolkit-0.8.9}/pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "unique_toolkit"
-version = "0.8.7"
+version = "0.8.9"
 description = ""
 authors = [
     "Martin Fadler <martin.fadler@unique.ch>",

{unique_toolkit-0.8.7 → unique_toolkit-0.8.9}/unique_toolkit/chat/schemas.py

@@ -2,6 +2,16 @@ from datetime import datetime
 from enum import StrEnum
 
 from humps import camelize
+from openai.types.chat import (
+    ChatCompletionAssistantMessageParam,
+    ChatCompletionUserMessageParam,
+)
+from openai.types.chat.chat_completion_message_function_tool_call_param import (
+    ChatCompletionMessageFunctionToolCallParam,
+)
+from openai.types.chat.chat_completion_message_function_tool_call_param import (
+    Function as OpenAIFunction,
+)
 from pydantic import (
     BaseModel,
     ConfigDict,
@@ -14,14 +24,15 @@ from unique_toolkit.content.schemas import ContentReference
 
 # set config to convert camelCase to snake_case
 model_config = ConfigDict(
-    alias_generator=camelize, populate_by_name=True, arbitrary_types_allowed=True
+    alias_generator=camelize,
+    populate_by_name=True,
 )
 
 
 class ChatMessageRole(StrEnum):
     USER = "user"
     ASSISTANT = "assistant"
-    TOOL = "tool"
+    TOOL = "tool"  # TODO: Unused according @unique-fabian. To be removed in separate PR
 
 
 class Function(BaseModel):
@@ -30,6 +41,12 @@ class Function(BaseModel):
     name: str
     arguments: str
 
+    def to_openai(self) -> OpenAIFunction:
+        return OpenAIFunction(
+            arguments=self.arguments,
+            name=self.name,
+        )
+
 
 class ToolCall(BaseModel):
     model_config = model_config
@@ -38,8 +55,16 @@ class ToolCall(BaseModel):
     type: str
     function: Function
 
+    def to_openai_param(self) -> ChatCompletionMessageFunctionToolCallParam:
+        return ChatCompletionMessageFunctionToolCallParam(
+            id=self.id,
+            function=self.function.to_openai(),
+            type="function",
+        )
+
 
 class ChatMessage(BaseModel):
+    # TODO: The below seems not to be True anymore @irina-unique. To be checked in separate PR
     # This model should strictly meets https://github.com/Unique-AG/monorepo/blob/master/node/apps/node-chat/src/public-api/2023-12-06/dtos/message/public-message.dto.ts
     model_config = model_config
 
@@ -71,6 +96,27 @@ class ChatMessage(BaseModel):
             raise ValueError("tool_call_ids is required when role is 'tool'")
         return self
 
+    def to_openai_param(self):
+        match self.role:
+            case ChatMessageRole.USER:
+                return ChatCompletionUserMessageParam(
+                    role="user",
+                    content=self.content or "",
+                )
+
+            case ChatMessageRole.ASSISTANT:
+                return ChatCompletionAssistantMessageParam(
+                    role="assistant",
+                    audio=None,
+                    content=self.content or "",
+                    function_call=None,
+                    refusal=None,
+                    tool_calls=[t.to_openai_param() for t in self.tool_calls or []],
+                )
+
+            case ChatMessageRole.TOOL:
+                raise NotImplementedError
+
 
 class ChatMessageAssessmentStatus(StrEnum):
     PENDING = "PENDING"

{unique_toolkit-0.8.7 → unique_toolkit-0.8.9}/unique_toolkit/chat/service.py

@@ -329,9 +329,9 @@ class ChatService:
 
         Args:
             content (str): The new content for the message.
-            references (list[ContentReference]): list of ContentReference objects. Defaults to [].
-            debug_info (dict[str, Any]]]): Debug information. Defaults to {}.
-            message_id (str, optional): The message ID. Defaults to None, then the ChatState user message id is used.
+            references (list[ContentReference]): list of ContentReference objects.
+            debug_info (dict[str, Any]]]): Debug information.
+            message_id (str, optional): The message ID, if not specified the last user message is edited.
             set_completed_at (Optional[bool]): Whether to set the completedAt field with the current date time. Defaults to False.
 
         Returns:

{unique_toolkit-0.8.7 → unique_toolkit-0.8.9}/unique_toolkit/language_model/infos.py

@@ -507,7 +507,9 @@ class LanguageModelInfo(BaseModel):
                     version="claude-3-7-sonnet",
                     encoder_name=EncoderName.O200K_BASE,  # TODO: Update encoder with litellm
                     token_limits=LanguageModelTokenLimits(
-                        token_limit_input=200_000, token_limit_output=128_000
+                        # Input limit is 200_000, we leave 20_000 tokens as buffer due to tokenizer mismatch
+                        token_limit_input=180_000,
+                        token_limit_output=128_000,
                     ),
                     info_cutoff_at=date(2024, 10, 31),
                     published_at=date(2025, 2, 24),
@@ -525,7 +527,9 @@ class LanguageModelInfo(BaseModel):
                     version="claude-3-7-sonnet-thinking",
                     encoder_name=EncoderName.O200K_BASE,  # TODO: Update encoder with litellm
                     token_limits=LanguageModelTokenLimits(
-                        token_limit_input=200_000, token_limit_output=128_000
+                        # Input limit is 200_000, we leave 20_000 tokens as buffer due to tokenizer mismatch
+                        token_limit_input=180_000,
+                        token_limit_output=128_000,
                     ),
                     info_cutoff_at=date(2024, 10, 31),
                     published_at=date(2025, 2, 24),
@@ -543,7 +547,9 @@ class LanguageModelInfo(BaseModel):
                     version="claude-sonnet-4",
                     encoder_name=EncoderName.O200K_BASE,  # TODO: Update encoder with litellm
                     token_limits=LanguageModelTokenLimits(
-                        token_limit_input=200_000, token_limit_output=64_000
+                        # Input limit is 200_000, we leave 20_000 tokens as buffer due to tokenizer mismatch
+                        token_limit_input=180_000,
+                        token_limit_output=64_000,
                     ),
                     info_cutoff_at=date(2025, 3, 1),
                     published_at=date(2025, 5, 1),
@@ -561,7 +567,9 @@ class LanguageModelInfo(BaseModel):
                     version="claude-opus-4",
                     encoder_name=EncoderName.O200K_BASE,  # TODO: Update encoder with litellm
                     token_limits=LanguageModelTokenLimits(
-                        token_limit_input=200_000, token_limit_output=32_000
+                        # Input limit is 200_000, we leave 20_000 tokens as buffer due to tokenizer mismatch
+                        token_limit_input=180_000,
+                        token_limit_output=32_000,
                     ),
                     info_cutoff_at=date(2025, 3, 1),
                     published_at=date(2025, 5, 1),

{unique_toolkit-0.8.7 → unique_toolkit-0.8.9}/unique_toolkit/language_model/schemas.py

@@ -5,6 +5,11 @@ from typing import Any, Self
 from uuid import uuid4
 
 from humps import camelize
+from openai.types.chat import ChatCompletionAssistantMessageParam
+from openai.types.chat.chat_completion_message_function_tool_call_param import (
+    ChatCompletionMessageFunctionToolCallParam,
+    Function,
+)
 from openai.types.chat.chat_completion_tool_param import ChatCompletionToolParam
 from openai.types.shared_params.function_definition import FunctionDefinition
 from pydantic import (
@@ -88,9 +93,7 @@ class LanguageModelFunction(BaseModel):
         return seralization
 
     def __eq__(self, other: Self) -> bool:
-        """
-        Compare two tool calls based on name and arguments.
-        """
+        """Compare two tool calls based on name and arguments."""
         if not isinstance(other, LanguageModelFunction):
             return False
 
@@ -105,6 +108,19 @@ class LanguageModelFunction(BaseModel):
 
         return True
 
+    def to_openai_param(self) -> ChatCompletionMessageFunctionToolCallParam:
+        arguments = ""
+        if isinstance(self.arguments, dict):
+            arguments = json.dumps(self.arguments)
+        elif isinstance(self.arguments, str):
+            arguments = self.arguments
+
+        return ChatCompletionMessageFunctionToolCallParam(
+            type="function",
+            id=self.id or "unknown_id",
+            function=Function(name=self.name, arguments=arguments),
+        )
+
 
 # This is tailored to the unique backend
 class LanguageModelStreamResponse(BaseModel):
@@ -113,6 +129,16 @@ class LanguageModelStreamResponse(BaseModel):
     message: LanguageModelStreamResponseMessage
     tool_calls: list[LanguageModelFunction] | None = None
 
+    def to_openai_param(self) -> ChatCompletionAssistantMessageParam:
+        return ChatCompletionAssistantMessageParam(
+            role="assistant",
+            audio=None,
+            content=self.message.text,
+            function_call=None,
+            refusal=None,
+            tool_calls=[t.to_openai_param() for t in self.tool_calls or []],
+        )
+
 
 class LanguageModelFunctionCall(BaseModel):
     model_config = model_config
@@ -126,7 +152,7 @@ class LanguageModelFunctionCall(BaseModel):
     @staticmethod
     def create_assistant_message_from_tool_calls(
         tool_calls: list[LanguageModelFunction],
-    ):
+    ) -> "LanguageModelAssistantMessage":
         assistant_message = LanguageModelAssistantMessage(
             content="",
             tool_calls=[