Merge pull request #16 from THUDM/feature/glm-4.5

feat: Feature/glm 4.5

pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "zai-sdk"
-version = "0.0.1b2"
+version = "0.0.1b3"
 description = "A SDK library for accessing big model apis from Z.ai"
 authors = ["Z.ai"]
 readme = "README.md"

src/zai/api_resource/chat/async_completions.py

@@ -56,6 +56,8 @@ class AsyncCompletions(BaseAPI):
 		extra_headers: Headers | None = None,
 		extra_body: Body | None = None,
 		timeout: float | httpx.Timeout | None | NotGiven = NOT_GIVEN,
+		response_format: object | None = None,
+		thinking: object | None = None,
 	) -> AsyncTaskStatus:
 		"""
 		Create an asynchronous chat completion task
@@ -79,6 +81,8 @@ class AsyncCompletions(BaseAPI):
 			extra_headers (Headers): Additional HTTP headers
 			extra_body (Body): Additional request body parameters
 			timeout (float | httpx.Timeout): Request timeout
+			response_format (Optional[object]): Response format specification
+			thinking (Optional[object]): Configuration parameters for model reasoning
 		"""
 		_cast_type = AsyncTaskStatus
 		logger.debug(f'temperature:{temperature}, top_p:{top_p}')
@@ -121,6 +125,8 @@ class AsyncCompletions(BaseAPI):
 			'tool_choice': tool_choice,
 			'meta': meta,
 			'extra': maybe_transform(extra, code_geex_params.CodeGeexExtra),
+			"response_format": response_format,
+			"thinking": thinking
 		}
 		return self._post(
 			'/async/chat/completions',

src/zai/api_resource/chat/completions.py

@@ -62,6 +62,7 @@ class Completions(BaseAPI):
 		extra_body: Body | None = None,
 		timeout: float | httpx.Timeout | None | NotGiven = NOT_GIVEN,
 		response_format: object | None = None,
+		thinking: object | None = None,
 	) -> Completion | StreamResponse[ChatCompletionChunk]:
 		"""
 		Create a chat completion
@@ -87,6 +88,7 @@ class Completions(BaseAPI):
 			extra_body (Body): Additional request body parameters
 			timeout (float | httpx.Timeout): Request timeout
 			response_format (object): Response format specification
+			thinking (Optional[object]): Configuration parameters for model reasoning
 		"""
 		logger.debug(f'temperature:{temperature}, top_p:{top_p}')
 		if temperature is not None and temperature != NOT_GIVEN:
@@ -131,6 +133,7 @@ class Completions(BaseAPI):
 				'meta': meta,
 				'extra': maybe_transform(extra, code_geex_params.CodeGeexExtra),
 				'response_format': response_format,
+				"thinking": thinking
 			}
 		)
 		return self._post(

src/zai/types/chat/chat_completion.py

@@ -45,17 +45,39 @@ class CompletionMessage(BaseModel):
 	tool_calls: Optional[List[CompletionMessageToolCall]] = None
 
 
+class PromptTokensDetails(BaseModel):
+	"""
+	Detailed breakdown of token usage for the input prompt
+
+	Attributes:
+		cached_tokens: Number of tokens reused from cache
+	"""
+	cached_tokens: int
+
+class CompletionTokensDetails(BaseModel):
+	"""
+	Detailed breakdown of token usage for the model completion
+
+	Attributes:
+	    reasoning_tokens: Number of tokens used for reasoning steps
+	"""
+	reasoning_tokens: int
+
 class CompletionUsage(BaseModel):
 	"""
 	Token usage information for completion
 	
 	Attributes:
 		prompt_tokens: Number of tokens in the prompt
+		prompt_tokens_details: Detailed breakdown of token usage for the input prompt
 		completion_tokens: Number of tokens in the completion
+		completion_tokens_details: Detailed breakdown of token usage for the model completion
 		total_tokens: Total number of tokens used
 	"""
 	prompt_tokens: int
+	prompt_tokens_details: Optional[PromptTokensDetails] = None
 	completion_tokens: int
+	completion_tokens_details: Optional[CompletionTokensDetails] = None
 	total_tokens: int