"""Code generated by Speakeasy (https://speakeasy.com). DO NOT EDIT.""" from .basesdk import BaseSDK from mistralai import models, utils from mistralai._hooks import HookContext from mistralai.models import ( agentscompletionrequest as models_agentscompletionrequest, agentscompletionstreamrequest as models_agentscompletionstreamrequest, mistralpromptmode as models_mistralpromptmode, prediction as models_prediction, responseformat as models_responseformat, tool as models_tool, ) from mistralai.types import OptionalNullable, UNSET from mistralai.utils import eventstreaming, get_security_from_env from mistralai.utils.unmarshal_json_response import unmarshal_json_response from typing import Any, Dict, List, Mapping, Optional, Union class Agents(BaseSDK): r"""Agents API.""" def complete( self, *, messages: Union[ List[models_agentscompletionrequest.AgentsCompletionRequestMessages], List[ models_agentscompletionrequest.AgentsCompletionRequestMessagesTypedDict ], ], agent_id: str, max_tokens: OptionalNullable[int] = UNSET, stream: Optional[bool] = False, stop: Optional[ Union[ models_agentscompletionrequest.AgentsCompletionRequestStop, models_agentscompletionrequest.AgentsCompletionRequestStopTypedDict, ] ] = None, random_seed: OptionalNullable[int] = UNSET, metadata: OptionalNullable[Dict[str, Any]] = UNSET, response_format: Optional[ Union[ models_responseformat.ResponseFormat, models_responseformat.ResponseFormatTypedDict, ] ] = None, tools: OptionalNullable[ Union[List[models_tool.Tool], List[models_tool.ToolTypedDict]] ] = UNSET, tool_choice: Optional[ Union[ models_agentscompletionrequest.AgentsCompletionRequestToolChoice, models_agentscompletionrequest.AgentsCompletionRequestToolChoiceTypedDict, ] ] = None, presence_penalty: Optional[float] = None, frequency_penalty: Optional[float] = None, n: OptionalNullable[int] = UNSET, prediction: Optional[ Union[models_prediction.Prediction, models_prediction.PredictionTypedDict] ] = None, parallel_tool_calls: Optional[bool] = None, prompt_mode: OptionalNullable[ models_mistralpromptmode.MistralPromptMode ] = UNSET, retries: OptionalNullable[utils.RetryConfig] = UNSET, server_url: Optional[str] = None, timeout_ms: Optional[int] = None, http_headers: Optional[Mapping[str, str]] = None, ) -> models.ChatCompletionResponse: r"""Agents Completion :param messages: The prompt(s) to generate completions for, encoded as a list of dict with role and content. :param agent_id: The ID of the agent to use for this completion. :param max_tokens: The maximum number of tokens to generate in the completion. The token count of your prompt plus `max_tokens` cannot exceed the model's context length. :param stream: Whether to stream back partial progress. If set, tokens will be sent as data-only server-side events as they become available, with the stream terminated by a data: [DONE] message. Otherwise, the server will hold the request open until the timeout or until completion, with the response containing the full result as JSON. :param stop: Stop generation if this token is detected. Or if one of these tokens is detected when providing an array :param random_seed: The seed to use for random sampling. If set, different calls will generate deterministic results. :param metadata: :param response_format: Specify the format that the model must output. By default it will use `{ \"type\": \"text\" }`. Setting to `{ \"type\": \"json_object\" }` enables JSON mode, which guarantees the message the model generates is in JSON. When using JSON mode you MUST also instruct the model to produce JSON yourself with a system or a user message. Setting to `{ \"type\": \"json_schema\" }` enables JSON schema mode, which guarantees the message the model generates is in JSON and follows the schema you provide. :param tools: :param tool_choice: :param presence_penalty: The `presence_penalty` determines how much the model penalizes the repetition of words or phrases. A higher presence penalty encourages the model to use a wider variety of words and phrases, making the output more diverse and creative. :param frequency_penalty: The `frequency_penalty` penalizes the repetition of words based on their frequency in the generated text. A higher frequency penalty discourages the model from repeating words that have already appeared frequently in the output, promoting diversity and reducing repetition. :param n: Number of completions to return for each request, input tokens are only billed once. :param prediction: Enable users to specify an expected completion, optimizing response times by leveraging known or predictable content. :param parallel_tool_calls: :param prompt_mode: Allows toggling between the reasoning mode and no system prompt. When set to `reasoning` the system prompt for reasoning models will be used. :param retries: Override the default retry configuration for this method :param server_url: Override the default server URL for this method :param timeout_ms: Override the default request timeout configuration for this method in milliseconds :param http_headers: Additional headers to set or replace on requests. """ base_url = None url_variables = None if timeout_ms is None: timeout_ms = self.sdk_configuration.timeout_ms if server_url is not None: base_url = server_url else: base_url = self._get_url(base_url, url_variables) request = models.AgentsCompletionRequest( max_tokens=max_tokens, stream=stream, stop=stop, random_seed=random_seed, metadata=metadata, messages=utils.get_pydantic_model( messages, List[models.AgentsCompletionRequestMessages] ), response_format=utils.get_pydantic_model( response_format, Optional[models.ResponseFormat] ), tools=utils.get_pydantic_model(tools, OptionalNullable[List[models.Tool]]), tool_choice=utils.get_pydantic_model( tool_choice, Optional[models.AgentsCompletionRequestToolChoice] ), presence_penalty=presence_penalty, frequency_penalty=frequency_penalty, n=n, prediction=utils.get_pydantic_model( prediction, Optional[models.Prediction] ), parallel_tool_calls=parallel_tool_calls, prompt_mode=prompt_mode, agent_id=agent_id, ) req = self._build_request( method="POST", path="/v1/agents/completions", base_url=base_url, url_variables=url_variables, request=request, request_body_required=True, request_has_path_params=False, request_has_query_params=True, user_agent_header="user-agent", accept_header_value="application/json", http_headers=http_headers, security=self.sdk_configuration.security, get_serialized_body=lambda: utils.serialize_request_body( request, False, False, "json", models.AgentsCompletionRequest ), allow_empty_value=None, timeout_ms=timeout_ms, ) if retries == UNSET: if self.sdk_configuration.retry_config is not UNSET: retries = self.sdk_configuration.retry_config retry_config = None if isinstance(retries, utils.RetryConfig): retry_config = (retries, ["429", "500", "502", "503", "504"]) http_res = self.do_request( hook_ctx=HookContext( config=self.sdk_configuration, base_url=base_url or "", operation_id="agents_completion_v1_agents_completions_post", oauth2_scopes=None, security_source=get_security_from_env( self.sdk_configuration.security, models.Security ), ), request=req, error_status_codes=["422", "4XX", "5XX"], retry_config=retry_config, ) response_data: Any = None if utils.match_response(http_res, "200", "application/json"): return unmarshal_json_response(models.ChatCompletionResponse, http_res) if utils.match_response(http_res, "422", "application/json"): response_data = unmarshal_json_response( models.HTTPValidationErrorData, http_res ) raise models.HTTPValidationError(response_data, http_res) if utils.match_response(http_res, "4XX", "*"): http_res_text = utils.stream_to_text(http_res) raise models.SDKError("API error occurred", http_res, http_res_text) if utils.match_response(http_res, "5XX", "*"): http_res_text = utils.stream_to_text(http_res) raise models.SDKError("API error occurred", http_res, http_res_text) raise models.SDKError("Unexpected response received", http_res) async def complete_async( self, *, messages: Union[ List[models_agentscompletionrequest.AgentsCompletionRequestMessages], List[ models_agentscompletionrequest.AgentsCompletionRequestMessagesTypedDict ], ], agent_id: str, max_tokens: OptionalNullable[int] = UNSET, stream: Optional[bool] = False, stop: Optional[ Union[ models_agentscompletionrequest.AgentsCompletionRequestStop, models_agentscompletionrequest.AgentsCompletionRequestStopTypedDict, ] ] = None, random_seed: OptionalNullable[int] = UNSET, metadata: OptionalNullable[Dict[str, Any]] = UNSET, response_format: Optional[ Union[ models_responseformat.ResponseFormat, models_responseformat.ResponseFormatTypedDict, ] ] = None, tools: OptionalNullable[ Union[List[models_tool.Tool], List[models_tool.ToolTypedDict]] ] = UNSET, tool_choice: Optional[ Union[ models_agentscompletionrequest.AgentsCompletionRequestToolChoice, models_agentscompletionrequest.AgentsCompletionRequestToolChoiceTypedDict, ] ] = None, presence_penalty: Optional[float] = None, frequency_penalty: Optional[float] = None, n: OptionalNullable[int] = UNSET, prediction: Optional[ Union[models_prediction.Prediction, models_prediction.PredictionTypedDict] ] = None, parallel_tool_calls: Optional[bool] = None, prompt_mode: OptionalNullable[ models_mistralpromptmode.MistralPromptMode ] = UNSET, retries: OptionalNullable[utils.RetryConfig] = UNSET, server_url: Optional[str] = None, timeout_ms: Optional[int] = None, http_headers: Optional[Mapping[str, str]] = None, ) -> models.ChatCompletionResponse: r"""Agents Completion :param messages: The prompt(s) to generate completions for, encoded as a list of dict with role and content. :param agent_id: The ID of the agent to use for this completion. :param max_tokens: The maximum number of tokens to generate in the completion. The token count of your prompt plus `max_tokens` cannot exceed the model's context length. :param stream: Whether to stream back partial progress. If set, tokens will be sent as data-only server-side events as they become available, with the stream terminated by a data: [DONE] message. Otherwise, the server will hold the request open until the timeout or until completion, with the response containing the full result as JSON. :param stop: Stop generation if this token is detected. Or if one of these tokens is detected when providing an array :param random_seed: The seed to use for random sampling. If set, different calls will generate deterministic results. :param metadata: :param response_format: Specify the format that the model must output. By default it will use `{ \"type\": \"text\" }`. Setting to `{ \"type\": \"json_object\" }` enables JSON mode, which guarantees the message the model generates is in JSON. When using JSON mode you MUST also instruct the model to produce JSON yourself with a system or a user message. Setting to `{ \"type\": \"json_schema\" }` enables JSON schema mode, which guarantees the message the model generates is in JSON and follows the schema you provide. :param tools: :param tool_choice: :param presence_penalty: The `presence_penalty` determines how much the model penalizes the repetition of words or phrases. A higher presence penalty encourages the model to use a wider variety of words and phrases, making the output more diverse and creative. :param frequency_penalty: The `frequency_penalty` penalizes the repetition of words based on their frequency in the generated text. A higher frequency penalty discourages the model from repeating words that have already appeared frequently in the output, promoting diversity and reducing repetition. :param n: Number of completions to return for each request, input tokens are only billed once. :param prediction: Enable users to specify an expected completion, optimizing response times by leveraging known or predictable content. :param parallel_tool_calls: :param prompt_mode: Allows toggling between the reasoning mode and no system prompt. When set to `reasoning` the system prompt for reasoning models will be used. :param retries: Override the default retry configuration for this method :param server_url: Override the default server URL for this method :param timeout_ms: Override the default request timeout configuration for this method in milliseconds :param http_headers: Additional headers to set or replace on requests. """ base_url = None url_variables = None if timeout_ms is None: timeout_ms = self.sdk_configuration.timeout_ms if server_url is not None: base_url = server_url else: base_url = self._get_url(base_url, url_variables) request = models.AgentsCompletionRequest( max_tokens=max_tokens, stream=stream, stop=stop, random_seed=random_seed, metadata=metadata, messages=utils.get_pydantic_model( messages, List[models.AgentsCompletionRequestMessages] ), response_format=utils.get_pydantic_model( response_format, Optional[models.ResponseFormat] ), tools=utils.get_pydantic_model(tools, OptionalNullable[List[models.Tool]]), tool_choice=utils.get_pydantic_model( tool_choice, Optional[models.AgentsCompletionRequestToolChoice] ), presence_penalty=presence_penalty, frequency_penalty=frequency_penalty, n=n, prediction=utils.get_pydantic_model( prediction, Optional[models.Prediction] ), parallel_tool_calls=parallel_tool_calls, prompt_mode=prompt_mode, agent_id=agent_id, ) req = self._build_request_async( method="POST", path="/v1/agents/completions", base_url=base_url, url_variables=url_variables, request=request, request_body_required=True, request_has_path_params=False, request_has_query_params=True, user_agent_header="user-agent", accept_header_value="application/json", http_headers=http_headers, security=self.sdk_configuration.security, get_serialized_body=lambda: utils.serialize_request_body( request, False, False, "json", models.AgentsCompletionRequest ), allow_empty_value=None, timeout_ms=timeout_ms, ) if retries == UNSET: if self.sdk_configuration.retry_config is not UNSET: retries = self.sdk_configuration.retry_config retry_config = None if isinstance(retries, utils.RetryConfig): retry_config = (retries, ["429", "500", "502", "503", "504"]) http_res = await self.do_request_async( hook_ctx=HookContext( config=self.sdk_configuration, base_url=base_url or "", operation_id="agents_completion_v1_agents_completions_post", oauth2_scopes=None, security_source=get_security_from_env( self.sdk_configuration.security, models.Security ), ), request=req, error_status_codes=["422", "4XX", "5XX"], retry_config=retry_config, ) response_data: Any = None if utils.match_response(http_res, "200", "application/json"): return unmarshal_json_response(models.ChatCompletionResponse, http_res) if utils.match_response(http_res, "422", "application/json"): response_data = unmarshal_json_response( models.HTTPValidationErrorData, http_res ) raise models.HTTPValidationError(response_data, http_res) if utils.match_response(http_res, "4XX", "*"): http_res_text = await utils.stream_to_text_async(http_res) raise models.SDKError("API error occurred", http_res, http_res_text) if utils.match_response(http_res, "5XX", "*"): http_res_text = await utils.stream_to_text_async(http_res) raise models.SDKError("API error occurred", http_res, http_res_text) raise models.SDKError("Unexpected response received", http_res) def stream( self, *, messages: Union[ List[ models_agentscompletionstreamrequest.AgentsCompletionStreamRequestMessages ], List[ models_agentscompletionstreamrequest.AgentsCompletionStreamRequestMessagesTypedDict ], ], agent_id: str, max_tokens: OptionalNullable[int] = UNSET, stream: Optional[bool] = True, stop: Optional[ Union[ models_agentscompletionstreamrequest.AgentsCompletionStreamRequestStop, models_agentscompletionstreamrequest.AgentsCompletionStreamRequestStopTypedDict, ] ] = None, random_seed: OptionalNullable[int] = UNSET, metadata: OptionalNullable[Dict[str, Any]] = UNSET, response_format: Optional[ Union[ models_responseformat.ResponseFormat, models_responseformat.ResponseFormatTypedDict, ] ] = None, tools: OptionalNullable[ Union[List[models_tool.Tool], List[models_tool.ToolTypedDict]] ] = UNSET, tool_choice: Optional[ Union[ models_agentscompletionstreamrequest.AgentsCompletionStreamRequestToolChoice, models_agentscompletionstreamrequest.AgentsCompletionStreamRequestToolChoiceTypedDict, ] ] = None, presence_penalty: Optional[float] = None, frequency_penalty: Optional[float] = None, n: OptionalNullable[int] = UNSET, prediction: Optional[ Union[models_prediction.Prediction, models_prediction.PredictionTypedDict] ] = None, parallel_tool_calls: Optional[bool] = None, prompt_mode: OptionalNullable[ models_mistralpromptmode.MistralPromptMode ] = UNSET, retries: OptionalNullable[utils.RetryConfig] = UNSET, server_url: Optional[str] = None, timeout_ms: Optional[int] = None, http_headers: Optional[Mapping[str, str]] = None, ) -> eventstreaming.EventStream[models.CompletionEvent]: r"""Stream Agents completion Mistral AI provides the ability to stream responses back to a client in order to allow partial results for certain requests. Tokens will be sent as data-only server-sent events as they become available, with the stream terminated by a data: [DONE] message. Otherwise, the server will hold the request open until the timeout or until completion, with the response containing the full result as JSON. :param messages: The prompt(s) to generate completions for, encoded as a list of dict with role and content. :param agent_id: The ID of the agent to use for this completion. :param max_tokens: The maximum number of tokens to generate in the completion. The token count of your prompt plus `max_tokens` cannot exceed the model's context length. :param stream: :param stop: Stop generation if this token is detected. Or if one of these tokens is detected when providing an array :param random_seed: The seed to use for random sampling. If set, different calls will generate deterministic results. :param metadata: :param response_format: Specify the format that the model must output. By default it will use `{ \"type\": \"text\" }`. Setting to `{ \"type\": \"json_object\" }` enables JSON mode, which guarantees the message the model generates is in JSON. When using JSON mode you MUST also instruct the model to produce JSON yourself with a system or a user message. Setting to `{ \"type\": \"json_schema\" }` enables JSON schema mode, which guarantees the message the model generates is in JSON and follows the schema you provide. :param tools: :param tool_choice: :param presence_penalty: The `presence_penalty` determines how much the model penalizes the repetition of words or phrases. A higher presence penalty encourages the model to use a wider variety of words and phrases, making the output more diverse and creative. :param frequency_penalty: The `frequency_penalty` penalizes the repetition of words based on their frequency in the generated text. A higher frequency penalty discourages the model from repeating words that have already appeared frequently in the output, promoting diversity and reducing repetition. :param n: Number of completions to return for each request, input tokens are only billed once. :param prediction: Enable users to specify an expected completion, optimizing response times by leveraging known or predictable content. :param parallel_tool_calls: :param prompt_mode: Allows toggling between the reasoning mode and no system prompt. When set to `reasoning` the system prompt for reasoning models will be used. :param retries: Override the default retry configuration for this method :param server_url: Override the default server URL for this method :param timeout_ms: Override the default request timeout configuration for this method in milliseconds :param http_headers: Additional headers to set or replace on requests. """ base_url = None url_variables = None if timeout_ms is None: timeout_ms = self.sdk_configuration.timeout_ms if server_url is not None: base_url = server_url else: base_url = self._get_url(base_url, url_variables) request = models.AgentsCompletionStreamRequest( max_tokens=max_tokens, stream=stream, stop=stop, random_seed=random_seed, metadata=metadata, messages=utils.get_pydantic_model( messages, List[models.AgentsCompletionStreamRequestMessages] ), response_format=utils.get_pydantic_model( response_format, Optional[models.ResponseFormat] ), tools=utils.get_pydantic_model(tools, OptionalNullable[List[models.Tool]]), tool_choice=utils.get_pydantic_model( tool_choice, Optional[models.AgentsCompletionStreamRequestToolChoice] ), presence_penalty=presence_penalty, frequency_penalty=frequency_penalty, n=n, prediction=utils.get_pydantic_model( prediction, Optional[models.Prediction] ), parallel_tool_calls=parallel_tool_calls, prompt_mode=prompt_mode, agent_id=agent_id, ) req = self._build_request( method="POST", path="/v1/agents/completions#stream", base_url=base_url, url_variables=url_variables, request=request, request_body_required=True, request_has_path_params=False, request_has_query_params=True, user_agent_header="user-agent", accept_header_value="text/event-stream", http_headers=http_headers, security=self.sdk_configuration.security, get_serialized_body=lambda: utils.serialize_request_body( request, False, False, "json", models.AgentsCompletionStreamRequest ), allow_empty_value=None, timeout_ms=timeout_ms, ) if retries == UNSET: if self.sdk_configuration.retry_config is not UNSET: retries = self.sdk_configuration.retry_config retry_config = None if isinstance(retries, utils.RetryConfig): retry_config = (retries, ["429", "500", "502", "503", "504"]) http_res = self.do_request( hook_ctx=HookContext( config=self.sdk_configuration, base_url=base_url or "", operation_id="stream_agents", oauth2_scopes=None, security_source=get_security_from_env( self.sdk_configuration.security, models.Security ), ), request=req, error_status_codes=["422", "4XX", "5XX"], stream=True, retry_config=retry_config, ) response_data: Any = None if utils.match_response(http_res, "200", "text/event-stream"): return eventstreaming.EventStream( http_res, lambda raw: utils.unmarshal_json(raw, models.CompletionEvent), sentinel="[DONE]", client_ref=self, ) if utils.match_response(http_res, "422", "application/json"): http_res_text = utils.stream_to_text(http_res) response_data = unmarshal_json_response( models.HTTPValidationErrorData, http_res, http_res_text ) raise models.HTTPValidationError(response_data, http_res, http_res_text) if utils.match_response(http_res, "4XX", "*"): http_res_text = utils.stream_to_text(http_res) raise models.SDKError("API error occurred", http_res, http_res_text) if utils.match_response(http_res, "5XX", "*"): http_res_text = utils.stream_to_text(http_res) raise models.SDKError("API error occurred", http_res, http_res_text) http_res_text = utils.stream_to_text(http_res) raise models.SDKError("Unexpected response received", http_res, http_res_text) async def stream_async( self, *, messages: Union[ List[ models_agentscompletionstreamrequest.AgentsCompletionStreamRequestMessages ], List[ models_agentscompletionstreamrequest.AgentsCompletionStreamRequestMessagesTypedDict ], ], agent_id: str, max_tokens: OptionalNullable[int] = UNSET, stream: Optional[bool] = True, stop: Optional[ Union[ models_agentscompletionstreamrequest.AgentsCompletionStreamRequestStop, models_agentscompletionstreamrequest.AgentsCompletionStreamRequestStopTypedDict, ] ] = None, random_seed: OptionalNullable[int] = UNSET, metadata: OptionalNullable[Dict[str, Any]] = UNSET, response_format: Optional[ Union[ models_responseformat.ResponseFormat, models_responseformat.ResponseFormatTypedDict, ] ] = None, tools: OptionalNullable[ Union[List[models_tool.Tool], List[models_tool.ToolTypedDict]] ] = UNSET, tool_choice: Optional[ Union[ models_agentscompletionstreamrequest.AgentsCompletionStreamRequestToolChoice, models_agentscompletionstreamrequest.AgentsCompletionStreamRequestToolChoiceTypedDict, ] ] = None, presence_penalty: Optional[float] = None, frequency_penalty: Optional[float] = None, n: OptionalNullable[int] = UNSET, prediction: Optional[ Union[models_prediction.Prediction, models_prediction.PredictionTypedDict] ] = None, parallel_tool_calls: Optional[bool] = None, prompt_mode: OptionalNullable[ models_mistralpromptmode.MistralPromptMode ] = UNSET, retries: OptionalNullable[utils.RetryConfig] = UNSET, server_url: Optional[str] = None, timeout_ms: Optional[int] = None, http_headers: Optional[Mapping[str, str]] = None, ) -> eventstreaming.EventStreamAsync[models.CompletionEvent]: r"""Stream Agents completion Mistral AI provides the ability to stream responses back to a client in order to allow partial results for certain requests. Tokens will be sent as data-only server-sent events as they become available, with the stream terminated by a data: [DONE] message. Otherwise, the server will hold the request open until the timeout or until completion, with the response containing the full result as JSON. :param messages: The prompt(s) to generate completions for, encoded as a list of dict with role and content. :param agent_id: The ID of the agent to use for this completion. :param max_tokens: The maximum number of tokens to generate in the completion. The token count of your prompt plus `max_tokens` cannot exceed the model's context length. :param stream: :param stop: Stop generation if this token is detected. Or if one of these tokens is detected when providing an array :param random_seed: The seed to use for random sampling. If set, different calls will generate deterministic results. :param metadata: :param response_format: Specify the format that the model must output. By default it will use `{ \"type\": \"text\" }`. Setting to `{ \"type\": \"json_object\" }` enables JSON mode, which guarantees the message the model generates is in JSON. When using JSON mode you MUST also instruct the model to produce JSON yourself with a system or a user message. Setting to `{ \"type\": \"json_schema\" }` enables JSON schema mode, which guarantees the message the model generates is in JSON and follows the schema you provide. :param tools: :param tool_choice: :param presence_penalty: The `presence_penalty` determines how much the model penalizes the repetition of words or phrases. A higher presence penalty encourages the model to use a wider variety of words and phrases, making the output more diverse and creative. :param frequency_penalty: The `frequency_penalty` penalizes the repetition of words based on their frequency in the generated text. A higher frequency penalty discourages the model from repeating words that have already appeared frequently in the output, promoting diversity and reducing repetition. :param n: Number of completions to return for each request, input tokens are only billed once. :param prediction: Enable users to specify an expected completion, optimizing response times by leveraging known or predictable content. :param parallel_tool_calls: :param prompt_mode: Allows toggling between the reasoning mode and no system prompt. When set to `reasoning` the system prompt for reasoning models will be used. :param retries: Override the default retry configuration for this method :param server_url: Override the default server URL for this method :param timeout_ms: Override the default request timeout configuration for this method in milliseconds :param http_headers: Additional headers to set or replace on requests. """ base_url = None url_variables = None if timeout_ms is None: timeout_ms = self.sdk_configuration.timeout_ms if server_url is not None: base_url = server_url else: base_url = self._get_url(base_url, url_variables) request = models.AgentsCompletionStreamRequest( max_tokens=max_tokens, stream=stream, stop=stop, random_seed=random_seed, metadata=metadata, messages=utils.get_pydantic_model( messages, List[models.AgentsCompletionStreamRequestMessages] ), response_format=utils.get_pydantic_model( response_format, Optional[models.ResponseFormat] ), tools=utils.get_pydantic_model(tools, OptionalNullable[List[models.Tool]]), tool_choice=utils.get_pydantic_model( tool_choice, Optional[models.AgentsCompletionStreamRequestToolChoice] ), presence_penalty=presence_penalty, frequency_penalty=frequency_penalty, n=n, prediction=utils.get_pydantic_model( prediction, Optional[models.Prediction] ), parallel_tool_calls=parallel_tool_calls, prompt_mode=prompt_mode, agent_id=agent_id, ) req = self._build_request_async( method="POST", path="/v1/agents/completions#stream", base_url=base_url, url_variables=url_variables, request=request, request_body_required=True, request_has_path_params=False, request_has_query_params=True, user_agent_header="user-agent", accept_header_value="text/event-stream", http_headers=http_headers, security=self.sdk_configuration.security, get_serialized_body=lambda: utils.serialize_request_body( request, False, False, "json", models.AgentsCompletionStreamRequest ), allow_empty_value=None, timeout_ms=timeout_ms, ) if retries == UNSET: if self.sdk_configuration.retry_config is not UNSET: retries = self.sdk_configuration.retry_config retry_config = None if isinstance(retries, utils.RetryConfig): retry_config = (retries, ["429", "500", "502", "503", "504"]) http_res = await self.do_request_async( hook_ctx=HookContext( config=self.sdk_configuration, base_url=base_url or "", operation_id="stream_agents", oauth2_scopes=None, security_source=get_security_from_env( self.sdk_configuration.security, models.Security ), ), request=req, error_status_codes=["422", "4XX", "5XX"], stream=True, retry_config=retry_config, ) response_data: Any = None if utils.match_response(http_res, "200", "text/event-stream"): return eventstreaming.EventStreamAsync( http_res, lambda raw: utils.unmarshal_json(raw, models.CompletionEvent), sentinel="[DONE]", client_ref=self, ) if utils.match_response(http_res, "422", "application/json"): http_res_text = await utils.stream_to_text_async(http_res) response_data = unmarshal_json_response( models.HTTPValidationErrorData, http_res, http_res_text ) raise models.HTTPValidationError(response_data, http_res, http_res_text) if utils.match_response(http_res, "4XX", "*"): http_res_text = await utils.stream_to_text_async(http_res) raise models.SDKError("API error occurred", http_res, http_res_text) if utils.match_response(http_res, "5XX", "*"): http_res_text = await utils.stream_to_text_async(http_res) raise models.SDKError("API error occurred", http_res, http_res_text) http_res_text = await utils.stream_to_text_async(http_res) raise models.SDKError("Unexpected response received", http_res, http_res_text)