fix(deps): update dependency mistralai to v3.4.0 #81
Loading…
Add table
Add a link
Reference in a new issue
No description provided.
Delete branch "renovate/mistralai-3.x"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
This PR contains the following updates:
==3.3.0→==3.4.0Release Notes
mistralai/client-python (mistralai)
v3.4.0Generated SDK baseline differences
The published SDK or generator baseline differs from the baseline used by the OpenAPI changelog.
Generated surface details
Stable compatibility changes
Now required on an existing stable model (1):
AudioAgentCreateRequest.configurationStable type changes
Changed types or signatures on the stable surface (4):
AudioAgentCreateRequest.configuration: OptionalNullable[AudioAgentConfigurationRequest] -> AudioAgentConfigurationRequestAudioAgentCreateRequestTypedDict.configuration: NotRequired[Nullable[AudioAgentConfigurationRequestTypedDict]] -> AudioAgentConfigurationRequestTypedDictOCRTableObject.format_: Annotated[Format, pydantic.Field(alias='format')] -> Annotated[FormatEnum, pydantic.Field(alias='format')]OCRTableObjectTypedDict.format_: Format -> FormatEnumBeta surface changes
Changed types or signatures in beta (2):
PublicAuthenticationMethod.has_default_credentials: bool -> Annotated[Optional[bool], pydantic.Field(deprecated='warning: ** DEPRECATED ** - This will be removed in a future release, please migrate away from it as soon as possible.')]PublicAuthenticationMethodTypedDict.has_default_credentials: bool -> NotRequired[bool]API changes
Breaking changes
sdk.audio.agents.create()now requiresconfiguration, amodels.AudioAgentConfigurationRequest. It can no longer be omitted or passed asNone.models.Format, the type ofOCRTableObject.format_, is renamed tomodels.FormatEnum. Imports ofFormatfrommistralai.client.modelsmust switch toFormatEnum.New: Responses API
sdk.responses.create()generates a model response from text or image input. The output can contain text, reasoning and function calls. It is stateless: for multi-turn conversations, send the full history ininput, including earlier output items and function results.modelandinput, either a string or a list of messages, output messages, function calls, function call outputs and reasoning items.instructions,tools(function and namespace tools; only client-executed function tools are supported),tool_choice("none","auto"(default),"required"or a specific function),parallel_tool_calls(defaultTrue),reasoning,text(outputformatandverbosity),max_output_tokens,max_tool_calls,temperature,top_p,metadataandprompt_cache_key.truncation="disabled",background=False,store=Falseandtop_logprobs=0.previous_response_id,conversation,promptandmoderationmust beNone. Responses are not stored.stream=Truereturns anEventStream[models.ResponsesStreamFrame]instead of amodels.CreateResponseResponse.models.CreateResponseResponseincludesid,status,output,usage,error,incomplete_details,service_tier("auto","default","flex","scale"or"priority"),truncation("auto"or"disabled") andprompt_cache_retention("in_memory"or"24h").errorand in error stream frames, are amodels.ResponsesErrorwithmessageandcode(models.Code):server_error,rate_limit_exceeded,invalid_prompt,bio_policy,vector_store_timeout, or one of 14 image errors:invalid_image,invalid_image_format,invalid_base64_image,invalid_image_url,image_too_large,image_too_small,image_parse_error,image_content_policy_violation,invalid_image_mode,image_file_too_large,unsupported_image_media_type,empty_image_file,failed_to_download_imageandimage_file_not_found.statusis"completed","failed","in_progress","cancelled","queued"or"incomplete", andincomplete_details.reason(models.Reason) is"max_output_tokens"or"content_filter".sdk.responses.count_input_tokens()counts the input tokens of a Responses request without generating a response, including the conversation, instructions and tool definitions. It takes the samemodel,input,instructions,tools,tool_choice,parallel_tool_calls,reasoningandtextascreate(), and returnsmodels.CountResponseInputTokensResponsewithinput_tokens.New fields
sdk.beta.rag.managed_indexes.search(), alone or inside anrrfretriever, take anoperator(models.Operator)."or", the default, matches chunks containing any of the query's words."and"matches only chunks containing all of them, the previous behaviour.models.AudioTTSConfig, thettsconfiguration ofsdk.audio.agents.create(),update()andsdk.audio.sessions.create(), takes an optionalexpressiveness, a number from 1 to 5.Deprecated fields
has_default_credentialson connector authentication methods is deprecated and alwaysfalse: connector-level default credentials were removed. It is returned bysdk.beta.connectors.create(),list(),get(),update()andget_authentication_methods(). Responses may now omit it, and the SDK reads a missing value asfalse.Added operations
Stable:
sdk.responses.create()sdk.responses.count_input_tokens()Schema changes
expressivenessinsdk.audio.agents.create(),sdk.audio.sessions.create(), andsdk.audio.agents.update().expressivenessacross audio agents (5 operations).has_default_credentialsacross beta connectors (5 operations).operatorinsdk.beta.rag.managed_indexes.search().ʁ,r,ɑ̃,ʀ,x,y,ɔ̃,ɐ,œ,ɛ̃,ʝ,ɲ,ts,ɥ,ʏ,ø,β,ɦ,pf,ʎ,ɣ,ɳ,ɔʏ,œ̃,ɟʝ,ɬ, andqinsdk.audio.speech.complete().Changes
Based on:
Generated
Releases
Configuration
📅 Schedule: (UTC)
🚦 Automerge: Enabled.
♻ Rebasing: Whenever PR is behind base branch, or you tick the rebase/retry checkbox.
🔕 Ignore: Close this PR and you won't be reminded about this update again.
This PR has been generated by Mend Renovate CLI.