Skip to content

changelog.md

Latest commit

 

History

History
203 lines (156 loc) · 12.1 KB

changelog.md

File metadata and controls

203 lines (156 loc) · 12.1 KB
Raw

3.0.0 - Unreleased

Breaking Changes

This release regenerates the Python SDK from the current Vapi API definition. The migration notes below compare 3.0.0 with the published 1.11.1 package. Client method names and parameters remain available, but public helper imports, model field declarations, and two return annotations change.

Update renamed helper imports

Update imports and type annotations that use these names:

Previous helper Current helper
CartesiaExperimentalControlsSpeedZero CartesiaSpeedControlZero
FallbackAzureVoiceVoiceIdZero FallbackAzureVoiceIdZero
GladiaTranscriberLanguages GladiaTranscriberLanguagesItem
FallbackGladiaTranscriberLanguages FallbackGladiaTranscriberLanguagesItem

VapiVoice.voice_id and FallbackVapiVoice.voice_id now use strings instead of the removed voice-ID helper types. Pass a supported voice ID directly. The OpenAI voice-ID helpers combine the current named voice choices with strings; voice availability depends on the selected model.

Gladia languages fields now use a list of language codes, such as ["en", "fr"]. The new LanguagesItem helpers describe one element of that list, including in embedded assistant, workflow, and fallback transcriber configurations.

Use tagged Vapi request models

For assistant requests using provider="vapi", use the request-specific model type. The standalone VapiModel no longer declares or defaults provider. The tagged request type supplies the discriminator required by the assistant request union.

import os

from vapi import Vapi
from vapi.types.create_assistant_dto_model import CreateAssistantDtoModel_Vapi

client = Vapi(token=os.environ["VAPI_API_KEY"])
assistant = client.assistants.create(
    model=CreateAssistantDtoModel_Vapi(version="latest"),
)
print(assistant.id)

For client.assistants.update(), use UpdateAssistantDtoModel_Vapi from vapi.assistants.types.update_assistant_dto_model. For AssistantOverrides.model, use AssistantOverridesModel_Vapi from vapi.types.assistant_overrides_model.

Handle structured-output run responses

The sync and async structured_output_controller_run() methods return StructuredOutputControllerRunResponse, a union of preview and rerun response models. They no longer declare a StructuredOutput return type.

Use preview_enabled=True to preview one call without updating its artifacts. A rerun response is a StructuredOutputRerunResponse with a message and an optional workflow_id. A preview response contains output-ID keys and an optional skipped field.

import os

from vapi import Vapi
from vapi.types.structured_output_rerun_response import StructuredOutputRerunResponse

client = Vapi(token=os.environ["VAPI_API_KEY"])
result = client.structured_outputs.structured_output_controller_run(
    call_ids=["YOUR_CALL_ID"],
    structured_output_id="YOUR_STRUCTURED_OUTPUT_ID",
    preview_enabled=True,
)
if isinstance(result, StructuredOutputRerunResponse):
    print(result.message, result.workflow_id)
else:
    print(result)

Replace both IDs with a call and saved structured output in your organization. See the Run Structured Output API reference for preview and rerun behavior.

Update simulation request construction

The removed simulation DTO helper classes below are no longer public imports. Supply request values through the corresponding client method's named parameters. For example, client.simulation_suites.simulation_suite_controller_create() accepts name and simulation_ids; client.simulation_scenarios.scenario_controller_create() accepts name, instructions, and evaluations.

Review changed model field declarations

The following fields are no longer declared on the listed models. Recheck code that accesses the Python field name or relies on its type or alias. Response properties outside the declared model fields can still be preserved as extra data under their wire names.

AssistantVersionPaginatedResponse.metadata now uses AssistantVersionPaginatedMetadata, with next_cursor, has_next_page, and limit. Its results are typed as AssistantVersion objects.

When constructing CreateSesameVoiceDto directly, supply file, voice_name, and transcription. file is a new required bytes field; voice_name and transcription are now required strings. SimulationSuite now declares a required target_assignments list. These model declarations do not add required parameters to existing client methods.

Several fields now use specific model types instead of generic dictionaries or strings:

Field Current type
Artifact.transfers Optional list of TransferArtifact objects, rather than strings
Call.phone_number Optional TransientTwilioPhoneNumber, rather than ImportTwilioPhoneNumberDto
Call.transport Optional CallTransport
CreateOutboundCallDto.transport Optional CreateOutboundCallDtoTransport
SayHookAction.exact and embedded say-hook actions Optional SayHookActionExact
CustomerSpeechTimeoutOptions.trigger_reset_mode Optional CustomerSpeechTimeoutOptionsTriggerResetMode
ScorecardMetric.conditions List of ScorecardMetricConditionsItem
RecordingConsent.type RecordingConsentType
VoiceLibraryVoiceResponse.age Optional VoiceLibraryVoiceResponseAge
ToolCallResult.message Optional ToolCallResultSpokenMessage, rather than ToolCallResultMessage
UserMessage.metadata Optional UserMessageMetadata

Update type annotations and code that expects dictionary indexing or string transfer values. ElevenLabsPronunciationDictionaryLocator.version_id, FallbackTranscriberPlan.transcribers, and VapiModel.model now allow None.

Removed declared field Affected models
eager_eot_threshold AssistantOverridesTranscriber_Deepgram, AssistantTranscriber_Deepgram, ConversationNodeTranscriber_Deepgram, CreateAssistantDtoTranscriber_Deepgram, CreateWorkflowDtoTranscriber_Deepgram, DeepgramTranscriber, FallbackDeepgramTranscriber, FallbackTranscriberPlanTranscribersItem_Deepgram, TransferAssistantTranscriber_Deepgram, UpdateAssistantDtoTranscriber_Deepgram, UpdateWorkflowDtoTranscriber_Deepgram, WorkflowTranscriber_Deepgram, WorkflowUserEditableTranscriber_Deepgram
fallback_plan AssistantOverridesVoice_Vapi, AssistantVoice_Vapi, ConversationNodeVoice_Vapi, CreateAssistantDtoVoice_Vapi, CreateWorkflowDtoVoice_Vapi, RecordingConsentPlanStayOnLineVoice_Vapi, RecordingConsentPlanVerbalVoice_Vapi, TransferAssistantVoice_Vapi, UpdateAssistantDtoVoice_Vapi, UpdateWorkflowDtoVoice_Vapi, VapiVoice, WorkflowUserEditableVoice_Vapi, WorkflowVoice_Vapi
max_tokens VapiModel
next_page_state AssistantVersionPaginatedResponse
provider VapiModel
sbc_configuration AssistantCredentialsItem_ByoSipTrunk, AssistantOverridesCredentialsItem_ByoSipTrunk, ByoSipTrunkCredential, CreateAssistantDtoCredentialsItem_ByoSipTrunk, CreateByoSipTrunkCredentialDto, CreateWorkflowDtoCredentialsItem_ByoSipTrunk, UpdateAssistantDtoCredentialsItem_ByoSipTrunk, UpdateByoSipTrunkCredentialDto, UpdateWorkflowDtoCredentialsItem_ByoSipTrunk, WorkflowCredentialsItem_ByoSipTrunk, WorkflowUserEditableCredentialsItem_ByoSipTrunk
slack_channel_id Subscription
slack_support_enabled Subscription
subscription_limits Call
type GhlTool

Trieve credential and knowledge-base types are no longer included in this SDK. This includes CreateTrieveCredentialDto, CreateTrieveKnowledgeBaseDto, UpdateTrieveKnowledgeBaseDto, TrieveKnowledgeBase, and TrieveKnowledgeBaseImport. No compatibility aliases or replacement Trieve models are provided. The current public API definition does not include Trieve schemas. If your integration constructs these types, verify that integration's request format before upgrading.

Removed public imports

The following names were exported by 1.11.1 and are absent from 3.0.0. The helper migrations above cover names with verified alternatives:

  • AssistantCredentialsItem_Trieve
  • AssistantOverridesCredentialsItem_Trieve
  • CartesiaExperimentalControlsSpeedZero
  • CreateAssistantDtoCredentialsItem_Trieve
  • CreateSimulationDto
  • CreateSimulationRunDto
  • CreateSimulationSuiteDto
  • CreateTrieveCredentialDto
  • CreateTrieveKnowledgeBaseDto
  • CreateTrieveKnowledgeBaseDtoProvider
  • CreateWorkflowDtoCredentialsItem_Trieve
  • FallbackAzureVoiceVoiceIdZero
  • FallbackGladiaTranscriberLanguages
  • FallbackVapiVoiceVoiceId
  • GenerateScenariosDto
  • GhlToolType
  • GladiaTranscriberLanguages
  • TrieveCredential
  • TrieveCredentialProvider
  • TrieveKnowledgeBase
  • TrieveKnowledgeBaseChunkPlan
  • TrieveKnowledgeBaseCreate
  • TrieveKnowledgeBaseCreateType
  • TrieveKnowledgeBaseImport
  • TrieveKnowledgeBaseImportType
  • TrieveKnowledgeBaseProvider
  • TrieveKnowledgeBaseSearchPlan
  • TrieveKnowledgeBaseSearchPlanSearchType
  • UpdateAssistantDtoCredentialsItem_Trieve
  • UpdatePersonalityDto
  • UpdateScenarioDto
  • UpdateSimulationDto
  • UpdateSimulationSuiteDto
  • UpdateTrieveCredentialDto
  • UpdateTrieveKnowledgeBaseDto
  • UpdateWorkflowDtoCredentialsItem_Trieve
  • VapiModelProvider
  • VapiVoiceVoiceId
  • WorkflowCredentialsItem_Trieve
  • WorkflowUserEditableCredentialsItem_Trieve

Added

  • Regenerated API types and methods from the current public API definition.
  • with_raw_response accessors for response status, headers, and parsed data.
  • Current model deprecation notices, including replacement_status and optional replacement_model.

Fixed

  • Structured-output rerun parsing preserves workflow_id, including responses without a workflow ID.
  • Code-tool request unions and GHL-tool response unions include their schema discriminator mappings.
  • Assistant server-message type hints include the current event names from the API definition.

Compatibility

  • Supported and locally tested Python versions: 3.10 through 3.14.
  • Local regression checks cover Pydantic 1 and 2 and the optional aiohttp transport.

[2.0.1] - 2026-10-07

2.0.0 - 2026-06-24

Breaking Changes

  • CartesiaExperimentalControlsSpeedZero has been removed and replaced by CartesiaSpeedControlZero. Update any imports or type annotations referencing CartesiaExperimentalControlsSpeedZero to use CartesiaSpeedControlZero instead.
  • FallbackAzureVoiceVoiceIdZero has been removed and replaced by FallbackAzureVoiceIdZero. Update any imports or type annotations referencing FallbackAzureVoiceVoiceIdZero to use FallbackAzureVoiceIdZero instead.

1.11.1 - 2026-05-20

  • chore: remove redundant content-type headers from raw clients
  • Remove explicitly set "content-type": "application/json" headers from
  • multiple raw client request calls across the SDK. These headers are
  • already handled by the underlying HTTP client when a JSON body is
  • present, making the explicit declarations redundant.
  • Key changes:
  • Remove hardcoded content-type: application/json headers from RawAssistantsClient and AsyncRawAssistantsClient
  • Remove same redundant headers from RawEvalClient, RawInsightClient, RawObservabilityScorecardClient, RawPhoneNumbersClient, RawSquadsClient, RawStructuredOutputsClient, and RawToolsClient
  • Applies to both sync and async variants of all affected clients
  • 🌿 Generated with Fern

1.11.0 - 2026-04-22

Added

  • Call.subscription_limits — new optional field that exposes the org's SubscriptionLimits (including concurrency limit information) at the time of a call.

1.10.0 - 2026-04-10

  • The SDK now supports aiohttp as an optional async HTTP transport backend. Install the new extra (pip install vapi_server_sdk[aiohttp]) to have AsyncVapi automatically use httpx-aiohttp under the hood. Two new convenience classes, DefaultAioHttpClient and DefaultAsyncHttpxClient, are also now available for users who want to configure the async HTTP client explicitly.

1.9.1 - 2026-04-07

  • SDK regeneration
  • Unable to analyze changes with AI, incrementing PATCH version.