The Conversational Agents console has launched for preview to create agents that can use a combination of generative and deterministic features. See how we intend on migrating to this new console.
]]>Dialogflow CX (Conversational Agents): Dialogflow CX has launched a new feature that allows you to auto-generate and auto-translate Intent training phrases, Entity synonyms, and Fulfillment phrases in the language of your choice. See the documentation for details.
Dialogflow CX (Conversational Agents): You can now require a full match for banned phrases in addition to a partial match. If enabled, a full match requires the input to be matched exactly in order to trigger a ban. For more information about setting banned phrases, see the documentation.
Dialogflow CX (Conversational Agents): Service directory support is now enabled for flexible webhooks. See the webhooks documentation for details.
]]>Dialogflow CX (Conversational Agents): You can now set either a partial match or a full match to banned phrases. This setting applies to playbooks, datastores, and generators. You can enable and test this feature in Agent Settings > Generative AI > Banned phrases > Match requirements.
]]>Dialogflow CX data stores: The following languages are now GA. See the language support page for details.
Dialogflow CX: You can now configure an access token name in Dialogflow Messenger to store the end user's authentication when they sign in, and then use it as the bearer token for tool authentication. See the Dialogflow Messenger documentation for more information about enabling this feature.
]]>Data store tools: You can now optimize the RAG performance of data store tools used by Playbooks. See the documentation for details.
Dialogflow CX: New feature Context token limits has been added to Agent Settings > Generative AI. You can use this feature to set a percentage of the token budget to be reserved for conversation history, as a maximum. See the Agent Settings documentation for details.
Generators and data store handlers: The model gemini-1.5-flash-002 is now GA.
Data store handlers: The default generative model has been changed to gemini-1.5-flash-001.
Dialogflow CX: As of August 2024, us-dialogflow.* has been re-introduced as the canonical endpoint for the US multi-region. The usa-dialogflow.* endpoint is still supported as an alias. See the regionalization documentation for details.
Dialogflow CX & ES: In order to increase the stability of Cloud Text-to-speech, out-of-quota requests for Journey Voices will now be fulfilled with a fallback voice of the same speaker persona. To opt out, contact your Google account team.
Dialogflow CX & ES: Dialogflow now supports A-law encoding in addition to Mu-law encoding for input and output audio. A-law and Mu-law are the two available formats in G.711.
]]>Conversational Agents: The gemini-1.5-flash-001 model is generally available (GA) for data store handlers as of August 20, 2024. The gemini-1.5-flash-002 model remains public Preview.
Starting in late 2024 and ending in early 2025, there will be changes to some Dialogflow product names, feature names, and consoles. See the documentation for complete details.
Dialogflow CX: You can now enable security checks to prevent prompt injection attacks. See the documentation for details.
Dialogflow CX: Cloud Text-to-Speech europe-west1 and europe-west3 regions for Neural2 voices now use corresponding regions.
Dialogflow CX and Vertex AI: The text-bison@002, text-bison and code-bison models will be deprecated on October 21, 2024 and automatically upgraded to the gemini-1.5-flash-001 model. This change applies to Vertex AI agents and the following Dialogflow CX Generative Features:
After the upgrade on October 21, 2024, gemini-1.5-flash-001 will be automatically selected in the console. We recommend that you upgrade to the new model early to allow enough time for testing and to ensure that your solution works as intended.
Dialogflow CX & ES: Text-to-speech Journey Voices now supports MULAW output audio_encoding (CX, ES) in addition to LINEAR16. Future updates to Journey Voices will appear in the Cloud Text-to-Speech documentation.
Dialogflow CX: Cloud Text-to-Speech europe-west1 and europe-west3 regions for Neural2 voices will temporarily use the eu mulit-region instead.
Dialogflow CX and Vertex AI Agents: Generative features will migrate to the gemini-1.5-flash-001 model on September 30, 2024. See the email notification.
]]>Dialogflow CX & ES: Text-to-Speech Journey Voices will get an update in the week of Sept 3. If you select a journey voice in your agent Text-to-Speech settings (CX, ES), only LINEAR16 output audio_encoding (CX, ES) will be supported starting from Sept 3. This model update will also include slight variations in pauses, tone, and so on of the synthesized journey voices.
]]>Dialogflow CX: You can now configure sensitivity levels of safety filters with respect to different Responsible AI (RAI) categories.
Vertex AI Agents: OpenAPI tool authentication now supports custom client certificates for mutual TLS authentication.
]]>Dialogflow CX: You can now enable consent-based end-user content redaction. If this setting is enabled, it lets you use a special session parameter to control whether end-user input and parameters are redacted from conversation history and Cloud logging. See the agent settings documentation for details.
Dialogflow CX: Dialogflow now uses corresponding Cloud Text-to-Speech regions for Neural2 voices.
The status of the Dialogflow Speech-to-Text model migrations has changed:
The Dialogflow speech model migration initially announced in December 2023 and later postponed to the new dates has been placed on hold until further notice (no ETA). For more information, see the migration documentation for Dialogflow CX and Dialogflow ES.
The Dialogflow phone_call speech model migration announced in March 2024 has been partially completed. The remaining migration parts have been placed on hold until further notice (no ETA). For more information, see the migration documentation for Dialogflow CX and Dialogflow ES.
Vertex AI Agents: Agent apps now provide generative settings for input token limit, output token limit, and temperature.
]]>Dialogflow ES: As of May 27 2024, Twilio no longer supports integrations with Dialogflow ES. For more details and information about migrating to Dialogflow CX, see the Twilio documentation.
Dialogflow CX: The gemini-1.5-flash generative model is now available for the generators feature.
]]>Vertex AI Agents: The following new regions are supported by agent apps:
europe-west1europe-west2europe-west3northamerica-northeast1us-west1The following was incorrectly announced: Dialogflow CX: The gemini-1.5-flash generative model is now available for the generators feature.
Dialogflow CX now offers custom webhook templates for integration with Salesforce. See the webhooks documentation for details.
Data store agents: You can now run self-service evaluation which will assess the quality of your data store agent and recommend changes.
All generative features: It was announced previously that the text-bison@001 model will be deprecated. In addition, the code-bison@001 model and fine-tuned text-bison@001 options will be deprecated. This deprecation will happen mid June. The deprecated models will be updated to gemini-1.0-pro-001, as previously announced. For more information, see the email announcement.
Data store agents: The gemini-1.5-flash generative model is now available for selection in the console.
]]>Dialogflow CX: You can now integrate with Soul Machines to create 3-D avatars.
]]>Vertex AI Agents: OpenAPI tools now support private network access
Vertex AI Agents: OpenAPI tool authentication now supports Bearer Token.
Dialogflow CX: VPC Service Controls now support Cloud Functions and Cloud Run.
]]>Dialogflow ES and Dialogflow CX: The us-dialogflow.googleapis.com endpoint and locations/us resource location, which served as aliases for global resources, will be discontinued starting May 21, 2024. We have changed the date
to update resource locations and endpoints from April 16, 2024 to May 21, 2024 to provide you with additional time. For more information, see the email announcement.
Note
global region (ES, CX) and only if you use the us alias in the API requests to these global-region agents. If you created agents in us-central1, us-east1, us-west1, and us (multi-region) regions, no action is required.us multi-region endpoint that was announced recently. Dialogflow CX and Vertex AI Agents: Effective June 15, 2024, the following generative features will be upgraded from text-bison-001 and fine-tuned text-bison@001 options to gemini-1.0-pro-001:
For more information, see the email announcement
]]>Data store agents: The default generative model has been changed to gemini-1.0-pro-001.
]]>Vertex AI Conversation has been renamed to Vertex AI Agents
Vertex AI Agents: Agent apps now support all languages supported by Vertex AI generative models.
Vertex AI Agents: Agent apps now support the eu multi-region.
Dialogflow CX: You can now access the session ID with built-in parameters.
]]>Dialogflow CX: The following new region is available:
us: United States multi-region accessed via usa-dialogflow.googleapis.com hostnameDialogflow CX: The Conversation history API is now available for public preview.
]]>Vertex AI Conversation: You can now create a data store in one language that is connected to an agent that uses different languages.
]]>Dialogflow CX: Language auto detect is now available for chat conversations. You can configure Dialogflow CX to detect an end-user's language and automatically respond in that same language. See the language reference page for a list of languages available for this feature.
]]>Dialogflow CX: Call companion is now generally available with new user interface settings.
Dialogflow CX: Dialogflow CX phone gateway is now generally available.
Vertex AI Conversation: Data stores now support parse and chunk configuration.
]]>