Quotas and limits

This document contains the commonly encountered quotas and limits for use of Dialogflow. Information on how to find a complete list of quotas and limits is provided below. We reserve the right to change these constraints, and this page will be updated to reflect any changes.

Quotas

Quotas are default constraints applied to your project. When using generative features, each combination of model and region has its own generative quota. If you are using a paid edition, you can request a quota increase. The purpose of quotas is to throttle service load per client, which protects a service from being overloaded and a client from unexpected resource usage.

Dialogflow quotas vary, depending on the edition of your agent. Quotas apply to each project and are shared across all applications and IP addresses using a project.

Quota time boundary details:

• Per-month quotas are replenished on the 1st of each month at 12:00 AM Pacific Time.
• Per-day quotas are replenished daily at 12:00 AM Pacific Time. When a Pacific Time daylight savings transition increases the length of a day, a proportional quota bonus will be granted. However, there is no equivalent deduction when a shorter day occurs due to the daylight savings transition.
• Per-minute quotas are refreshed every 60 seconds on the minute.

The quotas listed in the quotas table are the commonly encountered quotas. To review all quotas, see the Google Cloud console Dialogflow quotas page.

Generative quotas

When your agent uses Dialogflow CX generative features, each combination of model and region have unique quotas.

Quota identifier

The ConversationalAgentLlmTokenConsumption quota is a regional quota that tracks token usage for large language models (LLMs) in Dialogflow CX. This quota is measured across different LLM base models, and currently supports the following models:

• gemini-2.5-flash
• gemini-2.5-flash-lite

The ConversationalAgentLlmTokenConsumption quota tracks LLM token usage across the following Dialogflow CX features:

• Generators
• Generative Fallback
• Playbook
• Datastore

The default limit for the ConversationalAgentLlmTokenConsumption quota is 600,000 tokens per minute.

Multi-region quota increase request

If you need to request a quota increase for a multi-region, submit your request for one of the following specific regions:

• For agents in the US and Global multi-regions, use us-central1.
• For agents in the EU multi-region, use europe-west4.

Visualize token usage by region and model

To gain insights into your LLM token consumption across different regions and models, follow these steps using Metrics Explorer:

1. Access Metrics Explorer: Open Metrics Explorer.
2. Select the Metric: Choose Conversational Agent LLM tokens quota usage as the metric you want to analyze.
3. Configure Aggregation and Aligner:
   • Set the aggregation function to Sum to obtain the total token usage.
   • Click Configure Aligner to enable the calculation of changes between consecutive data points in your time series. This helps in understanding usage trends over time.
4. Define Aggregation Dimensions: To break down the token usage, select location and base_model as the dimensions for aggregation. This will allow you to see how token consumption varies across different regions and models.
5. Set Aligner Function: Choose Sum as the aligner function to calculate the total change in token usage between consecutive data points.

Quota increase request

You can request a quota increase adjustment if necessary.

If you use different projects for Dialogflow agents and for authenticating with service accounts, request a quota increase for the consumer project. For more information, see Using multiple projects for Dialogflow ES or Using multiple projects for Dialogflow CX.

For Dialogflow ES agents, make sure to upgrade the agent to the Dialogflow Essentials edition before requesting a quota increase.

Quotas table

The following terms are used to describe pricing and quotas:

• Request: A request is defined as any API call to the Dialogflow CX service, whether direct with API usage or indirect with integration or console usage. Depending on the task and design of the agent, the number of requests needed for an end-user to accomplish a task with a Dialogflow CX agent can vary greatly.
• Session: A session is a conversation between an end-user and a Dialogflow CX agent. A session remains active and its data is stored for 30 minutes after the last request is sent for the session. A session can be either a chat session or a voice session.
  • Chat session: A chat session only uses text for both requests and responses.
  • Voice session: A voice session uses audio for requests, responses, or both.
• Consumer projects and resource projects: If you use multiple projects, it is possible that the project associated with your request authentication (consumer project) is not the same project that is associated with the agent in the request (resource project). In this case, the consumer project is used to determine prices and quotas. For more information, see Using multiple projects for Dialogflow ES or Using multiple projects for Dialogflow CX.

The following tables provide a quota comparison for editions by agent type. Unless a feature is indicated as included, quotas are cumulative for all features used by a request.