Skip to content
MindflowMindflowMindflow Help Center

Chupacabra

Versioned deployments, faster AI··Task, new LLM models, and LLM tokens exposed. The chupacabra, a legendary creature from Latin American folklore, is said to attack and drink the blood of livestock, particularly goats

NEW

Section titled “NEW”

Deploy a specific flow version 🔢

Section titled “Deploy a specific flow version 🔢”

Ref: #7222

You can now deploy a specific version of a flow, ensuring that your production deployments remains stable while you continue to make edits.

After creating a new version of a flow, click on the ··· menu at the top right of the flow card and select Deploy flow...

Deploy specific flow version

A popup will appear, asking for the following information:

  • The destination environment

  • The version to use

    Version use

This allows you to have multiple versions of the same flow deployed.

Version use

In the designer, you can still quickly switch between different deployments using the top bar.

Version use

Here are a few things to keep in mind:

  • Automatic updates for Latest version

    If a flow is set to Latest, all its deployments will be updated automatically when the flow is modified. This can be very useful depending on the use case, as it allows you to keep multiple deployments in sync.

    For example, if you create a flow and deploy it in two environments using the Latest version, any changes made to the flow will affect both deployments.

  • Immutable Versions

    A specific version (e.g., V1) cannot be modified. Therefore, if you have deployments set to a specific version of a flow, they will not be affected by changes made to the flow Latest version.

    For example, if you create a flow and deploy it in two environments using version V1 any subsequent changes to the flow will not affect these deployments.

New LLM models for AI··Task 🧠

Section titled “New LLM models for AI··Task 🧠”

Ref: #8600

We have added several new LLM models to enhance your AI··Task capabilities:

  • Amazon Nova Lite - Optimized for enterprise tasks with balanced performance and cost efficiency. Great for text summarization and general content generation.

  • Amazon Nova Pro - Higher capability model with enhanced reasoning and complex task handling. Ideal for technical content creation and advanced analytical responses.

  • Llama 3.2 3B - Compact but powerful open-source model that excels at instruction following and creative writing while being resource-efficient.

  • Mistral Pixtral Large - Powerful large language model with advanced text processing capabilities, excellent reasoning abilities, and sophisticated content generation for complex tasks.

    Mistral pixtral large

CHANGES

Section titled “CHANGES”

Add LLM token usage for AI··Task 📊

Section titled “Add LLM token usage for AI··Task 📊”

Ref: #8604

Thanks to your feedback, we have added metadata to help you understand AI token usage, including input tokens, output tokens, and total tokens, for improved tracking and analysis.

Mistral pixtral large

Reduce AI··Task latency ⏱️

Section titled “Reduce AI··Task latency ⏱️”

Ref: #8538

We have improved AI··Task performance by eliminating redundant streaming events, resulting in reduced latency.

FIXES

Section titled “FIXES”
  • AI··Task: Fixed an issue where shared memory across executions caused tool configuration errors; memory is now isolated per task. (#8618)
  • Fixed a crash when pasting a step inside a For Each block, which prevented the flow from loading properly. (#8563)
  • Infinite execution in the synchronous webhook response step. (#8158)

INTEGRATIONS

Section titled “INTEGRATIONS”