TL;DR

Langflow now streams tokens incrementally for Agents, improving responsiveness when using LLMs (Large Language Models).

Fixes Worth Knowing

• Agent LLM responses now stream tokens incrementally for a faster user experience.
• Docling remote component now correctly handles task IDs and processing errors.
• Model handling for tool calling in agents has been updated, including support for IBM models.
• The AGENTS documentation has been significantly revised to be more organized and informative.
• Docker base images have been updated to Trixie for improved stability.

TL;DR

Langflow now supports flow version control, enabling tracking and rollback of changes to your Langflow projects.

Breaking

• Kubernetes Deployment: A fix was implemented to resolve crashes related to runtime port parsing, requiring review of existing Kubernetes deployments.

New

• Flow Version Control: Track and revert changes to your Langflow flows.
• DataFrame Operations: Added Concatenate and Merge operations for working with DataFrames.
• Plugin Support: Re-added OSS/enterprise plugin support with SSO migration.

Fixes Worth Knowing

• UI Improvements: Resolved several UI issues including scrolling in the playground, clipped table row icons, and visual lag.
• Component Updates: Fixed issues with custom component generation, API component requests, and model metadata for Anthropic, Google, and OpenAI.
• Knowledge Base: Improved Knowledge Base functionality, including ingestion and polymorphic job tracking.

Before You Upgrade

• Review your Kubernetes deployment configurations to ensure compatibility with the runtime port parsing fix.

TL;DR

Langflow now supports OpenAI’s Assistants API (AI-powered virtual assistants), enabling more complex conversational flows.

Fixes Worth Knowing

• Resolved an issue where the application would crash when attempting to load a flow with missing components.
• Fixed a bug preventing users from deleting flows containing nodes with no connections.
• Corrected a problem with incorrect display of component descriptions in the sidebar.

TL;DR

Langflow now correctly integrates with Pinecone (vector database), resolving previous issues with its implementation.

Fixes Worth Knowing

Pinecone integration is now functional, allowing users to reliably connect and utilize Pinecone for storing and retrieving vector embeddings.

TL;DR

Langflow now correctly integrates with Anthropic LLMs (large language models), resolving a previous issue.

Fixes Worth Knowing

Anthropic LLM integration is now functional, allowing users to utilize this model within Langflow.