FloTorch Enterprise V1.2 Release: Advancing Model Evaluation, Agent Intelligence, and UI Consistency

At FloTorch, our mission is to empower enterprises to build reliable, production-ready AI systems at scale. With the launch of FloTorch Enterprise V1.2, we are introducing a new wave of enhancements that simplify model evaluation, strengthen agent intelligence, and improve user experience. This release continues to enhance performance, scalability, and seamless integration, vital elements that, alongside the robust governance provided, are crucial for organizations accelerating their AI adoption.

FloTorch is designed to make it easier for enterprises to evaluate models at scale, configure agents with memory and session management in a no-code manner, and integrate seamlessly into existing systems. This release is about giving Builders the tools they need to move from experimentation to production with confidence, while providing Decision Makers a single pane of glass dashboard to monitor and govern their AI Workspaces.

New Evaluation Runtime

The upgraded evaluation runtime enables flexible and scalable benchmarking of models. Validated backend APIs ensure execution accuracy, allowing teams to compare models with confidence and speed. This capability helps enterprises shorten experimentation cycles and make informed model adoption decisions.

Enhanced Trace Visibility

Detailed span-level metadata is now available in the UI for traces, providing greater transparency into model and agent performance. This improvement enhances observability, making it easier to identify inefficiencies and optimize AI workflows at scale.

Unified Design System

A new design system has been rolled out across the platform, ensuring visual and functional consistency. The unified UI improves accessibility, simplifies navigation, and reduces friction for users, resulting in faster onboarding and more efficient operations.

Expanded MCP Integration

MCP (Model Control Protocol) can now be configured seamlessly via both the UI and Gateway. This enhancement streamlines integration with external tools and providers, helping enterprises expand their AI ecosystems without compromising reliability.

Memory and Session Management

Agents are now equipped with memory and session persistence capabilities. By integrating memory providers like mem0 and centralizing session management in the Gateway, agents can retain context across multi-step workflows. This enables stateful, enterprise-grade interactions for conversational and decision-making agents.

Visual Model Builder

With the introduction of the visual model builder, users can now build, configure, and manage models through an intuitive interface. This lowers technical barriers, accelerates experimentation, and allows cross-functional teams to collaborate more effectively.

Real-Time Streaming

The Gateway now supports real-time response streaming, delivering faster and more interactive experiences for long-form outputs such as live chats, document drafting, and reporting. This capability improves usability in time-sensitive applications.

MCP Proxy in Gateway

MCP Proxy support ensures that external tool and service requests are routed securely and efficiently through the Gateway. This strengthens compliance and trust for enterprises deploying AI across regulated industries.

Agent Lifecycle Management

Enterprises can now fully manage agent lifecycles through the UI and APIs. From creation and configuration to assigning tools and models, this feature streamlines orchestration and simplifies scaling of intelligent workflows.

New SDKs and Plugins

To accelerate integration with existing enterprise ecosystems, we are introducing new SDKs and plugins:

FloTorch SDK (PyPI): Provides simple access to chat completion, memory management, vector search, and session handling via the Gateway.
FloTorch ADK Plugin (PyPI): Extends Google ADK with memory persistence, session management, and LLM orchestration for building production-ready agents.
FloTorch CrewAI Plugin (PyPI): Enables advanced multi-agent orchestration, memory persistence, and session handling within CrewAI environments.

Current Limitations

As with any major release, a few limitations are under review. These include breadcrumb navigation inconsistencies, graph filtering issues, missing fields in project views, cache eviction under high-load conditions, and experiment errors with certain DeepSeek models. Our team is actively working on these areas, with fixes scheduled for upcoming updates.

Get Started

FloTorch Enterprise provides enterprises with a stronger foundation to evaluate, deploy, and manage AI systems at scale. By combining advanced model benchmarking, stateful agents, real-time interactivity, and seamless integrations, this release equips organizations to accelerate outcomes with confidence.

You can explore the new features today by selecting your preferred cloud option: