Forward-merge release/1.4 into develop

README.md

@@ -33,41 +33,46 @@ limitations under the License.
 
 <div align="center">
 
-*NVIDIA NeMo Agent Toolkit is a flexible, lightweight, and unifying library that allows you to easily connect existing enterprise agents to data sources and tools across any framework.*
+*NVIDIA NeMo Agent Toolkit adds intelligence to AI agents across any framework—enhancing speed, accuracy, and decision-making through enterprise-grade instrumentation, observability, and continuous learning.*
 
 </div>
 
 ## 🔥 New Features
 
-- [**A2A Support**](./docs/source/components/integrations/a2a.md) NeMo Agent Toolkit now supports deploying and consuming agents using the A2A protocol.
+- [**LangGraph Agent Automatic Wrapper:**](./examples/frameworks/auto_wrapper/langchain_deep_research/README.md) Easily onboard existing LangGraph agents to NeMo Agent Toolkit. Use the automatic wrapper to access NeMo Agent Toolkit advanced features with very little modification of LangGraph agents.
 
-- [**Amazon Bedrock AgentCore and Strands Agents Support:**](./docs/source/components/integrations/frameworks.md#strands) NeMo Agent Toolkit now supports building agents using Strands Agents framework and deploying them securely on Amazon Bedrock AgentCore runtime.
+- [**Automatic Reinforcement Learning (RL):**](./docs/source/improve-workflows/finetuning/index.md) Improve your agent quality by fine-tuning open LLMs to better understand your agent's workflows, tools, and prompts. Perform GRPO with [OpenPipe ART](./docs/source/improve-workflows/finetuning/rl_with_openpipe.md) or DPO with [NeMo Customizer](./docs/source/improve-workflows/finetuning/dpo_with_nemo_customizer.md) using NeMo Agent Toolkit built-in evaluation system as a verifier.
 
-- [**LangChain Agent Automatic Wrapper:**](./examples/frameworks/auto_wrapper/langchain_deep_research/README.md) NeMo Agent Toolkit now supports automatic wrapping of existing LangChain/LangGraph Agents.
+- [**Initial NVIDIA Dynamo Integration:**](./examples/dynamo_integration/README.md) Accelerate end-to-end deployment of agentic workflows with initial Dynamo support. Utilize the new agent-aware router to improve worker latency by predicting future agent behavior.
 
-- [**Microsoft AutoGen Support**](./docs/source/components/integrations/frameworks.md#autogen) NeMo Agent Toolkit now supports building agents using AutoGen framework.
+- [**A2A Support:**](./docs/source/components/integrations/a2a.md) Build teams of distributed agents using the A2A protocol.
 
-- [**Initial NVIDIA Dynamo Integration:**](https://docs.nvidia.com/dynamo/latest/) NeMo Agent Toolkit now has initial Dynamo support for end-to-end deployment acceleration of agentic workflows.
+- [**Safety and Security Engine:**](./examples/safety_and_security/retail_agent/README.md) Strengthen safety and security workflows by simulating scenario-based attacks, profiling risk, running guardrail-ready evaluations, and applying defenses with red teaming. Validate defenses, profile risk, monitor behavior, and harden agents across any framework.
 
-- [**Per-User Functions**](./docs/source/extend/custom-components/custom-functions/per-user-functions.md) NeMo Agent Toolkit now supports per-user functions for deferred instantiation, enabling per-user stateful functions, per-user resources, and other features.
+- [**Amazon Bedrock AgentCore and Strands Agents Support:**](./docs/source/components/integrations/frameworks.md#strands) Build agents using Strands Agents framework and deploy them securely on Amazon Bedrock AgentCore runtime.
 
-## ✨ Key Features
-
-- 🧩 [**Framework Agnostic:**](./docs/source/components/integrations/frameworks.md) NeMo Agent Toolkit works side-by-side and around existing agentic frameworks, such as [LangChain](https://www.langchain.com/), [LlamaIndex](https://www.llamaindex.ai/), [CrewAI](https://www.crewai.com/), [Microsoft Semantic Kernel](https://learn.microsoft.com/en-us/semantic-kernel/), and [Google ADK](https://google.github.io/adk-docs/), as well as custom enterprise agentic frameworks and simple Python agents. This allows you to use your current technology stack without replatforming. NeMo Agent Toolkit complements any existing agentic framework or memory tool you're using and isn't tied to any specific agentic framework, LLM provider, or data source.
-
-- 🔁 [**Reusability:**](./docs/source/components/sharing-components.md) Every agent, tool, and agentic workflow in this library exists as a function call that works together in complex software applications. The composability between these agents, tools, and workflows allows you to build once and reuse in different scenarios.
-
-- ⚡ [**Rapid Development:**](docs/source/get-started/tutorials/customize-a-workflow.md) Start with a pre-built agent, tool, or workflow, and customize it to your needs. This allows you and your development teams to move quickly if you're already developing with agents.
+- [**Microsoft AutoGen Support:**](./docs/source/components/integrations/frameworks.md#autogen) Build agents using the Microsoft AutoGen framework.
 
-- 📈 [**Profiling:**](./docs/source/improve-workflows/profiler.md) Use the profiler to profile entire workflows down to the tool and agent level, track input/output tokens and timings, and identify bottlenecks. While we encourage you to wrap (decorate) every tool and agent to get the most out of the profiler, you have the freedom to integrate your tools, agents, and workflows to whatever level you want. You start small and go to where you believe you'll see the most value and expand from there.
+- [**Per-User Functions:**](./docs/source/extend/custom-components/custom-functions/per-user-functions.md) Use per-user functions for deferred instantiation, enabling per-user stateful functions, per-user resources, and other features.
 
-- 🔎 [**Observability:**](./docs/source/run-workflows/observe/observe.md) Monitor and debug your workflows with dedicated integrations for popular observability platforms such as Phoenix, Weave, and Langfuse, plus compatibility with OpenTelemetry-based observability platforms. Track performance, trace execution flows, and gain insights into your agent behaviors.
-
-- 🧪 [**Evaluation System:**](./docs/source/improve-workflows/evaluate.md) Validate and maintain accuracy of agentic workflows with built-in evaluation tools.
-
-- 💬 [**User Interface:**](./docs/source/run-workflows/launching-ui.md) Use the NeMo Agent Toolkit UI chat interface to interact with your agents, visualize output, and debug workflows.
+## ✨ Key Features
 
-- 🔗 [**Full MCP Support:**](./docs/source/build-workflows/mcp-client.md) Compatible with [Model Context Protocol (MCP)](https://modelcontextprotocol.io/). You can use NeMo Agent Toolkit as an [MCP client](./docs/source/build-workflows/mcp-client.md) to connect to and use tools served by remote MCP servers. You can also use NeMo Agent Toolkit as an [MCP server](./docs/source/run-workflows/mcp-server.md) to publish tools via MCP.
+- 🛠️ **Building Agents**: Accelerate your agent development with tools that make it easier to get your agent into production.
+  - 🧩 [**Framework Agnostic:**](./docs/source/components/integrations/frameworks.md) Work side-by-side with agentic frameworks to add the instrumentation necessary for observing, profiling, and optimizing your agents. Use the toolkit with popular frameworks such as [LangChain](https://www.langchain.com/), [LlamaIndex](https://www.llamaindex.ai/), [CrewAI](https://www.crewai.com/), [Microsoft Semantic Kernel](https://learn.microsoft.com/en-us/semantic-kernel/), and [Google ADK](https://google.github.io/adk-docs/), as well as custom enterprise agentic frameworks and simple Python agents.
+  - 🔁 [**Reusability:**](./docs/source/components/sharing-components.md) Build components once and use them multiple times to maximize the value from development effort.
+  - ⚡ [**Customization:**](docs/source/get-started/tutorials/customize-a-workflow.md) Start with a pre-built agent, tool, or workflow, and customize it to your needs.
+  - 💬 [**Built-In User Interface:**](./docs/source/run-workflows/launching-ui.md) Use the NeMo Agent Toolkit UI chat interface to interact with your agents, visualize output, and debug workflows.
+- 📈 **Agent Insights:** Utilize NeMo Agent Toolkit instrumentation to better understand how your agents function at runtime.
+  - 📊 [**Profiling:**](./docs/source/improve-workflows/profiler.md) Profile entire workflows from the agent level all the way down to individual tokens to identify bottlenecks, analyze token efficiency, and guide developers in optimizing their agents.
+  - 🔎 [**Observability:**](./docs/source/run-workflows/observe/observe.md) Track performance, trace execution flows, and gain insights into your agent behaviors in production.
+- 🚀 **Agent Optimization:** Improve your agent's quality, accuracy, and performance with a suite of tools for all phases of the agent lifecycle.
+  - 🧪 [**Evaluation System:**](./docs/source/improve-workflows/evaluate.md) Validate and maintain accuracy of agentic workflows with a suite of tools for offline evaluation.
+  - 🎯 [**Hyper-Parameter and Prompt Optimizer:**](./docs/source/improve-workflows/optimizer.md) Automatically identify the best configuration and prompts to ensure you are getting the most out of your agent.
+  - 🧠 [**Fine-tuning with Reinforcement Learning:**](./docs/source/improve-workflows/finetuning/index.md) Fine-tune LLMs specifically for your agent and train intrinsic information about your workflow directly into the model.
+  - ⚡ [**NVIDIA Dynamo Integration:**](./examples/dynamo_integration/README.md) Use Dynamo and NeMo Agent Toolkit together to improve agent performance at scale.
+- 🔌 **Protocol Support:** Integrate with common protocols used to build agents.
+  - 🔗 [**Model Context Protocol (MCP):**](./docs/source/build-workflows/mcp-client.md) Integrate [MCP tools](./docs/source/build-workflows/mcp-client.md) into your agents or serve your tools and agents as an [MCP server](./docs/source/run-workflows/mcp-server.md) for others to consume.
+  - 🤝 [**Agent-to-Agent (A2A) Protocol:**](./docs/source/components/integrations/a2a.md) Build teams of distributed agents with full support for authentication.
 
 With NeMo Agent Toolkit, you can move quickly, experiment freely, and ensure reliability across all your agent-driven projects.
 
@@ -84,7 +89,7 @@ To install the latest stable version of NeMo Agent Toolkit from PyPI, run the fo
 pip install nvidia-nat
 ```
 
-NeMo Agent Toolkit has many optional dependencies which can be installed with the core package. Optional dependencies are grouped by framework and can be installed with the core package. For example, to install the LangChain/LangGraph plugin, run the following:
+NeMo Agent Toolkit has many optional dependencies that can be installed with the core package. Optional dependencies are grouped by framework. For example, to install the LangChain/LangGraph plugin, run the following:
 
 ```bash
 pip install "nvidia-nat[langchain]"
@@ -160,8 +165,10 @@ Before getting started, it's possible to run this simple workflow and many other
 ## 🛣️ Roadmap
 
 - [x] Automatic Reinforcement Learning (RL) to fine-tune LLMs for a specific agent.
-- [ ] Integration with [NeMo Guardrails](https://github.com/NVIDIA/NeMo-Guardrails) to secure any function in an agent workflow.
-- [x] End-to-end acceleration using intelligent integrations with [NVIDIA Dynamo](https://github.com/ai-dynamo/dynamo).
+- [x] Integration with [NVIDIA Dynamo](https://github.com/ai-dynamo/dynamo) to reduce LLM latency at scale.
+- [ ] Improve agent throughput with KV-Cache optimization.
+- [ ] Integration with [NeMo Guardrails](https://github.com/NVIDIA/NeMo-Guardrails) to improve agent safety and security.
+- [ ] Improved memory interface to support self-improving agents.
 
 ## 💬 Feedback
 

docs/source/improve-workflows/evaluate.md

@@ -15,7 +15,7 @@ See the License for the specific language governing permissions and
 limitations under the License.
 -->
 
-# Evaluating NVIDIA NeMo Agent Toolkit Workflows
+# Agent Evaluation in NVIDIA NeMo Agent Toolkit
 
 Evaluation is the process of executing [workflows](../build-workflows/about-building-workflows.md) ([agents](../components/agents/index.md), [tools](../build-workflows/functions-and-function-groups/functions.md#agents-and-tools), or pipelines) on curated test data and measuring their quality using quantitative metrics such as accuracy, reliability, and latency. Each of these metrics in turn is produced by an evaluator.
 

docs/source/improve-workflows/finetuning/index.md

@@ -18,7 +18,7 @@ limitations under the License.
 
 <!-- path-check-skip-begin -->
 
-# Finetuning Harness
+# NVIDIA NeMo Agent Toolkit Finetuning Harness for Reinforcement Learning
 
 :::{warning}
 **Experimental Feature**: The Finetuning Harness is experimental and may change in future releases. Future versions may introduce breaking changes without notice.

docs/source/improve-workflows/finetuning/rl_with_openpipe.md

@@ -17,7 +17,7 @@ limitations under the License.
 
 <!-- path-check-skip-begin -->
 
-# OpenPipe ART Integration
+# GRPO with OpenPipe ART
 
 This guide covers the integration between the NVIDIA NeMo Agent Toolkit finetuning harness and [OpenPipe ART](https://art.openpipe.ai/) (Agent Reinforcement Trainer), an open-source framework for teaching [LLMs](../../build-workflows/llms/index.md) through reinforcement learning.
 
@@ -76,7 +76,7 @@ NeMo Agent Toolkit connects to this backend through the `ARTTrainerAdapter`, whi
 
 The following table highlights the current support matrix for using ART with different agent frameworks in the NeMo Agent Toolkit:
 
-| Agent Framework        | Support | 
+| Agent Framework        | Support |
 |------------------------|--------------------------------------------------|
 | LangChain or LangGraph | ✅ Supported                                      |
 | Google ADK             | ✅ Supported                                      |
@@ -444,7 +444,7 @@ The `ARTTrainer` orchestrates the complete training loop:
 
 ### Running Training
 
-You must have OpenPipe ART plugin installed (`nvidia-nat-openpipe-art`), and an OpenPipe ART server running 
+You must have OpenPipe ART plugin installed (`nvidia-nat-openpipe-art`), and an OpenPipe ART server running
 and configured to accept training jobs.
 
 ```bash

docs/source/improve-workflows/optimizer.md

@@ -14,7 +14,7 @@ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 See the License for the specific language governing permissions and
 limitations under the License.
 -->
-# NVIDIA NeMo Agent Toolkit Optimizer Guide
+# NVIDIA NeMo Agent Toolkit Optimizer
 
 This document provides a comprehensive overview of how to use the NeMo Agent Toolkit Optimizer to tune your NeMo Agent Toolkit [workflows](../build-workflows/about-building-workflows.md).
 
@@ -630,7 +630,7 @@ This matrix visualization shows:
 **Example decision process**:
 - If latency is critical: Choose a Pareto optimal point with the lowest latency that still meets your accuracy requirements
 - If accuracy is paramount: Select the highest accuracy configuration and accept the latency trade-off
-- For balanced performance: Pick a point in the middle of the Pareto front 
+- For balanced performance: Pick a point in the middle of the Pareto front
 
 ## A Complete Example of Optimization