Forward-merge release/1.3 into develop

.coderabbit.yaml

@@ -21,7 +21,7 @@ reviews:
   collapse_walkthrough: false
   pre_merge_checks:
     title:
-      mode: warning
+      mode: error
       requirements: "Title should be concise, and descriptive, and use imperative mood (max ~72 chars)."
     description:
       mode: warning

.gitlab-ci.yml

@@ -42,6 +42,7 @@ variables:
   NAT_CI_ETCD_HOST: "etcd"
   NAT_CI_MILVUS_HOST: "milvus"
   NAT_CI_MYSQL_HOST: "mysql"
+  NAT_CI_OPENSEARCH_URL: "http://opensearch:9200"
   NAT_CI_REDIS_HOST: "redis"
   NAT_CI_S3_HOST: "minio"
   UV_CACHE_DIR: .uv-cache
@@ -81,19 +82,32 @@ test:python_tests:
     - name: minio/minio:RELEASE.2025-07-18T21-56-31Z
       alias: minio
       command: ["server", "/data", "--console-address", ":9001"]
-    - mysql:9.3
+      pull_policy: if-not-present
+    - name: mysql:9.3
+      pull_policy: if-not-present
     - name: arizephoenix/phoenix:latest
       alias: phoenix
-    - redis:7-alpine
+      pull_policy: if-not-present
+    - name: redis:8.0
+      alias: redis
+      pull_policy: if-not-present
     - name: quay.io/coreos/etcd:v3.5.5
       alias: etcd
       command: ["etcd", "--advertise-client-urls", "http://0.0.0.0:2379", "--listen-client-urls", "http://0.0.0.0:2379"]
+      pull_policy: if-not-present
     - name: milvusdb/milvus:v2.3.1
       alias: milvus
       variables:
         ETCD_ENDPOINTS: etcd:2379
         MINIO_ADDRESS: minio:9000
       command: ["milvus", "run", "standalone"]
+      pull_policy: if-not-present
+    - name: opensearchproject/opensearch:2.11.1
+      alias: opensearch
+      pull_policy: if-not-present
+      variables:
+        discovery.type: "single-node"
+        plugins.security.disabled: "true"
 
   script:
     - echo "Running tests"
@@ -147,3 +161,4 @@ upload:artifactory:
     - if: $CI_COMMIT_BRANCH == $CI_DEFAULT_BRANCH
     - if: $CI_COMMIT_BRANCH == 'main'
     - if: $CI_COMMIT_TAG
+    - if: $CI_COMMIT_BRANCH =~ /^release\/.*$/

README.md

@@ -37,7 +37,7 @@ NVIDIA NeMo Agent toolkit is a flexible, lightweight, and unifying library that
 
 ## ✨ Key Features
 
-- 🧩 [**Framework Agnostic:**](./docs/source/quick-start/installing.md#framework-integrations) 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/), and [Microsoft Semantic Kernel](https://learn.microsoft.com/en-us/semantic-kernel/), as well as customer enterprise 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, long-term memory, or data source.
+- 🧩 [**Framework Agnostic:**](./docs/source/quick-start/installing.md#framework-integrations) 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 customer enterprise 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, long-term memory, or data source.
 
 - 🔁 [**Reusability:**](./docs/source/extend/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.
 
@@ -72,6 +72,13 @@ Before you begin using NeMo Agent Toolkit, ensure that you have Python 3.11, 3.1
 
 ### Stable Version
 
+Based on your system settings, you may need to configure and activate a Python virtual environment. On macOS or Linux, you can run the following commands:
+
+```bash
+python3 -m venv .venv
+source .venv/bin/activate
+```
+
 To install the latest stable version of NeMo Agent Toolkit, run the following command:
 
 ```bash
@@ -108,7 +115,7 @@ More information on how to install the latest development version and contribute
    export NVIDIA_API_KEY=<your_api_key>
    ```
 
-2. Create the NeMo Agent toolkit workflow configuration file. This file will define the agents, tools, and workflows that will be used in the example. Save the following as `workflow.yaml`:
+2. Create the NeMo Agent toolkit workflow configuration file. This file will define the agents, tools, and workflows that will be used in the example. Save the following as `workflow.yml`:
 
    ```yaml
    functions:
@@ -137,10 +144,10 @@ More information on how to install the latest development version and contribute
       parse_agent_response_max_retries: 3
    ```
 
-3. Run the Hello World example using the `nat` CLI and the `workflow.yaml` file.
+3. Run the Hello World example using the `nat` CLI and the `workflow.yml` file.
 
    ```bash
-   nat run --config_file workflow.yaml --input "List five subspecies of Aardvarks"
+   nat run --config_file workflow.yml --input "List five subspecies of Aardvarks"
    ```
 
    This will run the workflow and output the results to the console.
@@ -168,9 +175,9 @@ The following diagram illustrates the key components of NeMo Agent toolkit and h
 ## 🛣️ Roadmap
 
 - [x] Integrate with [NeMo DataFlywheel](https://github.com/NVIDIA-AI-Blueprints/data-flywheel) for continuous model improvement from production data.
-- [ ] Add support for [Google ADK](https://google.github.io/adk-docs/) framework.
+- [x] Add support for [Google ADK](https://google.github.io/adk-docs/) framework.
 - [x] Add an agent optimizer to auto-tune hyperparameters and prompts to maximize performance.
-- [ ] MCP authorization and streamable HTTP support.
+- [x] MCP authorization and streamable HTTP support.
 - [ ] Integration with [NeMo Guardrails](https://github.com/NVIDIA/NeMo-Guardrails) to secure any function in an agent workflow.
 - [ ] End-to-end acceleration using intelligent integrations with [NVIDIA Dynamo](https://github.com/ai-dynamo/dynamo).
 
@@ -184,9 +191,16 @@ We would like to thank the following open source projects that made NeMo Agent t
 
 - [CrewAI](https://github.com/crewAIInc/crewAI)
 - [FastAPI](https://github.com/tiangolo/fastapi)
+- [Google Agent Development Kit (ADK)](https://github.com/google/adk-python)
 - [LangChain](https://github.com/langchain-ai/langchain)
 - [Llama-Index](https://github.com/run-llama/llama_index)
 - [Mem0ai](https://github.com/mem0ai/mem0)
+- [Model Context Protocol (MCP)](https://github.com/modelcontextprotocol/modelcontextprotocol)
+- [MinIO](https://github.com/minio/minio)
+- [OpenTelemetry](https://github.com/open-telemetry/opentelemetry-python)
+- [Phoenix](https://github.com/arize-ai/phoenix)
 - [Ragas](https://github.com/explodinggradients/ragas)
+- [Redis](https://github.com/redis/redis-py)
 - [Semantic Kernel](https://github.com/microsoft/semantic-kernel)
+- [Weave](https://github.com/wandb/weave)
 - [uv](https://github.com/astral-sh/uv)

ci/scripts/github/tests.sh

@@ -30,7 +30,7 @@ rapids-logger "Running tests with Python version $(python --version) and pytest
 set +e
 
 REPORT_IDENT_SLUG="$(arch)-py${PYTHON_VERSION}"
-pytest --junit-xml=${REPORTS_DIR}/report-${REPORT_IDENT_SLUG}_pytest.xml \
+pytest --run_slow --junit-xml=${REPORTS_DIR}/report-${REPORT_IDENT_SLUG}_pytest.xml \
        --cov=nat --cov-report term-missing \
        --cov-report=xml:${REPORTS_DIR}/report-${REPORT_IDENT_SLUG}_pytest_coverage.xml
 PYTEST_RESULTS=$?

ci/scripts/gitlab/build_wheel.sh

@@ -59,7 +59,7 @@ if [[ "${BUILD_NAT_COMPAT}" == "true" ]]; then
     done
 fi
 
-if [[ "${CI_COMMIT_BRANCH}" == "${CI_DEFAULT_BRANCH}" || "${CI_COMMIT_BRANCH}" == "main" ]]; then
+if [[ "${CI_COMMIT_BRANCH}" == "${CI_DEFAULT_BRANCH}" || "${CI_COMMIT_BRANCH}" == "main" || "${CI_COMMIT_BRANCH}" == "release/"* ]]; then
     rapids-logger "Uploading Wheels"
 
     # Find and upload all .whl files from nested directories

docs/source/conf.py

@@ -141,7 +141,11 @@ def _build_api_tree() -> Path:
 nbsphinx_allow_errors = True  # Continue through Jupyter errors
 add_module_names = False  # Remove namespaces from class/method signatures
 myst_heading_anchors = 4  # Generate links for markdown headers
-copybutton_prompt_text = ">>> |$ |# "  # characters to be stripped from the copied text
+copybutton_prompt_text = ">>> |$ "  # characters to be stripped from the copied text
+
+# Allow GitHub-style mermaid fence code blocks to be used in markdown files
+# see https://myst-parser.readthedocs.io/en/latest/configuration.html
+myst_fence_as_directive = ["mermaid"]
 
 suppress_warnings = [
     "myst.header"  # Allow header increases from h2 to h4 (skipping h3)

docs/source/extend/telemetry-exporters.md

@@ -160,7 +160,7 @@ Telemetry exporters in NeMo Agent toolkit are responsible for:
 
 The flexible telemetry export system routes workflow events through different exporter types to various destinations:
 
-```{mermaid}
+```mermaid
 graph TD
     A[Workflow Events] --> B[Event Stream]
     B --> C[Telemetry Exporter]
@@ -182,7 +182,7 @@ graph TD
 
 NeMo Agent toolkit supports several types of exporters based on the data they handle:
 
-```{mermaid}
+```mermaid
 graph LR
     A["IntermediateStep"] --> B["Raw Exporter"]
     A --> C["Span Exporter"]

docs/source/index.md

@@ -131,7 +131,7 @@ Adding a Telemetry Exporter <./extend/telemetry-exporters.md>
 API Authentication <./reference/api-authentication.md>
 Interactive Models <./reference/interactive-models.md>
 API Server Endpoints <./reference/api-server-endpoints.md>
-Websockets <./reference/websockets.md>
+WebSockets <./reference/websockets.md>
 Command Line Interface (CLI) <./reference/cli.md>
 Cursor Rules Reference <./reference/cursor-rules-reference.md>
 Evaluation <./reference/evaluate.md>

docs/source/quick-start/launching-ui.md

@@ -16,43 +16,54 @@ limitations under the License.
 -->
 
 # Launching the NVIDIA NeMo Agent Toolkit API Server and User Interface
+
 NVIDIA NeMo Agent toolkit provides a user interface for interacting with your running workflow. This guide walks you through starting the API server and launching the web-based user interface to interact with your workflows.
 
 ## User Interface Features
-- Chat history
-- Interact with a workflow via the HTTP API
-- Interact with a workflow via a WebSocket
-- Enable or disable a workflow's intermediate steps
-- Expand all workflow intermediate steps by default
-- Override intermediate steps with the same ID
+
+- Modern and responsive user interface
+- Real-time streaming responses
+- Human-in-the-loop workflow support
+- Chat history and conversation management
+- Light and Dark theme support
+- WebSocket and HTTP API integration
+- Intermediate steps visualization
+- Docker deployment support
 
 ## Walk-through
+
 This walk-through guides you through the steps to set up and configure the NeMo Agent toolkit user interface.
 
 ### Prerequisites
+
 Before starting, ensure you have:
+
 - NeMo Agent toolkit installed and configured
 - Set up the simple calculator workflow by following the instructions in `examples/getting_started/simple_calculator/README.md`
 - Node.js v18+ installed (required for the web interface)
 
-
 The NeMo Agent toolkit UI is located in a Git submodule at `external/nat-ui`. Ensure you have checked out all of the Git submodules by running the following:
+
 ```bash
 git submodule update --init --recursive
 ```
 
 ### Start the NeMo Agent Toolkit Server
+
 You can start the NeMo Agent toolkit server using the `nat serve` command with the appropriate configuration file.
 
 ```bash
 nat serve --config_file=examples/getting_started/simple_calculator/configs/config.yml
 ```
+
 Running this command will produce the expected output as shown below (truncated for brevity):
+
 ```bash
 INFO:     Uvicorn running on http://localhost:8000 (Press CTRL+C to quit)
 ```
 
 ### Verify the NeMo Agent Toolkit Server is Running
+
 After the server is running, you can make HTTP requests to interact with the workflow. This step confirms that the server is properly configured and can process requests.
 
 ```bash
@@ -66,47 +77,76 @@ curl --request POST \
 ```
 
 Running this command will produce the following expected output:
-> **Note:** The response depends on the current time of day that the command is run.
+:::note
+The response depends on the current time of day that the command is run.
+:::
 ```bash
 {
   "value": "No, 8 is less than the current hour of the day (4)."
 }
 ```
 
 ### Launch the NeMo Agent Toolkit User Interface
+
 After the NeMo Agent toolkit server starts, launch the web user interface. Launching the UI requires that Node.js v18+ is installed. Instructions for downloading and installing Node.js can be found in the official [Node.js documentation](https://nodejs.org/en/download).
 
 For comprehensive information about the NeMo Agent Toolkit UI, including setup instructions, configuration options, and UI components documentation, see:
+
 - [NeMo Agent Toolkit UI README](https://github.com/NVIDIA/NeMo-Agent-Toolkit-UI/blob/main/README.md) - Complete UI documentation and setup guide
 - [UI Components Documentation](https://github.com/NVIDIA/NeMo-Agent-Toolkit-UI/tree/main/docs/ui) - Detailed information about components, features, and interface elements
 
+#### Local Development
+
 ```bash
 cd external/nat-ui
-npm install
+npm ci
 npm run dev
 ```
+
 After the web development server starts, open a web browser and navigate to [`http://localhost:3000/`](http://localhost:3000/).
 Port `3001` is an alternative port if port `3000` (default) is in use.
 
 ![NeMo Agent toolkit Web User Interface](../_static/ui_home_page.png)
 
+:::important
+Workflows requiring human input or interaction (such as human-in-the-loop workflows, OAuth authentication, or interactive prompts) must use WebSocket connections. HTTP requests are the default method of communication, but human-in-the-loop functionality is not supported through HTTP. Ensure that `WebSocket` mode is enabled in the UI by navigating to the top-right corner and selecting the `WebSocket` option in the arrow pop-out.
+:::
+
 ### Connect the User Interface to the NeMo Agent Toolkit Server Using HTTP API
+
 Configure the settings by selecting the *Settings* icon located on the bottom left corner of the home page.
 
 ![NeMo Agent toolkit Web UI Settings](../_static/ui_generate_example_settings.png)
 
 #### Settings Options
-**Note:** It's recommended to select `/chat/stream` for intermediate results streaming.
-- `Theme`: Light or Dark Theme.
-- `HTTP URL for Chat Completion`: REST API enpoint.
-  - /generate
-  - /generate/stream
-  - /chat
-  - /chat/stream
-- `WebSocket URL for Completion`: WebSocket URL to connect to running NeMo Agent toolkit server.
-- `WebSocket Schema`: Workflow schema type over WebSocket connection.
+
+**Appearance:**
+
+- `Theme`: Switch between Light and Dark mode
+
+**API Configuration:**
+
+- `HTTP Endpoint`: Select API endpoint type:
+  - **Chat Completions — Streaming** - Real-time OpenAI Chat Completions compatible API endpoint with streaming responses (recommended for intermediate results)
+  - **Chat Completions — Non-Streaming** - Standard OpenAI Chat Completions compatible API endpoint
+  - **Generate — Streaming** - Text generation with streaming
+  - **Generate — Non-Streaming** - Standard text generation
+- `Optional Generation Parameters`: OpenAI Chat Completions compatible JSON parameters that can be added to the request body (available for chat endpoints)
+
+**WebSocket Configuration:**
+
+- `WebSocket Schema`: Select schema for real-time connections:
+  - **Chat Completions — Streaming** - Streaming chat over WebSocket (recommended for intermediate results)
+  - **Chat Completions — Non-Streaming** - Non-streaming chat over WebSocket
+  - **Generate — Streaming** - Streaming generation over WebSocket
+  - **Generate — Non-Streaming** - Non-streaming generation over WebSocket
+
+:::note
+For intermediate results streaming, use **Chat Completions — Streaming** (`/chat/stream`) or **Generate — Streaming** (`/generate/stream`).
+:::
 
 ### Simple Calculator Example Conversation
+
 Interact with the chat interface by prompting the Agent with the
 message: `Is 4 + 4 greater than the current hour of the day?`
 

docs/source/reference/evaluate-api.md

@@ -23,7 +23,7 @@ It is recommended that the [Evaluating NeMo Agent toolkit Workflows](./evaluate.
 The evaluation endpoint can be used to start evaluation jobs on a remote NeMo Agent toolkit server. This endpoint is only available when the `async_endpoints` optional dependency extra is installed. For users installing from source, this can be done by running `uv pip install -e '.[async_endpoints]'` from the root directory of the NeMo Agent toolkit library. Similarly, for users installing from PyPI, this can be done by running `pip install "nvidia-nat[async_endpoints]"`.
 
 ## Evaluation Endpoint Overview
-```{mermaid}
+```mermaid
 graph TD
   A["POST /evaluate"] --> B["Background Job Created"]
   B --> C["GET /evaluate/job/{job_id}"]