Add Tools to a Workflow

The Customizing a Workflow tutorial demonstrates how to customize a workflow by overriding parameters. This tutorial will demonstrate how to add new tools to a workflow. Adding a new tool to a workflow requires copying and modifying the workflow configuration file, which, in effect, creates a new customized workflow.

NeMo Agent toolkit includes several built-in tools (functions) that can be used in any workflow. To query for a list of installed tools, run the following command:

nat info components -t function

The examples/getting_started/simple_web_query/configs/config.yml workflow defines a tool to query the LangSmith User Guide. This is defined in the functions section of the configuration file:

functions:
  webpage_query:
    _type: webpage_query
    webpage_url: https://docs.smith.langchain.com
    description: "Search for information about LangSmith. For any questions about LangSmith, you must use this tool!"
    embedder_name: nv-embedqa-e5-v5
    chunk_size: 512

However, the workflow is unaware of some related technologies, such as LangChain, if you run:

nat run --config_file examples/getting_started/simple_web_query/configs/config.yml --input "How do I trace only specific parts of my LangChain application?"

The output may be similar to the following:

Workflow Result:
["Unfortunately, the provided webpages do not provide specific instructions on how to trace only specific parts of a LangChain application using LangSmith. However, they do provide information on how to set up LangSmith tracing with LangChain and how to use LangSmith's observability features to analyze traces and configure metrics, dashboards, and alerts. It is recommended to refer to the how-to guide for setting up LangSmith with LangChain or LangGraph for more information."]

You can solve this by updating the workflow to also query the LangGraph Quickstart guide.

To do this, create a copy of the original workflow configuration file. To add the LangGraph query tool to the workflow, update the YAML file updating the functions section from:

functions:
  webpage_query:
    _type: webpage_query
    webpage_url: https://docs.smith.langchain.com
    description: "Search for information about LangSmith. For any questions about LangSmith, you must use this tool!"
    embedder_name: nv-embedqa-e5-v5
    chunk_size: 512

to:

functions:
  langsmith_query:
    _type: webpage_query
    webpage_url: https://docs.smith.langchain.com
    description: "Search for information about LangSmith. For any questions about LangSmith, you must use this tool!"
    embedder_name: nv-embedqa-e5-v5
    chunk_size: 512
  langchain_query:
    _type: webpage_query
    webpage_url: https://docs.smith.langchain.com/observability/how_to_guides/trace_with_langchain
    description: "Search for information about LangChain. For any questions about LangChain, you must use this tool!"
    embedder_name: nv-embedqa-e5-v5
    chunk_size: 512

Since you now have two instances of the webpage_query tool, you need to update the name of the first tool to langsmith_query.

Finally, update the workflow.tool_names section to include the new tool from:

workflow:
  _type: react_agent
  tool_names: [webpage_query, current_datetime]

to:

workflow:
  _type: react_agent
  tool_names: [langsmith_query, langchain_query, current_datetime]

Note

The resulting YAML is located at examples/documentation_guides/workflows/custom_workflow/custom_config.yml in the NeMo Agent toolkit repository.

When you rerun the workflow with the updated configuration file:

nat run --config_file examples/documentation_guides/workflows/custom_workflow/custom_config.yml \
  --input "How do I trace only specific parts of my LangChain application?"

We should receive output similar to:

Workflow Result:
['To trace only specific parts of a LangChain application, you can either manually pass in a LangChainTracer instance as a callback or use the tracing_v2_enabled context manager. Additionally, you can configure a LangChainTracer instance to trace a specific invocation.']

Alternate Method Using a Web Search Tool

Adding individual web pages to a workflow can be cumbersome, especially when dealing with multiple web pages. An alternative method is to use a web search tool. One of the tools available in NeMo Agent toolkit is the tavily_internet_search tool, which utilizes the Tavily Search API.

The tavily_internet_search tool is part of the nvidia-nat[langchain] package, to install the package run:

# local package install from source
uv pip install -e '.[langchain]'

Prior to using the tavily_internet_search tool, create an account at tavily.com and obtain an API key. Once obtained, set the TAVILY_API_KEY environment variable to the API key:

export TAVILY_API_KEY=<YOUR_TAVILY_API_KEY>

We will now update the functions section of the configuration file replacing the two webpage_query tools with a single tavily_internet_search tool entry:

functions:
  internet_search:
    _type: tavily_internet_search
  current_datetime:
    _type: current_datetime

Next, update the workflow.tool_names section to include the new tool:

workflow:
  _type: react_agent
  tool_names: [internet_search, current_datetime]

The resulting configuration file is located at examples/documentation_guides/workflows/custom_workflow/search_config.yml in the NeMo Agent toolkit repository.

When you re-run the workflow with the updated configuration file:

nat run --config_file examples/documentation_guides/workflows/custom_workflow/search_config.yml \
  --input "How do I trace only specific parts of my LangChain application?"

Which will then yield a slightly different result to the same question:

Workflow Result:
['To trace only specific parts of a LangChain application, users can use the `@traceable` decorator to mark specific functions or methods as traceable. Additionally, users can configure the tracing functionality to log traces to a specific project, add metadata and tags to traces, and customize the run name and ID. Users can also use the `LangChainTracer` class to trace specific invocations or parts of their application. Furthermore, users can use the `tracing_v2_enabled` context manager to trace a specific block of code.']