Agentic
Learn to set up your agentic AI application with NLX in minutes
What's an agentic application?
Agentic is an advanced form of conversational AI applications that don't require predefined flows to execute processes and generation tasks. Instead, agents in your NLX workspace employ tools to perform web search, email delivery, as well as image, text, and PDF generation asynchronously. You can also provide custom workflows for advanced services or structured processes you want an agent to access. Collectively, these tools allow an agentic app to seamlessly retrieve and send information for a variety of tasks:
Searches online for information and compiles the results into a PDF
Discovers trending topics online and generates a blog post and promo image
Create an agentic app
Import the NLX agentic template into your workspace to see a sample use case you can follow as a guide.
Developing a conversational AI app that's agentic is simple:
Select Applications in your workspace menu > Choose New application
Select Agentic from the available options > Provide a name for your agentic app
Click Create application
Configuration
The Configuration tab of your agentic app allows you to provide the necessary instructions, tools, and communication method for your agent to operate.
Behavior
Choose the generative AI provider in the Provider dropdown. You may choose the built-in NLX model or select a managed LLM provider you previously integrated into your workspace
Model (optional): For managed LLM providers, you can specify the particular model you want to power your agentic AI's intelligence
Provide a complete set of instructions in the Prompt field that details the purpose and context of the agentic AI's role, its core capabilities, any guidelines you want it to adhere to while executing with given tools and requests, formatting rules or quality standards, and any other rules or notes that are pertinent to its task
Delivery
Set up the communication details for your agent
Delivery channel: Choose the delivery service that your agent will use to correspond with users > Click Save
User name: Enter a name for your agent that will be visible through its delivery channel
Email: Taken from the User name entered with the NLX suffix
[user name]@dev.nlxa.gent
Allow list: Enter all user emails who are permitted to trigger the agent through email correspondence and receive its executed output
Language: Specify the language the agent should use
Tools
Select one or more of the following to imbue your agent with different actions for completing its task:
Custom flow: Choose one or more flows created in your workspace that the agent can follow. This allows you to provide access to your business systems or other integration services using Data request nodes
Generate image: Creates a custom image
Generate PDF: Assembles content into a downloadable PDF document
Generate text: Writes short or long-form content like summaries, posts, or instructions
Search web: Looks up real-time information from the internet to answer questions or gather content
Send email: Sends a message or file to an email address with any of the outputs created
Click Save
Executions
The Executions tab reveals all tasks that have been triggered with your agent and detailed information on its time to process and ultimate status of the task:
Goal: Lists the request message a user gave to the agent
Status: Lists the current progress state as it performs a request
Completed: Finished the defined goal request
Waiting for user: Awaiting additional information from the user to complete a request
Running: Currently performing the request
Failed: An error occurred during execution and the task request could not be completed
Runtime: The length of time it takes the agent to complete a request
Timestamp: The date and time when a request came in and an execution plan was created

Deployment
An agentic application's Deployment tab allows you to construct a build that contains a package of the agent's tools, prompt, and delivery details in the state they look like at the time the build is created. You may then deploy a successful build to make it live or roll back to a previous deployment.
To create a build:
Click Deployment tab of your agent app > Select Create build
Review Validation check for critical errors or detected UX issues in custom flows. Before each new build initiates, a validation check is run to provide a preview of potential errors that may cause failed builds Detected issues are listed with descriptions and potential solutions
Select and ensure NLP engine is correct
Optional
Enable NLX Boost: Enhances the performance of your NLP engine with generative AI intent classification
Allow NLX Boost to override the NLP: Enabling this relies on NLX's built-in generative AI to detect user intent (based on your training data) and route to a flow accordingly, regardless of NLP's detected match
Click Create build
Click Deploy option next to the successful build
Optional
Deployment languages: Select the languages to include in the deployment, if multiple are available
Hosting: Host your application as a Touchpoint app via the conversational.app domain. Ideal for previewing its final look during development and sharing your app externally with collaborators. Enable the Hosting option to configure the URL (e.g.,
mybusiness.conversational.app
)
Click Create deployment
Using your Agentic app
Your agentic application runs tasks asynchronously to achieve its goal. You can interact with your agent after a successful deployment in two ways:
Email delivery channel: Open your email client and send a request to the agent’s email address (found in the Configuration tab of your agent app). Make sure:
The sender email address is on the agent’s Allow list on the Configuration tab
The sender confirmed a subscription to the agent from the email notification sent
Touchpoint app (hosted chat): Use the hosting URL found in the deployment link details to open a chat interface and talk to your agent. The hosting option must be enabled when deploying your build

Settings
Your agentic application's Settings tab allows you to update basic application information, set criteria for timeouts, and delete your application.
Open the Application you wish to edit and select its Settings tab.
Details section allows you to add or edit an optional description and resource tags
Agentic section allows you to set a custom timeout (in seconds) for connecting to the AI model service. You may also specify the number of successive attempts you permit connection attempts before the request is marked as a fail
Danger zone allows you to delete your agent application. Note that custom flows will not be deleted and remain available for use with other applications in your workspace
Last updated