This add-on is operated by Stratis, LLC

Compose Agentic experiences, Extend Agentforce, deliver Private AI.

Last updated February 13, 2025

Stratis AI is an add-on that brings conversational AI to your Heroku applications. With Stratis AI, you can enable natural language interactions with your data, websites, and documents, to provide users access to knowledge and insights.

Adding Stratis AI to your Heroku app can:

• Turn Websites Into Knowledge: Make your website’s information accessible through conversational interactions.
• Search Documents Instantly: Integrate AWS S3 documents for instant AI-powered search.
• Talk to Your Heroku Database: Query Heroku Postgres databases in plain language with no SQL required (available at GA).
• Engage Conversationally with Data Cloud & Customer 360: Unlock Salesforce Data Cloud insights with conversational access (available at GA).
• Build Custom Agents: Create agentic experiences tailored to your business needs.
• Bring the Conversation Anywhere: Deploy chat functionality to any Heroku app with a simple JavaScript integration.
• Converse in Any Language: Create agentic experiences in any language around the globe.

Getting Started: Core Concepts

Here are some Stratis AI concepts to get familiar with:

• An Agent is a combination of one of more Actions that powers a digital assistant or chatbot.
• An Action determines what your agents can do with a Knowledge Source.
• Knowledge Sources are where an agent retrieves information from. In the beta, you can create knowledge sources from documents, such as a S3 drive, and web content, such as website URLs.

In the beta release, you can create an agentic digital assistant that knows about two types of knowledge: documents and web content. After provisioning our add-on, see the Building Your Agents section where we:

• Create two Knowledge Sources: an uploaded PDF and a specified website
• Create two Actions: a doc action and a web action
• Create an Agent and add the two Actions to it
• Create a digital assistant or chatbot Deployment and see our new Agent in action.
• Optionally, deploy our Agent to a sample Heroku app using the included Heroku Button so we can test it live.

Provisioning the Add-on

Attach Stratis AI to a Heroku application via the CLI:

$ heroku addons:create stratisai:PLAN --app HEROKU_APP_NAME
Creating stratisai on ⬢ HEROKU_APP_NAME... free
All set! To get started, run `heroku addons:open stratisai -a HEROKU_APP_NAME`.

After provisioning Stratis AI, the STRATISAI_ORG_ID config var is available in the attached app’s configuration. It contains the unique identifier for the Stratis organization created for your Heroku app. For now, you can ignore this config var, but it can come in handy for referencing it during a support issue during troubleshooting. You can see the config var via the heroku config:get command:

$ heroku config:get STRATISAI_ORG_ID -a HEROKU_APP_NAME
f9b807aa-84ba-4b00-b234-00806c5f0155

Accessing the Admin Dashboard

After provision the add-on, access the admin dashboard via the browser or CLI:

Via Browser

1. Navigate to the Resources tab of your app in the Heroku Dashboard.
2. Select the Stratis AI resource to open the Stratis AI dashboard in a new window.

Via Command Line

Access the dashboard via the CLI:

$ heroku addons:open stratisai -a HEROKU_APP_NAME

Building Your Agents

The Setup Wizard guides you through configuring and deploying your conversational agents in four steps. This process ensures an easy experience, even for users with limited technical expertise.

1. Connect Your Knowledge Sources
2. Define Your Actions (Tools)
3. Assemble Agents
4. Deploying Conversational Agents

Connect Your Knowledge Sources

The Stratis AI beta supports multiple knowledge sources including:

• AWS S3 Buckets: add and schedule periodic ingestion from your S3 documents.
• Websites: enable web ingestion for real-time data updates.

To create a knowledge source:

1. Select the type of source, for example, AWS S3 or website.
2. Enter the required credentials, for example, S3 keys or API URLs.
3. Choose a synchronization schedule, for example, daily, weekly, or monthly.
4. Click Submit.

Our example shows the configuration for creating a web source with an hourly sync schedule.

You can view and edit your web sources via the list view after creating them.

After configuring your knowledge sources, an automatic process triggers to crawl the source and then vectorize the information into pgvector. During this process, you can select the knowledge source and view the progress. Usually, the crawl is very fast and depending on the size of the source, the vectorization follows shortly.

Define Your Actions (Tools)

Actions are the building blocks of your agents. They determine what your agents can do with the configured data sources. For example:

• Retrieving specific documents from AWS S3.
• Searching website content.

To create an action in the setup wizard:

1. Select a knowledge source you created.
2. Select the specific additional configuration, such as the query or function the action performs if applicable.
3. Add a meaningful name and description to help agents use it.
4. Click Submit.

Assemble Agents

Agents combine specific actions to perform meaningful tasks. You can customize agents to fit different workflows or business needs.

To create an agent in the wizard:

1. Assign one or more actions you created to the agent.
2. Select a language or processing capabilities for the agent.
3. Add optional details like a custom persona or branding logo.
4. Click Submit.

In the Agents page, you can test an ad-hoc conversation with the agent by clicking the message icon. Use this feature to validate your agent before making it available to a deployed chat interface.

Deploying Conversational Agents

Deploy your agents to interact with users via Conversational Agents. You can embed these interfaces into your website via manual code insertion or a Heroku Button.

During this beta, you can deploy your conversational agents by placing a small JavaScript snippet the Stratis AI add-on provides into your web or mobile app. If you have any questions on where to place this snippet, you can reach out to us at support@stratisglobal.com.

To simplify this process, we’re providing a sample site deployment method that includes your created agent using a Heroku Button. To see your agent running on this sample site, click the Heroku Button on the Deploy Agent Chat Interfaces page, or from Operate → Deployments in the Admin UI.

Either way, begin by creating the conversational agent in the setup wizard:

1. Add a meaningful name and icon for your chat interface.
2. Assign one or more agents you created to the chat interface.
3. Optionally, provide a list of website domains your chat interface can be on.
4. Click Submit.

Deploying via Code Change

To deploy via a code change:

1. Grab the generated code snippet from the Setup Wizard by clicking the copy icon under the Message column on the chat interface’s row. You can also visually inspect the code snippet by clicking the pencil icon.
2. Insert the code snippet into your website’s code, specifically inside the <body> of the HTML. This step varies based on your chosen web framework. Reach out at support@stratisglobal.com for help.

Deploying via Heroku Button

To deploy via a Heroku Button, click the Deploy to Heroku button on the chat interface row you want to deploy. This button opens up a new page where you enter a name for your app and click Deploy.

After the new app deploys, click the View button to chat with your agent.

Next Steps

After successfully creating and deploying your first conversational agent, you can iterate and improve it by:

• Adding knowledge sources
• Monitoring the crawling and vectorizing of those sources
• Adding more actions
• Adding configurations and parameters to how your agent performs
• Testing your agents for accuracy and performance

Removing the Add-on

Remove Stratis AI via the CLI:

$ heroku addons:destroy stratisai
Destroying stratisai-trapezoidal-84750 on ⬢ HEROKU_APP_NAME... done

Support

Submit all Stratis AI support and runtime issues via one of the Heroku Support channels. Any non-support-related issues or product feedback is welcome at support@stratisglobal.com.