Skip to content

Latest commit

 

History

History
107 lines (80 loc) · 5.34 KB

README.md

File metadata and controls

107 lines (80 loc) · 5.34 KB

Spring AI Content Creation Workflow with Camunda, Tavily, and SendGrid

This project demonstrates a content creation workflow orchestrated by Camunda, leveraging Spring AI, Tavily Search API, and SendGrid. The workflow automates the process of generating a blog post based on a user-defined topic, searching for relevant information, writing the post, formatting it as an HTML email, and sending the email.

Workflow Overview

The workflow, defined in content_creation_workflow.bpmn, consists of the following steps:

  1. Start Event: The workflow begins with a start event, triggered manually or by an external system.
  2. Search Content (Service Task - searchAgent):
    • Uses Spring AI's ChatClient to formulate a structured search query for the Tavily Search API based on a user-provided topic (defaulting to "Spring AI and Camunda").
    • Calls the Tavily API (via TavilyApiClient) to retrieve search results.
    • Stores the search results in a process variable (searchResults).
  3. Write Blog Post (Service Task - writerAgent):
    • Uses Spring AI's ChatClient with a system prompt designed to instruct the AI to write a concise blog post (under 250 words).
    • Takes the user topic and search results as input.
    • Generates the blog post content.
    • Stores the blog post content in a process variable (blogPostContent).
  4. Convert to Email (Service Task - emailAgent):
    • Uses Spring AI's ChatClient with a system prompt to convert the plain-text blog post into well-formatted HTML, suitable for an email. It uses <h2>, <p>, <ul>, and <li> tags.
    • Takes the blog post content as input.
    • Generates the HTML email content.
    • Stores the HTML content in a process variable (emailContent).
    • Calls the SendGrid API (via the sendEmail method) to send the formatted email to a predefined recipient.
  5. End Event: The workflow concludes after the email has been sent.

Prerequisites

  • Java 17 or higher.
  • Camunda Spring-boot.
  • OpenAI API Key (set as environment variable OPENAI_API_KEY).
  • Tavily Search API Key (set as environment variable TAVILY_API_KEY).
  • SendGrid API Key (set as environment variable SENDGRID_API_KEY).
  • Define two environment variables (EMAIL_RECIPIENT, EMAIL_SENDER).

Getting Started

  1. Clone the repository:

    git clone https://github.com/apappascs/spring-ai-examples.git
cd spring-ai-examples/spring-ai-workflow-camunda

  2. Set environment variables:

    export OPENAI_API_KEY=<your-openai-api-key>
export TAVILY_API_KEY=<your-tavily-api-key>
export SENDGRID_API_KEY=<your-sendgrid-api-key>
export EMAIL_RECIPIENT=<recipient-email-address>
export EMAIL_SENDER=<sender-email-address>

  3. Build the application (using Maven):

    mvn clean install

  4. Run the Spring Boot application:

    mvn spring-boot:run

  5. Start a workflow instance:

    • You can start a new workflow instance through the Camunda Tasklist, Cockpit, or REST API.

    • Example using the REST API (replace with your Camunda deployment URL):

      curl -X POST \
  http://localhost:8080/engine-rest/process-definition/key/spring-ai-workflow-camunda-process/start \
  -H 'Content-Type: application/json' \
  -d '{}'

      This will start a new instance of the workflow, triggering the search, blog post generation, and email sending process.

Key Components

  • content_creation_workflow.bpmn: The BPMN 2.0 workflow definition. This file can be opened and edited with the Camunda Modeler.
  • service/ package: Contains the Java delegates that implement the workflow's service tasks.
    • SearchAgent.java: Formulates the Tavily search query and retrieves search results.
    • WriterAgent.java: Generates the blog post content.
    • EmailFormatterAgent.java: Converts the blog post to HTML and sends the email via SendGrid.
    • TavilyApiClient.java: A client for interacting with the Tavily Search API, handling requests and responses.
  • application.yaml: Contains configuration for Spring AI (OpenAI), SendGrid, and Tavily API keys, along with Camunda administrator credentials.

Technologies Used

Notes

  • Remember to correctly set the environment variables mentioned above with the corresponding API keys.
  • The prompts used by the ChatClient in each service task can be further refined to improve the quality of the generated content.

Contributing

Contributions are welcome! Please feel free to submit pull requests or open issues to improve the project.