How to Use Flex to Build Custom Content Workflows

We’ve just launched Flex, a new feature now available for beta testing in the app. Since it’s still in beta, do expect some hiccups while we continue refining it. Please share your feedback at hello@surgegraph.io to help us improve Flex.

This quick guide covers everything you need to start using Flex. Here’s what we’ll cover:

• What is Flex
• Definition of Terminologies
• How to Build a Workflow
• Custom Prompts and Variables
• Tips

What is Flex?

Flex is a drag-and-drop custom workflow builder that lets you create a step-by-step process for generating content.

Think of it as a flowchart: each step, or “building block”, handles a specific task, like crafting a title, creating an outline, or writing a full article.

You can control the sequence, connect the blocks to direct the flow of information, and customize each one with your own prompts.

Definition of Terminologies

Here’s a breakdown of key terms you’ll encounter in Flex, organized for quick reference:

Workflow Components

• Building Blocks: Cards you drag and drop into the builder. Each represents a step or task (e.g., generating a title).
• Sockets: Connection points between building blocks. They control the sequence of the workflow and how information flows from one step to the next.
• Variables: A placeholder used to automatically pull and insert values in your prompts and workflow. Variables allow you to reuse data across different steps and automate inputs without manual entry. (learn more in the Custom Prompts and Variables section)
• System Variable: A pre-defined variable that already exists in SurgeGraph. For example, keyword, author, and headings. (learn more in the Custom Prompts and Variables section)
• Custom Variable: A user-defined variable that allows you to set and reuse specific values within your workflow. In other words, it’s a new, unique variable you can define yourself. (learn more in the Custom Prompts and Variables section)
• System Message: An instruction given to the AI model that sets the overall behavior, tone, or role it should adopt when generating content. It provides context and guidelines that influence how the AI responds but is not visible in the final output. System messages are used in most LLMs like ChatGPT and Claude. For example, take a look at Claude’s System Message here: https://docs.anthropic.com/en/release-notes/system-prompts
• User Prompt: The direct instruction that tells the AI what to generate. It defines the specific task, request, or instruction for the AI to follow and directly impacts the output produced.

Building Blocks

• Input: The starting point of your workflow where you feed initial data (eg. the keyword).
• Data: External information fed into the workflow.
  • SERP Research: Crawls the top-ranking pages for your keyword and retrieves the headings and Contextual Terms (commonly found keywords).
• Models: The AI model used to perform tasks for any of the Prompt blocks. Currently, we support OpenAI and Gemini, but more models will be added soon.
• Logic: Customizes workflow behavior.
  • Conditional: Adjusts flow based on conditions you specify.
  • Merge: Combines multiple inputs from multiple blocks into one.
• Prompts: Performs tasks based on prompts you give the AI. The output flows to the next connected block.
  • Custom Prompt: Create a custom task using your own prompt.
  • Title: Generate a title based on your keyword.
  • Content Goal: Defines and sets the purpose of your content.
  • Outline: Generates an outline with headings (H2, H3) only.
  • Talking Point: Generates key points for each heading to guide the writing.
  • Longform: Generates the long-form article section by section (eg. introduction, H2s, conclusion), then combines all the sections into the final article.

• Outputs: Decide what happens to your final output. You can add multiple Output blocks.
  • Save as Document: Stores the output in the Vertex dashboard. Each socket (Title, Content Goal, Outline, Longform) must connect to a respective block (e.g., Title block for the title socket). Unconnected sockets will result in that part being blank.
  • Export: Downloads the output in your chosen format.
  • Webhook: Sends the output to an external system (eg. WordPress or any CMS).
• Preset Prompts: Reusable custom prompts you created in your workflow.

How to Build a Workflow

Follow these steps to create a custom workflow in Flex:

1. Start with Input, End with Output

Every workflow begins with an Input block and ends with an Output block.

The Input block has two sockets:

• Research Socket: Connect this to a Data socket. (currently, only SERP Research is supported – more options coming soon.)
• Starting Point Socket: Link this to the next building block you want to start with, like a Prompt block (e.g., Title or Outline).

2. Fill in the Middle With Your Process

Next, build your workflow by dragging and dropping building blocks in the order you want. Each block handles a task, like creating an outline.

Connect blocks using sockets to define the flow. The output of one block (eg. outline) feeds into the next block (eg. generating content).

You can either choose from the preset Prompt blocks (Title, Outline, Longform) for ready-made tasks, or create a unique task with your own instructions using a Custom Prompt block.

3. Choose the LLM Model For Each Prompt Block

Every Prompt block has an LLM model socket.

Select the LLM model that handles each task by dragging an LLM model block (e.g., GPT-4o) and connecting it to the socket.

4. Test and Execute

When your workflow is ready, click on Execute.

To ensure it works, you can perform a Test Execution first.

First, on the top right corner, toggle to “Test” and click on “Test Execution”. (Note: you can’t edit the keyword or other input here, as it’s only a test).

Then, check the State column. Green means the workflow ran successfully with no errors. Red means there’s an issue to fix (e.g., unconnected sockets or invalid prompts).

Note: Test runs don’t generate output. They only confirm whether the workflow can run successfully.

Once you’re ready to run the execution, toggle to “Live”, fill in the keyword, adjust author or intent if needed (expand via the Action button), and click Execute.

You can track the execution status as it is processed. Once it’s done, you’ll see a link in the Output column.

The output depends on your Output block. If you added a Save as Document block, the content will appear in your Vertex dashboard’s document list.

Custom Prompts and Variables

If you choose to use the preset blocks like Outline or Longform, you can use our default prompt or write your own prompts for specific tasks.

For example, in the Longform block, you can use your own prompt for generating the introduction like so:

“Generate an engaging introduction for a blog post on #{keyword} using the problem-solution hook framework. Start by highlighting a common pain point related to #{keyword}, then introduce the blog post as the solution. Only output the introduction. Do not include explanations, instructions, or any other text, including quotation marks.”

Additionally, you can build a unique task using the Custom Prompt block. This block allows you to write your own prompt that instructs the AI on what to do.

For example, you can create a custom prompt that generates the title like so:

“Generate a compelling blog post title for #{keyword} using one of the following frameworks:

1. The Question Hook – Pose a thought-provoking question that sparks curiosity.

2. The Benefit-Driven Title – Highlight a key advantage or solution the post provides.

Choose the most suitable framework for #{keyword}, ensuring the title is clear, engaging, and optimized for reader interest. Only output the title. Do not include explanations, instructions, or any other text, including quotation marks.”

You can use variables in your prompts for easier information handling.

Variables

Variables are placeholders you insert into prompts to pull in specific data, like keyword or headings.

There are two types of Variables:

1. System Variables

These are predefined by us and are ready to use. The format is: #{system_variable}.

Here’s the list:

• #{keyword}: The keyword you enter before running the workflow.
• #{location}: The location you set.
• #{language}: The language you select.
• #{reader_intent}: The reader intent you choose.
• #{author}: The author persona you select.
• #{headings}: A list of headings pulled from the top 10 ranking pages for your keyword.
• #{contextual_terms}: A list of commonly found keywords pulled from the top 10 ranking pages for your keyword.

Example prompt:

“Generate an engaging introduction for a blog post on #{keyword} using the problem-solution hook framework. Start by highlighting a common pain point related to #{keyword}, then introduce the blog post as the solution. Only output the introduction. Do not include explanations, instructions, or any other text, including quotation marks.”

What we’ll do here is pull the “keyword” from your keyword input and replace it in the prompt. For example, if your keyword input is “copywriting”, we’ll replace the #{keyword} with “copywriting”.

2. Custom Variables

You can also define and create your own variables. The format is: {custom_variable}.

When you add a custom variable to a prompt, Flex creates an input socket on that block.

For that socket, you can either:

• Connect another block to feed data into the custom variable socket.
• Or leave it blank – it’ll become a field you fill in manually before execution.

Example prompt:

“Write a product review introduction for {product_name}. Only output the introduction. Do not include explanations, instructions, or any other text, including quotation marks.”

What happens here is that a new input socket for product_name. Then, you can either:

• Connect a block that outputs the product name to the socket – let’s say it’s “Acme Widget. The prompt then becomes “Write a product review introduction for Acme Widget.”
• Or leave it blank. Before execution, you’ll enter the product name in the custom text field that appears.

Tips

Here’s a few tips to ensure your workflow run smoothly:

1. Prompting: When using custom prompts, instruct the AI to only output the result without any other text. Eg:

“Only output the title. Do not include explanations, instructions, or any other text, including quotation marks.”

You don’t have to worry about this if you use our default prompts.

2. Socket Compatibility: Not all sockets connect to every block. For example, Prompt blocks have an LLM model socket that only links to an LLM model block.

3. Default LLM Model: If you don’t connect an LLM model to a Prompt block, we’ll use GPT-4o as the default.

4. Own API Key Needed: Flex requires you use your own API key. Add your API key in Settings > Organization Settings > LLM API Key.

Try Flex & Share Your Feedback

And that’s all you need to know to get started with Flex – do try it out!

Since it’s still in beta, we’re fine-tuning it to improve it, plus more features are coming soon.

If you run into hiccups or have ideas, please share your feedback at hello@surgegraph.io. Your input helps us improve Flex for you!

Log in to your SurgeGraph account to start using Flex here: https://app.surgegraph.io