Drafter AI's purpose is to make AI features to implement as easy as possible for no-code developers.
Let's figure out, how Drafter AI works.
Drafter AI consists of 3 major parts:
Workflow (for creators only)
A core of Drafter AI, where the logic of the algorithm is defined.
You can maintain multiple workflows and switch between them using the left sidebar:
Input (for all users)
A place for you or your users to input necessary data in the form for the logic to start working.
Output (for all users)
It's where final results are displayed, filtered and can be exported for further use.
Let's build a simple web page's keyword extractor.
Its purpose is to get metadata from any web page and identify keywords from it.
Part 1: Create
Start creating a new workflow by clicking the "+" in the bottom left corner.
Provide the next information about the new workflow:
- Workflow Name - The name to identify workflow to get to it inside the system and from outside integration
- Workflow Status - when you start working on the workflow, it's "Draft". nobody except the owner can access and use it. When it's tested and works, switch to "Published"
- Workflow Description (optional) - to describe what the workflow does for yourself and other users
- Workflow Icon - to identify the workflow easier at the workflow navigation sidebar.
When the data is filled - click "Start Creating" to proceed.
Part 2: Design
After the workflow is created, you'll see the core of Drafter AI - workflow builder.
It enables you to define how your workflow should work.
To do that, you need 2 things to do:
As for today, there are two types of input (many more to come later):
- Manual Input. For single operations and 3rd party integrations. It generates a manual form for the users to input data on their own. In this case, define what data do you need to process the workflow.
- Dataset. To scale workflow execution for multiple entries. Just upload a CSV table with defined column names and multiple entries in it to process workflow for every entry automatically.
In this tutorial let's start with manual input.
Step 1. Add "Form/Manual" block
Drag and drop the "Form / Manual" block from the "Inputs Types" area on the right side to the builder area.
Step 2. Provide general info
Click on the block and provide the general info required.
Step 3. Define input data fields
In the "Design Form" area define, what data fields do you need and select the proper data types for each field using the dropdown at left.
Important! In the current version, only one input block per workflow is allowed.
Step 4. Add actions
Actions are the blocks that define what operations should be performed using the input data.
There are a lot of them, but there is a search to make navigation easier.
In our case, let's find "Get URL Data" which extracts data from the provided web pages.
Drag and drop it to the builder area.
After that, connect dots between input and action blocks.
Step 5. Configure action
For the action blocks, we should define both input and output data.
Every action block accepts a specific data type only, depending on the tasks it should perform.
For example, to extract the data from the web page it's mandatory to have its URL (address).
Let's configure the "Get URL Data" block.
Important! For Data Input we can select any of the data outputs from the previous steps.
In this case, let's select the "Website" (URL) data field which is the output from the input block.
Having it, we can extract a lot of different information about the web page.
Let's extract metadata like title, description and favicon.
Step 6. Setup Keyword Extractor
Using the same approach, let's add the "Extract Keyword" action block.
It gets URL description from the "Get URL Data" action block as an input and extracts Keywords from it using GPT-3 engine.
Important! Refer to credits estimations in the bottom left to understand, how much time and credits could specific workflow configuration consume for a single entry.
Keep in mind that those estimations are approximate as it's impossible to estimate the amount of real data to be processed for every entry before the processing starts.
This workflow is finished.
Part 3: Test
Step 1. Proceed to input
After the workflow is ended, click the "Run Flow" button in the top right to proceed to the input form.
Important! The input form is displayed ONLY for the workflows with the "Form / Manual" type of the input.
If you use "Dataset" as an input, "Run Flow" will start workflow execution for all the entries in it instantly.
Step 2. Input data in the form
Input the necessary data (URL, in our case) using the automatically generated form.
Click "Submit" for the workflow to start.
Step 3. Wait for the workflow to be executed
It may take some seconds or minutes to get things ready, depending on the complexity of the workflow and the number of entries in the input.
Use the progress bar in the bottom left to understand the progress.
Step 4. Setup visualization
When you use workflow for the first time, your output area is empty.
After the data is processed, drag and drop visualizations in the output area.
There are 3 types of visualizations:
- Cards - several cards in the same raw to display data vertically
- Rows - full-width containers which work the best for displaying data with a lot of texts or large images in it
- Table - a classic table that displays all the data available for all the entries
Important! You can use all the types of visualizations on the page together.
For example, it's possible to use "Cards" to display top-3 results and the Table for all other results.
Step 5. Display data output
When the visualizations are added, fill them with data generated during the workflow execution.
At the pane at the right you see every data field from input or generated during the workflow.
It enables you or users to display, check and compare any data needed on the go using multiple visualizations.
Important! If things went wrong - you can easily get back to the workflow builder area by clicking the "Workflow" tab (below the title)
Part 4: Publish and Integrate
If you're happy with the results - you have plenty of options to use them.
To provide access to the workflow for the other users and 3rd party apps (through API or integrations) the workflow should be published.
To do that, open the "General Settings" tab and change "Workflow Status" from "Draft" to "Publish".
Don't forget to click on "Apply Changes" to finalize.
In the case of processing a dataset as an input, you get a lot of data as the output.
You can export output data as CSV for further use or re-use on other apps.
Access via API (Advanced)
Warning: it requires some development skills (low-code).
Every workflow has a pre-generated API call to copy-paste by developers.
Access it by clicking the "Generate API Call" in the top right.
The last thing you need to make it happen is Access Key.
It's also important for Zapier integration.
You can find it at "Account" > "Settings" > "Integrations" section.
After the key is generated, just copy it to paste anywhere you need.
Important! For security reasons, you can generate and copy it once.
Integrate using Zapier
We have published a Zapier connector, so you're able to get input data from and send output data (results) to any of the apps integrated with Zapier.
Drafter AI - Zapier Tutorial coming soon.
That's all folks.
Let's get started!