---
id: ds
title: Data Source Node
---
The **Data Source** node allows you to connect your workflow to external data sources, such as databases, APIs, or third-party services. With the **Data Source** node, you can:
- Retrieve data from databases like PostgreSQL, MySQL, or MongoDB
- Fetch information from REST or GraphQL APIs
- Integrate with SaaS platforms such as ClickUp, Salesforce, or Slack
- Integrate AI APIs such as OpenAI or Anthropic
- Perform queries, filters, or transformations before passing data to other nodes
The **Data Source** node is essential for building dynamic workflows that depend on real-time data. It enables enterprise-level automation, reporting, and decision-making by seamlessly bringing external data into your workflow.
Each **Data Source** node has specific configurations based on its type:
- **PostgreSQL**: SQL query fields
- **REST API**: HTTP method, endpoint, headers
- **Twilio**: SMS configuration fields
- And more based on your configured sources
## Example 1 - Customer Support Automation Workflow
Consider a workflow that automatically drafts an AI generated response to a customer support ticket.
Here's an overview of the workflow.
**Input**
For this example, the workflow receives input data in the following structure:
```js
{
"from": "johndoe@example.com",
"name": "John Doe",
"data": "Hi, I've been using ToolJet and it's the best tool in the market!"
}
```
This payload comes from the source that triggers the workflow. In this case, the workflow is triggered through a Webhook, so the incoming request body provides this data. However, the same workflow could also be triggered by other sources such as another workflow or a ToolJet application as long as they send parameters in this format.
**Step 1** : From the Trigger node, drag a connection and add an OpenAI data source node. Name this node `generateResponse`.
**Step 2** : Select the _Operation_ and _Model_ and configure the prompt.
Example prompt:
```
You are a customer support representative at ToolJet. Write a response to the following ticket that the customer raised.
Ticket Data - {{startTrigger.params.data}}
Strictly provide only a JSON with the following structure without any fillers or markdown.
{
"data":
}
```
**Step 3** : Create a JavaScript node to sanitise the response.
Paste the following code into the JavaScript node. This code sanitises the response by removing any backticks and forwards only the necessary data.
````js
let openAIResponse = generateResponse.data;
openAIResponse = openAIResponse.replace("```json", "");
openAIResponse = openAIResponse.replace("```", "");
openAIResponse = JSON.parse(openAIResponse.trim());
return openAIResponse.data;
````
**Step 4** : Mail the response back to the user.
Configure an SMTP node to send the response back to the user.
**Result**
Now whenever we trigger the workflow with a support ticket, the customer will receive an automated AI response.
## Example 2 - System Health Check Alerts using Prometheus
Consider a workflow that queries a Prometheus server every 30 mins for a health check and emails the DevOps team if the system is unhealthy.
To make our workflow run, we have kept the metrics for an unhealthy system as follows:
```
CPU Usage > 0.0005%
Up Time < 95%
Memory Usage > 5MB
```
These thresholds are intentionally low so the demo can easily show an alert.
Here's an overview of the workflow:
**Step - 1** : Create a new workflow with a scheduled trigger that runs every 30 minutes into the hour.
**Step - 2** : Add three outgoing prometheus data source nodes.
We'll name these nodes as - `checkCPUUsage`, `checkUptime`, `checkMemoryUsage` and perform an **Instant Query with PromQL** operation.
To query the statistics, we'll add the following queries respectively:
1. `checkCPUUsage` Node:
```js
rate(process_cpu_seconds_total{job="prometheus"}[5m])
```
2. `checkUptime` Node:
```js
avg_over_time(up{job="prometheus"}[1h]) * 100
```
3. `checkMemoryUsage` Node:
```js
go_memstats_alloc_bytes{job="prometheus"} / 1024 / 1024
```
**Step - 3** : Add an `If condition` node to check health status named `checkOverallHealth`.
Prometheus returns results in a nested format. To get the actual metric value, we access:
`.data.data.result[0].value[1]`. This is the numeric value we compare in the If condition.
We'll add the following condition to check if the system is healthy:
```js
checkCPUUsage.data.data.result[0].value[1] > 0.0005 ||
checkUptime.data.data.result[0].value[1] < 95 ||
checkMemoryUsage.data.data.result[0].value[1] > 5;
```
If this evaluates to true, it means the system is unhealthy, else the system is healthy.
**Step - 4** : Configure the SMTP Node
Create an outgoing SMTP node from the green port of `checkOverallHealth` and configure the node.
At minimum, fill the following fields:
- **From address**: The email you want alerts to come from
- **From Name**: A friendly label (e.g., Monitoring System)
- **To/CC/BCC**: Recipients of the alert
- **Subject**: Email subject (e.g., Health Check Alert!)
- **HTML/Text**: The email content you want to send
To learn more about configuring an SMTP data source, refer [this documentation](/docs/data-sources/smtp).
**Output** :
After running the workflow, we get the following email:
