How to Integrate with Power Automate

Learn how to use the SnackPrompt AI Engine API to build powerful automations and AI-powered flows in Microsoft Power Automate.

Overview

Power Automate (formerly Microsoft Flow) offers several ways to integrate external APIs:

Method	Use Case	Description
HTTP Action	Direct API calls	Make HTTP requests to any API
Custom Connector	Reusable integration	Create a connector for your organization
AI Builder	AI integration	Combine with Microsoft AI capabilities
Copilot Studio	Chatbots	Build intelligent chatbots with external knowledge

Integration Architecture

┌─────────────────────────────────────────────────────────┐
│                    Power Automate                       │
│  ┌─────────────┐    ┌─────────────┐    ┌────────────┐   │
│  │   Trigger   │───▶│ HTTP Action │───▶│   Action  │   │
│  └─────────────┘    └──────┬──────┘    └────────────┘   │
│                            │                            │
│                     ┌──────▼──────┐                     │
│                     │   Parse     │                     │
│                     │    JSON     │                     │
│                     └──────┬──────┘                     │
└────────────────────────────┼────────────────────────────┘
                             │
                             ▼
              ┌──────────────────────────────┐
              │  SnackPrompt AI Engine API   │
              │  /v1/kb/search or /v1/kb/chat│
              └──────────────────────────────┘

---

Method 1: HTTP Action (Recommended)

Use the HTTP action to call the SnackPrompt AI Engine API directly.

Step 1: Create a New Flow

1. Go to make.powerautomate.com

2. Click Create > Automated cloud flow or Instant cloud flow

3. Choose your trigger (e.g., When an email arrives, When an item is created)

Step 2: Add HTTP Action

1. Click + New step

2. Search for HTTP

3. Select HTTP (not HTTP + Swagger)

Step 3: Configure the Request

For Search Endpoint:

Field	Value
Method	POST
URI	https://api-integrations.snackprompt.com/v1/kb/search

Headers:

Key	Value
Content-Type	application/json
x-api-key	@{variables('ApiKey')} or your key

Body:

{
  "query": "@{triggerBody()?['subject']}",
  "filters": {
    "tenant_id": "YOUR_TENANT_ID"
  },
  "limit": 5
}

Step 4: Parse the Response

Add Parse JSON action after HTTP:

1. Click + New step

2. Search for Parse JSON

3. Content: Select the Body from HTTP action

4. Schema: Click "Generate from sample" and paste:

{
  "items": [
    {
      "payload": {
        "original_text": "Sample text",
        "snack_item_id": "item_123"
      },
      "score": 0.95
    }
  ]
}

Step 5: Use the Results

Now you can use the parsed results in subsequent actions:

• @{body('Parse_JSON')?['items']} - All items

• @{first(body('Parse_JSON')?['items'])?['payload']?['original_text']} - First result text

---

Method 2: Chat Endpoint for Complete Responses

Use the /v1/kb/chat endpoint when you want the API to handle all the RAG.

HTTP Configuration

Field	Value
Method	POST
URI	https://api-integrations.snackprompt.com/v1/kb/chat

Headers:

Key	Value
Content-Type	application/json
x-api-key	YOUR_API_KEY

Body:

{
  "query": "@{triggerBody()?['message']}",
  "filters": {
    "tenant_id": "YOUR_TENANT_ID",
    "tag_names": ["Support", "FAQ"]
  }
}

Response

The API returns:

• answer: Ready-to-use AI response

• sources: Sources used to generate the response

Access with: @{body('HTTP')?['answer']}

---

Method 3: Custom Connector (Reusable)

Create a custom connector for your organization to simplify reuse.

Step 1: Create Custom Connector

1. Go to Data > Custom connectors

2. Click + New custom connector > Create from blank

Step 2: General Information

Field	Value
Connector name	SnackPrompt AI Engine
Host	integrations.api.snackprompt.com.br
Base URL	/v1/kb

Step 3: Security

Field	Value
Authentication type	API Key
Parameter label	API Key
Parameter name	x-api-key
Parameter location	Header

Step 4: Definition - Search Action

General:

Field	Value
Summary	Search Knowledge Base
Operation ID	SearchKnowledgeBase

Request:

• Method: POST

• URL: /search

Body Parameters:

Name	Type	Required	Description
query	string	Yes	Search query
tenant_id	string	Yes	Tenant ID
tag_names	array	No	Filter by tags
limit	integer	No	Max results

Step 5: Definition - Chat Action

General:

Field	Value
Summary	Chat with Knowledge Base
Operation ID	ChatWithKnowledgeBase

Request:

• Method: POST

• URL: /chat

Step 6: Create and Test

1. Click Create connector

2. Go to Test tab

3. Create a new connection with your API key

4. Test both operations

Step 7: Use in Flows

The connector now appears under your custom connectors:

[Trigger] → [SnackPrompt: Search Knowledge Base] → [Action]

---

Method 4: Integration with Copilot Studio

Build chatbots that use your knowledge base.

Step 1: Create a Power Automate Flow

Create a flow that can be called from Copilot Studio:

1. Create Instant cloud flow

2. Trigger: When Power Virtual Agents calls a flow

3. Add HTTP action to call SnackPrompt API

4. Add Return value(s) to Power Virtual Agents

Step 2: Configure Flow

Input from Copilot:

• UserQuestion (text) - The user's question

HTTP Action:

{
  "query": "@{triggerBody()?['text']}",
  "filters": {
    "tenant_id": "YOUR_TENANT_ID"
  }
}

Return to Copilot:

• Answer: @{body('HTTP')?['answer']}

• Sources: @{body('HTTP')?['sources']}

Step 3: Connect in Copilot Studio

1. In your Copilot, create a new Topic

2. Add Call an action > Select your flow

3. Pass the user's message

4. Display the returned answer

---

Practical Use Cases

1. Email Auto-Responder

[When email arrives] → [HTTP: /chat] → [Send reply email]

Automatically respond to customer emails using knowledge base.

2. Teams Bot Integration

[When message in Teams] → [HTTP: /search] → [Post reply in Teams]

Answer questions posted in Microsoft Teams channels.

3. SharePoint Document Assistant

[When item created in SharePoint] → [HTTP: /search] → [Update item with related docs]

Automatically find and link related documents.

4. Forms Response Handler

[When Forms response submitted] → [HTTP: /chat] → [Send email with answer]

Process form questions and send personalized responses.

5. Scheduled Knowledge Report

[Recurrence: Weekly] → [HTTP: /search] → [Create Excel row] → [Send email]

Generate weekly reports from knowledge base queries.

6. Multi-Step Support Flow

[When email arrives]
        │
        ▼
[HTTP: /search]
        │
        ▼
[Condition: items length > 0]
        │
    ┌───┴───┐
    │       │
   Yes      No
    │       │
    ▼       ▼
[Reply]  [Create Planner task]

---

Working with Variables

Store API Key Securely

Use Environment Variables or Azure Key Vault:

Option 1: Environment Variables

1. Go to Solutions

2. Add Environment variable

3. Name: SnackPromptApiKey

4. Reference: @{parameters('SnackPromptApiKey')}

Option 2: Azure Key Vault

1. Add Azure Key Vault connector

2. Get secret action

3. Use secret value in HTTP header

Initialize Variables

At the start of your flow:

[Initialize variable: ApiKey]
[Initialize variable: TenantId]

Reference in HTTP body:

{
  "filters": {
    "tenant_id": "@{variables('TenantId')}"
  }
}

---

Configuration Tips

1. Use Expressions for Dynamic Values

@{if(empty(triggerBody()?['category']),
     json('[]'),
     createArray(triggerBody()?['category']))}

2. Handle Empty Results

Add a Condition after Parse JSON:

Condition: length(body('Parse_JSON')?['items']) is greater than 0

3. Limit Results for Efficiency

{
  "limit": 3
}

4. Error Handling

Configure Run after settings:

1. Click ... on the action after HTTP

2. Select Configure run after

3. Check has failed to handle errors

Or use Scope with Try-Catch pattern:

[Scope: Try]
    └── [HTTP Action]
    └── [Process Results]
[Scope: Catch] (Run after: Try has failed)
    └── [Send error notification]

5. Apply to Each for Multiple Results

When processing all results:

[Apply to each: items]
    └── Current item: @{items('Apply_to_each')?['payload']?['original_text']}

---

Complete Example: Support Ticket Automation

Flow Overview

1. Email arrives with support question

2. Search knowledge base for answer

3. If found, send automated reply

4. If not found, create support ticket in Planner

Step-by-Step

Trigger: When a new email arrives (V3)

• Folder: Inbox

• Include Attachments: No

Action 1: HTTP

Method: POST
URI: https://api-integrations.snackprompt.com/v1/kb/chat
Headers:
  Content-Type: application/json
  x-api-key: YOUR_API_KEY
Body:
{
  "query": "@{triggerBody()?['subject']} @{triggerBody()?['bodyPreview']}",
  "filters": {
    "tenant_id": "YOUR_TENANT_ID",
    "tag_names": ["Support"]
  }
}

Action 2: Parse JSON

• Content: @{body('HTTP')}

• Schema: (generate from sample response)

Action 3: Condition

• @{length(body('Parse_JSON')?['sources'])} is greater than 0

If Yes - Send Email (V2)

• To: @{triggerBody()?['from']}

• Subject: Re: @{triggerBody()?['subject']}

• Body: @{body('Parse_JSON')?['answer']}

If No - Create Task (Planner)

• Title: Support: @{triggerBody()?['subject']}

• Details: @{triggerBody()?['body']}

---

Expressions Reference

Common Expressions

Expression	Description
@{body('HTTP')}	Full HTTP response body
@{body('HTTP')?['answer']}	Answer field from response
@{first(body('Parse_JSON')?['items'])}	First item from array
@{length(body('Parse_JSON')?['items'])}	Count of items
@{join(body('Parse_JSON')?['items'], ', ')}	Join items as string

Working with Arrays

// Get first result text
@{first(body('Parse_JSON')?['items'])?['payload']?['original_text']}

// Get all texts as array
@{body('Parse_JSON')?['items']?['payload']?['original_text']}

// Check if has results
@{greater(length(body('Parse_JSON')?['items']), 0)}

---

Troubleshooting

Error: "tenant_id is required"

Ensure tenant_id is inside the filters object:

// ❌ Wrong
{ "query": "...", "tenant_id": "..." }

// ✅ Correct
{ "query": "...", "filters": { "tenant_id": "..." } }

Error: "InvalidJson" in Parse JSON

1. Check HTTP response is valid JSON

2. Verify schema matches actual response

3. Add Compose action before Parse JSON to debug:

   • Input: @{body('HTTP')}

HTTP Action Fails

1. Check URI is correct (no extra spaces)

2. Verify headers are properly formatted

3. Test API with Postman first

4. Check firewall/network access

Empty Results

1. Verify tenant_id is correct

2. Remove tag_names filter to search all content

3. Increase limit parameter

4. Test query directly in API

Flow Runs Slowly

1. Reduce limit parameter

2. Use parallel branches where possible

3. Consider using async patterns for long operations

---

Premium vs Standard Licensing

The HTTP action requires a Premium license. Alternatives for Standard:

Feature	Premium	Standard Alternative
HTTP Action	✅	Use Custom Connector (also Premium)
Custom Connectors	✅	Request IT to create certified connector
Azure Functions	✅	Azure Functions with Standard trigger

---

Related

• Endpoints Reference

• Available Filters

• Error Hand

External Resources

• Power Automate Documentation

• HTTP Action Reference

• Custom Connectors Guide

• Copilot Studio Documentation

• Expression Reference