Open connector to add integrations to Relevance AI.
Install the Relevance Connect CLI:
pip install relevance_connect
For local development, you can install the package from the source:
pip install -e .
Login to your Relevance AI account: Walkthrough:
relevance-connect login
Checkout the example for a complete example.
Follow these steps to create and submit a new integration:
Create a new folder for your integration and navigate to it.
Create a metadata.json
file that defines your integration's configuration:
{
"name": "Firecrawl",
"description": "Firecrawl is a tool that allows you to run firecrawl.",
"inputs": [
{
"input_name": "website_url",
"type": "string",
"title": "Website URL",
"description": "The URL of the website to crawl.",
"default": "https://www.firecrawl.dev"
},
{
"input_name": "firecrawl",
"type": "string",
"title": "Firecrawl API Key",
"description": "The API key for firecrawl.",
"metadata": {
"content_type": "api_key",
"is_fixed_param": true
}
}
],
"required": ["website_url", "firecrawl"],
"requirements": ["firecrawl"],
"icon": "https://firecrawl.dev/favicon.ico",
"long_output_mode": true,
"timeout": 300
}
- name: Name of the integration
- description: Description of what the integration does
- inputs: Array of input configurations for the integration
- required: Array of required input field names
- requirements: (Optional) Python packages required by the integration
- icon: (Optional) URL to an icon for the integration
- long_output_mode: (Optional) Set to
true
if your code returns output greater than 10 million characters - timeout: (Optional) Timeout in seconds (default: 300)
📖 For detailed information about all available input types and their schemas, see INPUTS.md.
Create a main.py
file containing your integration logic:
from firecrawl import FirecrawlApp
# Initialize the FirecrawlApp with API key from secrets
app = FirecrawlApp(api_key="{{secrets.chains_firecrawl}}")
# Use the website URL from the params
scrape_status = app.scrape_url(params['website_url'])
# Return the scraped content
return scrape_status.markdown
- Single file only: The entire integration must be in one
main.py
file - Return statement required: Your script must end with a
return
statement - Access inputs: Use the
params
dictionary to access inputs (e.g.,params["website_url"]
) - API keys: Reference API key inputs using the pattern
{{secrets.chains_XXX}}
whereXXX
is the input name
Create an inputs.json
file with test data:
{
"website_url": "https://www.example.com",
"firecrawl": "your-test-api-key"
}
Execute your integration locally:
relevance-connect run
If you want to save the integration to your Relevance AI account, you can do so with the following command:
relevance-connect save
Once your integration is working correctly, submit it to Relevance AI!
relevance-connect logout
Javascript is also supported. For javascript, packages are not supported. Checkout the js_example for an example.
relevance-connect run-js