⚡️🛰️🚀The most sought after Workflow
This template demonstrates a Proof of Concept for a ReAct AI Agent that can fetch random web pages (not limited to Wikipedia or Google search results).
At the top, there is a manual chat node connected to a LangChain ReAct Agent. The agent uses a workflow tool to access and retrieve page content.
The page content extraction process begins with converting query parameters into a JSON object, with three predefined parameters:
- url – the address of the page to fetch
- method – full or simplified view of the page content
- maxlimit – maximum length for the retrieved page. If the content exceeds this length, an error message is returned.
Fetching page content involves several steps:
- An HTTP Request node attempts to retrieve the page content.
If the page content is successfully retrieved, post-processing steps include:
- Extracting the HTML BODY content
- Removing unnecessary tags to reduce the page size
- Further removing external URLs and image sources (depending on the method parameter)
- Converting the remaining HTML to Markdown, further reducing length while preserving the page structure
- If the content length is within the limit (default maxlimit = 70000; see CONFIG node), it is sent back to the Agent.
Note:
- The HTTP Request portion can be isolated into a separate workflow.
- Check the Workflow Tool description, which guides the agent to use a query string with multiple parameters instead of a JSON object.
If you need further assistance with rantir workflows and automations, please reach out to the community here: https://discord.gg/qwcYvMVt
To use this template, make sure you are on rantir version 2.0 or later.
Other Workflows like this one
Your connected stack awaits to automate AI workflows with 24-7 uptime performance and engagement