I want to build a system where one agent scrapes a LinkedIn profile and another agent compares that data to our ICP (Ideal Customer Profile). Can Flowise handle this kind of multi-agent collaboration? I’m looking specifically at the "Worker" and "Manager" agent nodes. Has anyone used these for sales automation tasks successfully?
3 answers
This is exactly what the new Agentic nodes in Flowise were designed for. You can set up a "Supervisor" node that delegates tasks to specialized workers. One worker can have a tool for Google Search or a specific scraper API, while the other is a "Judge" agent that evaluates the findings against your criteria. I’ve found that using GPT-4 for the Manager and GPT-3.5 for the Workers is a great way to save money while maintaining high-quality logic. Just be careful with scraping policies—ensure you are using a proxy service to avoid getting your IP banned.
Does the Flowise supervisor handle the "handoff" automatically, or do you have to manually define the loop for when an agent fails a task?
The "Loop" node can also be helpful here if you need to score a list of leads one by one from a CSV upload.
That’s a smart addition, Brandon. Batch processing leads through Flowise instead of one-by-one saves a lot of manual trigger time for our sales team.
It's mostly automatic, Kevin. You define the "tools" available to each worker, and the Flowise Manager decides which tool to use based on the user's prompt. It’s surprisingly effective at self-correction.