Hi everyone!
I’m building a customer support agent for WhatsApp using Activepieces. It works great for simple tasks, but I’m hitting a wall when it comes to giving it real memory and contextual knowledge.
In n8n, I built an agent that:
- Stores chat history in a Postgres database
- Retrieves past messages and prepends them for context
- Uses a vector store (Supabase) with OpenAI embeddings to provide business-specific knowledge as a tool to the agent
In Activepieces, I don’t see an option to pass tools like in n8n (e.g., embedding-based memory or extra knowledge sources). I’d like to migrate this agent to Activepieces and preserve that memory + knowledge capability.
My questions:
- Is it currently possible to give an Activepieces agent tools like a vector DB or external memory context?
- Is there a way to simulate memory by injecting history or vector-based results into the agent prompt?
- Are there any best practices or examples to follow for this kind of setup?
Thanks in advance! Happy to share more about my setup if helpful.
(Attached is a screenshot of what it looks like in n8n.)
