n8n RAG Demo Workflow: EcoHash Chat + BGE Reranker

Instantly demo Retrieval-Augmented Generation (RAG) in n8n with EcoHash

This n8n RAG Demo Workflow creates a tiny in-memory vector store, generates embeddings with EcoHash, and uses an EcoHash chat model + BGE reranker to retrieve the most relevant sentence—via an AI Agent tool call.

What this workflow does

• Starts on “Test workflow” in n8n.
• Creates five sample documents as plain text items.
• Loads the text as documents, generates embeddings with EcoHash, and inserts them into an in-memory vector store.
• Exposes the vector store as a retriever tool that returns top matches and reranks results using the EcoHash BGE reranker.
• Runs an AI Agent on the EcoHash chat model that calls the retriever tool to find the document sentence mentioning the capital of France, then outputs only the exact matching sentence.

Use cases

• RAG prototyping for n8n users who want to see end-to-end retrieval + reranking quickly.
• SaaS operators evaluating EcoHash embeddings and relevance reranking behavior on short text corpora.
• Automation engineers testing agent/tool patterns in n8n: retriever tool → rerank → exact sentence output.

Technical details

• Workflow trigger: Manual trigger (run via Test workflow in n8n).
• Nodes / stack: code, sticky note, n8nn8n-nodes-langchainagent (AI Agent), n8n-nodes-ecohashlm chat eco hash (EcoHash Chat Model), n8n-nodes-ecohash embeddings (EcoHash embeddings), and n8n-nodes-ecohash reranker (BGE reranker).
• Key requirement: Use the same in-memory vector store memory key (example: ecohash_smoke_vs) in both the insert and retrieve nodes.

Setup (requirements)

Create an EcoHash account at https://ecohash.com (new accounts include free starter credit for testing), generate an API key in https://console.ecohash.com, and add it to n8n as an EcoHash credential. Select it on the EcoHash Chat Model, Embeddings EcoHash, and EcoHash Reranker nodes.