Cohere in Langchain Code Node to Improve RAG with Reranker

[Paul_Bogemann]

I am tryingt to improve my RAG results with a reranker. I do not think there is an appropriate node yet. Therefore I would like to implement it with a Langchain code node. I would want to do so using a Cohere reranker.
My question is regarding the correct steps to import the packages needed and if someone possibly has done this before?
Thanks in Advance

[n8n]

It looks like your topic is missing some important information. Could you provide the following if applicable.

• n8n version:
• Database (default: SQLite):
• n8n EXECUTIONS_PROCESS setting (default: own, main):
• Running n8n via (Docker, npm, n8n cloud, desktop app):
• Operating system:

[Jim_Le]

Hey Paul,

Great question. I’m pleased to say that the cohere libraries are available in n8n and you can use it the langchain code node like so:

(Example taken from: Cohere Rerank | 🦜️🔗 Langchain)

const { CohereRerank } = require("@langchain/cohere");

const cohereRerank = new CohereRerank({
  apiKey: '...', // You'll have to hardcode your key here
  model: "rerank-english-v2.0", // Default
});

To use it in the example from my other post on using Langchain code node, we can do:

[Deleted]

Hope this helps!

Edit: Sorry, that was a bad example!

[Jim_Le]

@Paul_Bogemann I’ve created a new Tutorial here which I think will help - Build A Custom Q&A Chain using Cohere Rerank (Using Langchain Code Node)

[Paul_Bogemann]

Hi Jim!
Thank you very much for the elaborated answer! This is a great Tutorial with strong level of detail!