Prompt Chain

Build Adaptive RAG with Cohere Command R

Adaptive RAG with Cohere Command R, routing across web search, iterative RAG, or LLM answers for efficient information retrieval.

Works with cohere

91
Spark score
out of 100
Updated 3 months ago
Version 1.0.0

Add to Favorites

Why it matters

Implement an adaptive Retrieval Augmented Generation (RAG) system using Cohere's Command R model. This system intelligently routes queries to optimize retrieval strategies, including web search and iterative RAG, for more accurate and efficient information retrieval.

Outcomes

What it gets done

01

Analyze queries to determine optimal retrieval strategy (no retrieval, web search, or iterative RAG).

02

Integrate Cohere Command R for advanced query understanding and tool utilization.

03

Orchestrate multi-phase RAG processes for complex information needs.

04

Leverage web search capabilities for dynamic information gathering.

Install

Add it to your toolbox

Run in your project directory:

curl -fsSL https://spark.entire.vc/get/lg-langgraphadaptiveragcohere | bash

Capabilities

What this chain does

Search the web

Searches the web and retrieves relevant sources.

RAG index

Chunks, embeds, and indexes documents for semantic retrieval.

Summarize

Condenses long documents or threads into key takeaways.

Classify

Labels or categorizes text, files, or data points.

Overview

Adaptive RAG Cohere Command R

What it does

This asset implements an Adaptive RAG strategy using Cohere Command R. It performs query analysis to route information retrieval tasks across three distinct paths: direct LLM answers (No Retrieval), web searches, or iterative RAG processes. This approach aims to optimize accuracy and efficiency in answering user queries.

How it connects

Utilize this asset when you need a sophisticated RAG system that can dynamically adapt its retrieval strategy based on the nature of the query. It is particularly useful for complex questions requiring external information (web search) or multi-step reasoning (iterative RAG), while also efficiently handling simpler queries with direct LLM responses.

Source README

This directory is retained purely for archival purposes and is no longer updated. Please see the newly consolidated LangChain documentation for the most current information and resources.

Adaptive RAG Cohere Command R

Adaptive RAG is a strategy for RAG that unites (1) query analysis with (2) active / self-corrective RAG.

In the paper, they report query analysis to route across:

  • No Retrieval (LLM answers)
  • Single-shot RAG
  • Iterative RAG

Let's build on this to perform query analysis to route across some more interesting cases:

  • No Retrieval (LLM answers)
  • Web-search
  • Iterative RAG

We'll use Command R, a recent release from Cohere that:

  • Has strong accuracy on RAG and Tool Use
  • Has 128k context
  • Has low latency

Environment

Index

LLMs

We use a router to pick between tools.

Cohere model decides which tool(s) to call, as well as the how to query them.

Generate

Web Search Tool

Graph

Capture the flow in as a graph.

Graph state

Graph Flow

Build Graph

Trace:

https://smith.langchain.com/public/623da7bb-84a7-4e53-a63e-7ccd77fb9be5/r

Trace:

https://smith.langchain.com/public/57f3973b-6879-4fbe-ae31-9ae524c3a697/r

Trace:

https://smith.langchain.com/public/1f628ee4-8d2d-451e-aeb1-5d5e0ede2b4f/r

Discussion

Questions & comments · 0

Sign In Sign in to leave a comment.