Notebooks
M
Milvus Bedrock Langchain Zilliz Rag

Bedrock Langchain Zilliz Rag

image-searchvector-databasesemantic-searchmilvusembeddingsunstructured-dataquestion-answeringLLMmilvus-bootcampdeep-learningimage-recognitionimage-classificationaudio-searchPythonbootcampragNLP
alph-notebooks/milvus-bootcamp / bedrock_langchain_zilliz_rag.ipynb

Build a RAG chain using Zilliz cloud and AWS bedrock

Combine Zilliz cloud and Aws bedrock to build a RAG(Retrieval-Augmented Generation) chain through Langchain framework.

The RAG chain consists of a retriever for retrieving relevant documents from a vector store, a prompt template for generating the input prompt, a language model for generating the AI response, and an output parser for formatting the generated response.

Zilliz cloud is used for vector storage and retrieval, and AWS bedrock is used for supporting language models and embedding models. Langchain is used to connect the components and build the RAG chain.

Prerequisites

Install the required packages and set the required environment variables.

[1]
[2]

Set the required environment variables.

[3]

The zilliz cloud uri and zilliz api key can be obtained from the Zilliz cloud console guide.

In simple terms, you can access them on your zilliz cloud cluster page.

Create LLM and Embedding models using aws bedrock

Create an aws bedrock instance and deploy language models and embedding models.

[4]

Load documents and split them into chunks

We use the Langchain WebBaseLoader to load documents from web sources and split them into chunks using the RecursiveCharacterTextSplitter.

[5]

Create the RAG chain and invoke it

We use the Langchain framework to create the RAG chain. The RAG chain consists of a retriever for retrieving relevant documents from a Zilliz vector store. When invoke from_documents function, the retriever will automatically create a Zilliz vector store from the loaded documents and embeddings.

[6]
[7]
Self-reflection is a vital capability that allows autonomous AI agents to improve iteratively by analyzing and refining their past actions, decisions, and mistakes. Some key aspects of self-reflection for AI agents include:

1. Evaluating the efficiency and effectiveness of past reasoning trajectories and action sequences to identify potential issues like inefficient planning or hallucinations (generating consecutive identical actions without progress).

2. Synthesizing observations and memories from past experiences into higher-level inferences or summaries to guide future behavior.

3. Generating reflective questions based on recent observations and attempting to answer those questions to gain insights.

4. Incorporating reflections into the agent's working memory to provide additional context for querying the language model and adjusting future plans.

For example, in the Reflexion framework (Shinn & Labash 2023), the agent computes a heuristic after each action to determine if the current trajectory is inefficient or contains hallucinations. If so, it can reset the environment and start a new trial, utilizing the self-reflections added to its memory to adjust its planning and reasoning process.

Looks great, the RAG chain answered the question through using relevant knowledge context from the web page and provide concise and correctly answers.