Structured Hierarchical Retrieval - LlamaIndex 🦙 0.9.20 #170
Labels
AI-Agents
Autonomous AI agents using LLMs
Algorithms
Sorting, Learning or Classifying. All algorithms go here.
Automation
Automate the things
embeddings
vector embeddings and related tools
llm
Large Language Models
llm-experiments
experiments with large language models
llm-function-calling
Function Calling with Large Language Models
RAG
Retrieval Augmented Generation for LLMs
unclassified
Choose this if none of the other labels (bar New Label) fit the content.
Structured Hierarchical Retrieval - LlamaIndex 🦙 0.9.20
Doing RAG well over multiple documents is hard. A general framework is given a user query, first select the relevant documents before selecting the content inside.
But selecting the documents can be tough - how can we dynamically select documents based on different properties depending on the user query?
In this notebook we show you our multi-document RAG architecture:
Represent each document as a concise metadata dictionary containing different properties: an extracted summary along with structured metadata.
Store this metadata dictionary as filters within a vector database.
Given a user query, first do auto-retrieval - infer the relevant semantic query and the set of filters to query this data (effectively combining text-to-SQL and semantic search).
The text was updated successfully, but these errors were encountered: