Welcome to the Marqo Cookbook!

The Marqo Cookbook is a collection of recipes, best practices, and tips for using Marqo to build vector search backed applications.

The cookbook is designed to be a practical guide to help you get the most out of Marqo and answer common questions. The guides contain a mix of Marqo specific information as well as concepts and techniques that are generally applicable.

All guide expect that you have Marqo running on localhost using the default port of 8882 and have the Python client installed.

The client can be installed using pip:

pip install marqo

Marqo can be run using Docker:

docker run --name marqo -it -p 8882:8882 marqoai/marqo:latest

Table of Contents

• Model Selection:
  • Text Search
  • Multimodal Search
• Best Practices:
  • Score Modifier Best Practices
  • Multimodal Combination Field Best Practices
  • Text Pre-Processing Best Practices
  • Query Prompt Engineering
  • Multi-term Queries
  • Analysing Processing Time
• Recipes:
  • Lexical Reciprocal Rank Fusion
  • Similar Item Recommendations
  • Diversifying Recommendations
  • Semantic Filtering Templates
  • Personalised Search and Recommendations
  • Embedding Static Context
  • Calculating Recall
  • Content Moderation
  • Duplicate Detection
  • Loading Sentence Transformers from Hugging Face
• Index Types:
  • Unstructured Vs Structured Indexes
  • Creating an Unstructured Index:
    • Text Only Unstructured Index
    • Multimodal Unstructured Index
    • Filtering in Unstructured Indexes
  • Creating a Structured Index:
    • Text Only Structured Index
    • Multimodal Structured Index
    • Structured Index with Other Features
  • Understanding HNSW Parameters