Rediscovering Apollo 11 🚀

Overview

Rediscovering Apollo 11 is a Java-based project that leverages Redis OM Spring, vector search, and retrieval-augmented generation (RAG) to explore Apollo 11 mission data. This project integrates structured and unstructured data, including transcripts, photographs, and extracted information, to enable powerful semantic search and querying capabilities.

Talk

This implementation is the support demo for the Rediscovering Apollo 11: Using Spring AI + Redis OM Spring to explore the trip to the moon! talk.

How It Works

• The ./src/resources/Apollo11_Data directory contains data sourced from Apollo in Real Time, including utterances, photographs, and the table of contents.
• We use Redis OM Spring, an extension of Spring Data Redis, to efficiently load and manage this data in Redis.
• The RDB is the file that allows us to recreate our Redis database with all the data already loaded in. It can be downloaded from: https://www.dropbox.com/scl/fi/x5hz3pfnizt4tntqe3yor/dump.rdb
• Download the images from (https://www.dropbox.com/scl/fi/qanbqn9084ull141tzqxe/apollo11.zip) and unzip them into resources/static/images/ which will result in the resources/static/images/apollo11 directory being created.

Data Processing Pipeline

1. Utterance Processing:
   • Each utterance is loaded into Redis and vectorized for similarity search.
2. Table of Contents Integration:
   • The table of contents is loaded, and utterances are grouped accordingly.
3. Summarization & Vectorization:
   • An LLM generates summaries for each group of utterances, which are then vectorized for efficient retrieval.
4. Question Extraction & Vectorization:
   • The LLM extracts relevant questions from the utterances and vectorizes them for searchability.
5. Photograph Processing:
   • Photographs are indexed by vectorizing both the images and their descriptions.

Search Capabilities

We expose REST controllers that allow searching across multiple dimensions:

• Text-based Search: Retrieve utterances using natural language queries.
• Question-based Search: Find answers using extracted and vectorized questions.
• Summary-based Search: Query key mission moments through LLM-generated summaries.
• Image-text Search: Search images using textual descriptions.
• Image-based Search: Find similar images by uploading a query image.

This setup ensures fast, efficient retrieval of Apollo 11 data using Redis vector search.

🌐 API Endpoints

Method	Endpoint	Description
POST	/search-by-text	Search mission logs using text similarity
POST	/search-by-question	Search using vectorized questions
POST	/search-by-summary	Retrieve summarized reports
POST	/search-by-image-text	Find images based on text queries
POST	/search-by-image	Search for images using vector embeddings

🐳 Running with Docker Compose

To start Redis with the preloaded RDB file, run:

docker-compose up -d

The RDB file contains all the data already loaded so that you can simply query it. Download it from: https://www.dropbox.com/scl/fi/x5hz3pfnizt4tntqe3yor/dump.rdb

---

🛠 Developed by: Raphael De Lio, Developer Advocate
📢 Powered by Redis OM Spring & Redis 8 🚀