Samples
Hands on exercises to get you started
Featured
-
Temporal Similarity Search for Pattern Matching and Outliers in kdb+
Unlock deeper time-series insights on your kdb+ databases by connecting them to KDB.AI, enabling pattern matching and anomaly detection upon massive kdb+ databases.
-
Complex Document RAG with Unstructured
Use Unstructured to ingest complex documents, partition them into useful elements like text, tables, and images, chunk the elements and embed them. Next insert the elements into the KDB.AI vector database to enable retrieval and…
-
Multi-Index Multimodal Search
KDB.AI enables multiple indexes to be defined within a table. This means that the indexes can be queried and searched independently, or simultaneously. Use-cases like Hybrid Search, and Multimodal search are perfectly fit for this…
-
Fuzzy Filtering
Fuzzy filtering increases the robustness and flexibility of vector search by allowing for more error-tolerant searches. This ensures that partially matched metadata still return relevant results.
-
qHNSW Index for Memory Efficient Search
qHSNW is a memory efficient index that stores the index on-disk instead of in-memory. This increases the scalability of retrieval operations and substantially reduces the memory footprint of both vector insertion and search.
-
qFlat Index for Memory Efficient Search
Perform document search using KDB.AI’s qFlat index, a novel approach to vector indexing by storing the index on-disk rather than in-memory. This enables substantially reduced memory footprint while maintaining the accuracy benefits of flat indices.
-
LlamaParse: RAG on PDFs
Use LlamaParse to extract embedded data from complex file formats such as PDFs and build RAG pipelines upon this data.
-
KDB.AI Hugging Face Integration
The KDB.AI Hugging Face integration guide helps you execute a wide variety of tasks with your KDB.AI vector database.
-
LlamaIndex Advanced RAG
Utilize LlamaIndex and KDB.AI for a RAG pipeline to semantically search and analyze financial publications over time and make better trading decisions.
-
Transformed Temporal Similarity Search
Our model compresses time series data by over 99% while preserving its shape, enabling high-speed searches on large temporal datasets with minimal memory use.
-
Non-Transformed Temporal Similarity Search
Efficiently analyze rapid time series data with direct similarity searches on columns, bypassing embedding and indexing—perfect for identifying trends and anomalies in financial markets.
-
Multimodal Retrieval Augmented Generation
Simultaneously manage and search multiple data types in KDB.AI. Create multimodal embeddings, retrieve diverse data, and integrate it into LLM-driven response generation.
-
Metadata Filtering
Improve search performance in vector databases by incorporating metadata filtering, enhancing both the speed and precision of searches.
-
Hybrid Search
Increase the relevancy of search results by combining keyword-based sparse vector search with the semantic understanding of dense vector search.
-
Document Search
Streamline semantic searches on unstructured texts with KDB.AI. From document ingestion to running similarity searches, discover how to effectively utilize our vector database.
-
Sentiment Analysis
Extract insights from reviews using sentiment analysis combined with vector search, helping you understand customer experiences and identify areas for improvement.
-
Pattern Matching and Outliers
Apply similarity search for pattern recognition in manufacturing data, supporting quality control, process optimization, and predictive maintenance.
-
Recommendation Systems
Develop a music recommendation engine using vector embeddings from categorical and numerical music data, quickly identifying songs that match user inputs.
-
Retrieval Augmented Generation (RAG) with LangChain
Construct a comprehensive RAG pipeline with LangChain, from data ingestion to leveraging retrieved information for LLM response generation.
-
Image Search
Utilize KDB.AI to quickly find similar brain scan images stored as vector embeddings. Learn the process from image embedding generation to executing image-based queries.
-