The vector database for contextual and time series search. Build AI apps, find patterns in your data, and mix structured with unstructured data.
### User Query
query = 'How is Acme Inc managing its loan losses as of late?'
### Create the dense query vector
dense_query = [embedding_model.encode(query).tolist()]
### Create the sparse query vector
sparse_query = [dict(Counter(ids)) for ids in tokenizer([query],
add_special_tokens=False, max_length=None)['input_ids']]
### Hybrid search
table.search(
vectors={"sparse_index": sparse_query,
"dense_index": dense_query},
index_params={"sparse_index":{'weight':0.5} ,
"dense_index":{'weight':0.5}}, n=5
)
table.search (
dense_query,
n=3,
filter=[
("=", "sym", "AAA"),
(">", "price", "100.00"),
]
)
### Time Series Window Query Vector
q = AAA_df['price'][100].tolist()
### Search for Similar Time Series Windows
table.search(vectors={'price': [q]}, n=10, type="tss")
### Outlier Search for Anomaly Detection
res = table.search(vectors={'price': [q]}, n=-10, type="tss")[0]
### Multi Search over Multiple Temporal Windows Simultaneously
table.search(vectors={'price': [q, q2, q3]}, n=10, type="tss")
Ability to handle GenAI complexities for modeling unstructured data such as text, video, audio, and images.
Unify multiple indexes at once for multi-layered embeddings creating flexibility and faster search.
Scale with purpose-built qHNSW and qFlat indexing to lower costs and memory requirements.
Perform search 17x faster with 12x less memory than HNSW without the need for embeddings for fast changing temporal data.
Reduce memory and on-disk storage by 100x for slow changing time-based data sets and accelerate search by 10x.
Combines similarity, exact, and literal search in a single query where query results remain relevant with content changes.
Get hands on with our code repositories and try out sample projects.
Get fast responses to your questions from community and KX experts.
New to vector databases? Master the basics and explore key use cases like semantic search, recommendation systems, and anomaly detection.
Compress and search over massive, fast-flowing time series datasets.
Combine semantic and keyword vector searches for increased search relevance.
Filter vectors based on unlimited metadata to refine search accuracy with structured data.
Understand how to optimize your retrieval of vectors in a variety of scenarios.
Build internal knowledge bases that extend the power of large language models.
Chunk documents and inject relevant data into LLMs to reduce hallucination.
Build real-world AI projects with our commercial offering configured to your specs.