The AI-native database built for LLM applications, providing incredibly fast vector and full-text search
Roadmap 2024 | Twitter | Discord | YouTube |
Infinity is a cutting-edge AI-native database that provides a wide range of search capabilities for rich data types such as vectors, full-text, and structured data. It provides robust support for various LLM applications, including search, recommenders, question-answering, conversational AI, copilot, content generation, and many more RAG (Retrieval-augmented Generation) applications.
Infinity comes with high performance, flexibility, ease-of-use, and many features designed to address the challenges facing the next-generation AI applications:
- Achieves 0.1 milliseconds query latency on million-scale vector datasets.
- Up to 10K QPS on million-scale vector datasets.
See the Benchmark.
Supports a fused search of multiple embeddings and full text, in addition to filtering.
Supports a wide range of data types including strings, numerics, vectors, and more.
- Intuitive Python API. See the Python client
- A single-binary architecture with no dependencies, making deployment a breeze.
- Operating system: Ubuntu 22.04 or higher
- Clang-17 or higher to support C 20 and modules
- CMake 3.28 or higher
import infinity
import infinity.index as index
from infinity.common import REMOTE_HOST
infinity_obj = infinity.connect(REMOTE_HOST)
db = infinity_obj.get_database("default")
# Drop my_table if it already exists
db.drop_table("my_table", if_exists=True)
# Create a table named "my_table"
db.create_table(
"my_table", {"num": "integer", "body": "varchar", "vec": "vector,4,float"}, None)
table.insert(
[{"num": 1, "body": "undesirable, unnecessary, and harmful", "vec": [1.0, 1.2, 0.8, 0.9]}])
table.insert(
[{"num": 2, "body": "publisher=US National Office for Harmful Algal Blooms", "vec": [4.0, 4.2, 4.3, 4.5]}])
res = table.query_builder().output(["*"]).knn("vec", [3.0, 2.8, 2.7, 3.1], "float", "ip", 2).to_pl()
print(res)
💡 For more information about the Python API, see the Python API Reference.
See Build from Source.