Back to Gap Map Infrastructure / Storage & Retrieval

Infinity

open source / Overall score: 2.8

Infinity is a database built for LLM applications that indexes dense vectors, sparse vectors, tensors and full text in one engine and fuses them into a single hybrid query. Results can be reranked with reciprocal rank fusion, weighted sum or ColBERT-style tensor rerankers, and structured data is queried in the same system. It is self-hosted.

Openness

5 high confidence

5.0

license: Apache-2.0(OSI)
source: public
core features withheld: no

Apache-2.0 license body confirmed. The repository is public and unarchived and builds the whole product, and the README describes no paid tier, enterprise edition or license-gated build beside it, so source is public and the core ungated.

https://raw.githubusercontent.com/infiniflow/infinity/main/LICENSE recorded 2026-09-30
LICENSE file is the Apache-2.0 text: "Apache License" and "Version 2.0, January 2004" and "TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION".
https://raw.githubusercontent.com/infiniflow/infinity/main/README.md recorded 2026-09-30
README describes "a cutting-edge AI-native database that provides a wide range of search capabilities" and offers install via Docker, a binary or "Build from Source"; searched the body for enterprise, paid, license key and cloud: none.

Adoption

2 low confidence

2.0

Infinity's adoption reading uses GitHub stars, since no server deployment figure is published. The `infinity-sdk` package on PyPI is the Python client for a running server rather than the engine itself, so its downloads would measure client installs rather than product use, and a star does not confirm production use either.

https://api.github.com/repos/infiniflow/infinity recorded 2026-08-18
Repo metadata - stargazers_count = 4,675 - for infiniflow/infinity.

Capability

4 medium confidence

4.0

Infinity fuses dense, sparse, tensor and full-text indexes into a single hybrid query, with tensor reranking and structured-data support in one self-hosted engine. Vespa covers more of the same persist-index-retrieve path at a larger, distributed scale, which keeps Infinity one step behind it.

https://infiniflow.org/docs/dev/configurations recorded 2026-09-30
Configuration reference says the server can start "as a cluster node in `ADMIN` mode" when storage is MinIO object storage.
https://raw.githubusercontent.com/infiniflow/infinity/main/README.md recorded 2026-09-30
README says Infinity supports "a hybrid search of dense embedding, sparse embedding, tensor, and full text, in addition to filtering" and "several types of rerankers including RRF, weighted sum and **ColBERT**".

Verified 2026-09-27