Every Paradigm. One Engine.
Relational SQL
Full transactional SQL with joins, aggregations, window functions, CTEs, and subqueries. PostgreSQL wire-protocol compatible.
Full-Text Search
BM25 and Bayesian BM25 scoring with WAND/BMW optimization. 13 tokenizers, 15 token filters, spell check, and autocomplete. Custom tokenizer and token filter support.
Vector Search
HNSW-based approximate nearest neighbor search with cosine, inner product, and Euclidean distance metrics. Probabilistic score calibration.
Graph Queries
Apache AGE-compatible graph traversal via SQL table functions. BFS, DFS, shortest path, and pattern matching.
Core Innovations
UQA — Unified Query Algebra
Cognica's engine is built on a single mathematical structure called UQA. One posting list abstraction composes SQL, full-text, vector, graph, and geospatial paradigms through Boolean algebra. Goes further with deep_fusion() to express neural networks.
Bayesian BM25 — Probabilistic Scoring
Transforms raw BM25 scores into calibrated probabilities [0, 1] via Bayesian inference. Enables stable log-odds fusion of lexical, vector, and neural signals. BEIR zero-shot NDCG@10 41.67% (+6.28pt over BM25), calibration error reduced by up to 77%.
Industry Adoption
Bayesian BM25 has been adopted by leading search and retrieval frameworks.
Apache Lucene
The most widely used open-source search library. Bayesian BM25 integrated as a core scoring option.
MTEB Baseline
Massive Text Embedding Benchmark adopted Bayesian BM25 as the official baseline for retrieval evaluation.
Vespa.ai
Yahoo's large-scale serving engine. Bayesian BM25 included as an official rank profile example.
txtai
All-in-one embeddings database. Added Bayesian BM25 for probabilistic hybrid search pipelines.
Built for Production
Flexible Deployment
Run Cognica wherever your application needs it.
Server Mode
Deploy as a standalone database server with PostgreSQL wire protocol. Connect from any language or tool that supports PostgreSQL.
Embedded Mode
Embed directly into your application as a library, like SQLite3 or DuckDB. Zero network overhead, single-process simplicity.
Latest from the Blog
Between models and hardware lies an invisible structural gap — a data layer that should exist but doesn't. We examine why the current AI infrastructure stack is fragmented, why this is a structural problem rather than a transitional one, and what conditions a proper solution must satisfy. We then present UQA and the Cognica engine as one concrete response to these conditions.
by Jaepil Jeong | 2026-04-10
Graph databases solve relationship-heavy problems elegantly, but adding a separate graph system alongside your relational database creates operational complexity. We explain how Cognica integrates graph queries into its unified algebra, enabling Cypher and SQL to compose in a single transaction without data duplication.
by Jaepil Jeong | 2026-03-26
A cosine similarity of 0.85 tells you an angle, not a probability. We show how to transform vector similarity scores into calibrated relevance probabilities using distributional statistics that ANN indexes already compute — completing the probabilistic unification of text and vector retrieval.
by Jaepil Jeong | 2026-03-25