Overview — Absolute DB

What is Absolute DB?

Absolute DB is a unified, AI-native database engine written entirely in pure C11. The current release ships as a single self-contained binary — no runtime dependencies, no package manager, no Docker required to get started. Drop it on any machine from a Raspberry Pi Zero to a 65,536-node planetary cluster and it works.

It is simultaneously a relational SQL database, a vector store, a graph engine, a full-text search engine, a time-series database, a key-value store, a document store, and a message queue — all sharing one storage layer, one transaction manager, one WAL, and one query engine. You do not stitch six products together. You deploy one.

# Download (macOS or Linux, x86-64 or ARM64)
curl -fsSL https://downloads.absolutedb.com/install.sh | bash

# Start
absdb-server --port 5433

# Connect with psql
psql -h localhost -p 5433 -U admin absdb

Why it exists

Modern applications are routinely forced to run four to eight specialised databases simultaneously: PostgreSQL for relational data, Pinecone or Weaviate for vectors, Neo4j for graph, Elasticsearch for full-text, TimescaleDB for time-series, Redis for caching and pub/sub, MongoDB for documents. Each product has its own operational model, backup strategy, security configuration, and licensing cost.

This fragmentation creates four compounding problems:

• Operational complexity: every database has its own upgrade cycle, monitoring agent, and failure mode
• Cross-model query tax: retrieving data from two databases always requires a round-trip through application code
• Consistency gaps: coordinating transactions across services requires distributed sagas or eventual consistency compromises
• Total cost: enterprise licences for five products easily exceed $50,000/month before any engineering time

Absolute DB eliminates every one of these problems with a single binary that is smaller than most JavaScript bundles.

Unified multi-model approach

Every data model in Absolute DB shares the same underlying storage layer (B+Tree + PAX columnar + LIRS buffer pool), the same MVCC transaction manager, the same WAL, and the same SQL engine. There is no impedance mismatch between models because they are not separate systems — they are different views over the same data.

A single SELECT statement can span relational JOIN + vector similarity + BM25 full-text + graph traversal — no application-side composition required:

-- Find documents about "consensus algorithms" that are:
--   1. semantically close to an embedding vector
--   2. authored by a user with >= 5 graph connections to "Raft"
SELECT d.id, d.title, d.body
FROM   documents d
JOIN   graph_edges g ON g.src = d.author_id AND g.edge_type = 'wrote'
WHERE  d.body MATCH 'consensus algorithms'
  AND  d.embedding <=> '[0.12, 0.98, ...]'::vector < 0.25
ORDER  BY d.embedding <=> '[0.12, 0.98, ...]'::vector
LIMIT  10;

Key differentiators

vs. PostgreSQL

PostgreSQL is exceptional but requires separate extensions for every capability: pgvector for vectors, TimescaleDB for time-series, pg_trgm for trigrams, Citus for sharding. Each extension adds operational risk and version compatibility constraints. Absolute DB ships all of these capabilities natively — fully integrated, tested together, and maintained as a single codebase.

vs. SQLite

SQLite (~600 KB) is the gold standard for embedded databases. Absolute DB's absdb-lite (~154 KB) is ~4× smaller and delivers vector search, BM25 full-text, graph traversal, SIMD-accelerated query execution, JIT compilation, and post-quantum cryptography in the same embedded zero-config form factor. For IoT and edge AI workloads, there is no comparable option.

vs. DuckDB

DuckDB (~12 MB embedded) is excellent for OLAP analytics on a single machine. Absolute DB's absdb-lite (~154 KB) is ~78× smaller, adds OLTP row-store, full ACID transactions, HNSW vector search, and a network server — while matching DuckDB's columnar scan throughput via PAX storage with AVX-512 SIMD kernels.

7 target classes

Absolute DB is designed to be the best database in each of seven distinct classes simultaneously — not a compromise that partially satisfies all of them.

Binary sizes

All binaries are stripped release builds with -Os + LTO. No external shared libraries are required beyond libc, libm, libpthread, and libdl (WASM JIT only).

Context: no other database ships TLS 1.3 + post-quantum crypto + HNSW + BM25 + Raft consensus + C-RAID in a single binary at this size.

make release
ls -lh bin/absdb bin/absdb-server bin/absdb-lite
# -rwxr-xr-x  absdb         ~8.0M
# -rwxr-xr-x  absdb-server  ~20.0M
# -rwxr-xr-x  absdb-lite    ~2.0M

Technical foundations

Pure C11

The entire codebase — 98 .c source files, ~120,000 lines — is written in pure C11 targeting POSIX-compliant systems. No C++, no Rust, no Go. The compiler flags include -Wall -Wextra -Wpedantic -Werror: zero warnings are permitted. Every commit must pass AddressSanitizer, UndefinedBehaviorSanitizer, and ThreadSanitizer with zero violations.

Zero external dependencies

Every capability — TLS 1.3, post-quantum cryptography, HTTP/2 + gRPC, Protobuf serialisation, Arrow IPC, Parquet, S3/GCS/Azure object storage — is implemented in-house in pure C11. There are no vendored third-party libraries. The only allowed linkage is libc, libm, libpthread, and libdl.

Performance highlights

License

Absolute DB is released under the Business Source License 1.1 (BSL-1.1). In summary:

• Free for development, testing, and non-production use — no restrictions
• Production use requires a commercial licence at the tier appropriate for your deployment (Community is free; see Pricing)
• On the change date (four years after each version's release), the version converts to the Apache License 2.0
• The full licence text is in the LICENSE file in the repository root

Owner: D.H.Maree (ABN 21 498 105 915) — sole IP holder. Licensed to SupportCALL AU as primary distributor. Website: absolutedb.com.