Documentation

AionDB is a PostgreSQL-wire database engine that keeps relational tables, graph labels, and vector search in one local system. These pages focus on using and evaluating the product. Implementation notes live separately.

New in v0.3: AionDB now ships the vector update: pgvector-style SQL, HNSW, IVF-flat, Qdrant-style filtered helpers, PostgreSQL ecosystem compatibility work, and a reproducible vector benchmark harness. Start with What's New in v0.3.

Crate-by-crate implementation notes live in Advanced Specification. They are useful for contributors, but they are not the recommended starting point for users evaluating the database.

Choose a path

v0.3 At A Glance

v0.3 ships the vector stack:

• HNSW raw reaches 0.996 recall@10 in the default vector benchmark.
• HNSW PQ reaches 0.994 recall@10.
• IVF-flat builds the default 50k-vector dataset in about 416-418 ms.
• IVF-flat with nprobe=32 reaches 0.863 recall@10 at about 2.57 ms mean query latency.
• Qdrant-style JSON filters bring metadata-aware retrieval into the vector helper layer.

Start

• Installation: build from source, create a local archive, run the container profile, and review the Kubernetes and systemd templates.
• Getting Started: build the server, create a local user, connect with psql, and run the first query.
• Tutorial: one small dataset using SQL, graph labels, and vector scoring.
• Example Workloads: practical schemas for product, support, and knowledge-base evaluations.

Learn

• Core Concepts: the mental model behind tables, labels, vectors, and catalog state.
• Architecture: how the server, engine, catalog, storage, and WAL fit together.
• Storage Format and WAL Contract: the current storage and write-ahead-log contracts.
• Query Lifecycle: what happens between a client query and execution.
• Tradeoffs: workloads where AionDB is a good fit, and workloads where it is not.

Build

• SQL, SQL Statements, Data Types, and Functions: the relational surface.
• Graph and Vector, Graph Reference, and Vector Reference: hybrid query features over ordinary tables.
• Indexes and Constraints, Transactions, and System Catalogs: behavior that affects application correctness.
• Interfaces, Client Drivers, PostgreSQL Compatibility, and Ecosystem Integrations: connecting through pgwire, integrating normal SQL tools, or embedding the engine.

Manage

• Configuration: command-line flags, environment variables, and local data directories.
• Administration, Control Plane, Operations, and Security: operating the server during evaluation.
• Observability, Explain JSON, Storage Compatibility, Backup and Recovery, and Troubleshooting: diagnosing and recovering local deployments.

Evaluate

• v0.3 Vector Performance, Benchmarks, Benchmark Results, Benchmark Reproducibility, and Performance Tuning: running vector, graph, SQL, and hybrid performance checks.
• Testing, Evaluation Checklist, Product Hardening Plan, and Migration Guide: deciding whether a workload is ready to try.
• Limitations, Error Reference, FAQ, and Glossary: boundaries and terminology.

Project

• What's New in v0.3, Roadmap, Roadmap to v1, Governance, GTM Evidence, Release Notes, and Release Process: where the product is going, how decisions are made, what evidence supports claims, and how releases are described.
• Contributing: how to work on the codebase without starting from crate-level internals.

Reading order for reviewers

For a fast technical review, read:

1. Core Concepts
2. Tradeoffs
3. Limitations
4. PostgreSQL Compatibility
5. Benchmarks

That path answers the main credibility questions before going deep into reference pages.