HTAP Roadmap¶
This roadmap is a reader-friendly snapshot of where ApexBase is today and what the project is optimizing for next. It is intentionally high level; implementation details live in issues, tests, and engineering documents.
Current Focus¶
ApexBase is focused on being a fast embedded HTAP engine with:
- Single-file table storage.
- Rust execution core and Python-first ergonomics.
- Analytical SQL over local data.
- Efficient DataFrame and Arrow interoperability.
- PostgreSQL Wire and Arrow Flight access for external tools.
- Full-text search and vector search as built-in capabilities.
Implemented Foundations¶
| Area | Status |
|---|---|
| Columnar storage | V4 row-group storage with .apex table files |
| Python API | ApexClient, ResultView, DataFrame import/export |
| SQL | DDL, DML, joins, CTEs, subqueries, windows, set operations |
| Multi-database | Named databases and database.table SQL references |
| Transactions | BEGIN / COMMIT / ROLLBACK, savepoints, optimistic concurrency |
| Indexing | B-Tree and Hash indexes |
| Full-text search | SQL-native FTS index management and MATCH() predicates |
| Vector search | TopK distance APIs and SQL integration |
| Protocols | PostgreSQL Wire and Arrow Flight servers |
| Rust embedding | Native Rust API for embedded use |
Near-Term Documentation Goals¶
- Keep the first-run path short and accurate.
- Split user-facing concepts from API reference material.
- Add focused guides for SQL, import paths, and protocol deployment.
- Keep contributor notes separate from application-user documentation.
- Build and publish docs automatically through GitHub Pages.
Product Direction¶
| Theme | Direction |
|---|---|
| Embedded reliability | Keep local persistence predictable and easy to reason about |
| Query performance | Preserve fast paths for common analytical and point lookup patterns |
| Interoperability | Make Arrow, Pandas, Polars, PostgreSQL clients, and Rust all feel native |
| Operational simplicity | Keep the default mode serverless while offering protocol servers when needed |
| Specialized search | Continue integrating structured SQL with text and vector search |
Contributor Entry Points¶
- Storage Architecture explains the storage stack.
- Engineering Guidelines records query-dispatch and fast-path rules.
- Rust Embedded API documents native integration points.
- Tests under
test/are the most concrete source of supported behavior.