Ridge DB

Ridge is seed's bounded, single-node relational database. It can run as an embedded library or as the loopback-only ridged service through a documented PostgreSQL-wire subset. Ridge Admin provides the browser-based administration workbench.

Ridge v0.3 gates V3-G0 through V3-G25 are complete for their recorded scopes. This is a local production foundation, not a PostgreSQL replacement and not a distributed database.

Choose a starting point

Goal	Guide
Start a local database	Quickstart
Understand storage, WAL, and MVCC	Architecture
Follow the complete Ridge and Ridge Admin runtime	Runtime architecture
Write supported SQL	SQL reference
Embed Ridge in a seed application	Embedded API
Connect with psql or an application driver	Pgwire and drivers
Back up, restore, inspect, or maintain data	Operations
Use the browser administration application	Ridge Admin
Check compatibility and unsupported features	Compatibility
Diagnose a failure	Troubleshooting
Equip an AI agent to work with Ridge	AI agents and the ridgedb skill

Capability summary

Area	Current bounded capability
Storage	Persistent recursive B+Trees, generation-2 typed rows and schemas
Durability	CRC-32 logical WAL, fsynced commit, committed-only recovery
Transactions	Atomic multi-table transactions, read committed and repeatable read
Concurrency	Disjoint staging and publication with deterministic conflict detection
SQL	Multi-column relational SQL, indexes, constraints, joins, aggregates, windows
Data workloads	JSON, time-series, full-text search, exact and bounded-HNSW vectors
Operations	Integrity, vacuum, backup/restore, PITR, metrics, quotas, read-only mode
Connectivity	Embedded API and bounded TLS/password pgwire on loopback
Administration	Ridge Admin local web UI with local and verified SSH connections

What Ridge intentionally does not claim

Ridge does not currently provide replication, failover, sharding, distributed SQL, stored procedures, triggers, extensions, COPY, SCRAM, general binary formats, or complete PostgreSQL catalogs. PostgreSQL drivers work only through the tested pgwire subset.

Unsupported syntax and protocol shapes fail explicitly. Do not infer a feature from PostgreSQL, MySQL, or SQLite merely because Ridge accepts familiar SQL.

Sources of truth

Public documentation summarizes the operational contracts. The authoritative implementation sources are:

grove/apps/ridge/CAPABILITIES.md for the closed capability inventory;
grove/apps/ridge/V3_CONTRACT.md for SQL, type, isolation, and error rules;
grove/libs/ridge_sql/README.md for the accepted SQL surface;
grove/apps/ridge/ROADMAP.md for gates and deferred work;
grove/apps/ridge-admin/README.md and RELEASE.md for Ridge Admin;
executable tests and benchmark artifacts under grove/apps/ridge/tests/, grove/libs/ridge_sql/tests/, and grove/apps/ridge/benchmarks/.

Capability claims are versioned and bounded. Benchmark results are regression evidence for a named host and workload, not universal database rankings.