StrataDB
Docs
Star on GitHub2
Docs Get Started Star on GitHub · 2

WHITEPAPERS

The internals, written down.

How StrataDB works, at the level of invariants and proofs — not promises.

The layered stack

Five crates from shared vocabulary to model execution, and the dependency rules the build enforces to keep meaning and mechanics apart.

The storage substrate

How StrataDB's storage layer persists generic versioned rows while knowing nothing about product meaning, and how the engine attaches through a stable runtime boundary.

Durability and recovery

Two storage modes, two durability policies, a write-ahead log that halts on fsync failure, and deterministic recovery on open.

Commits and versioning

Writes auto-commit into versioned commit units, storage keeps a per-branch version-to-timestamp timeline, and there are no manual transactions.

Data capabilities

How the engine turns one branch-aware MVCC row into six capabilities — KV, JSON, event, vector, graph, and branches — plus derived state.

Runtime modes

Durable, cache, read-only, and IPC-backed access, plus the resource-profile model that runs one binary from edge devices to servers.

Errors and diagnostics

How detailed storage mechanics become a small, stable set of engine-owned public errors, and what observability the engine exposes.

Product

  • Strata Foundry
  • Changelog

Documentation

  • Getting Started
  • Guides
  • Cookbook
  • Reference

Internals

  • Architecture
  • strata-core
  • strata-foundry
  • stratahub

For agents

  • llms.txt
  • For AI agents
  • MCP reference

© 2026 StrataDB · Apache-2.0 · Research preview