Modules Overview

Graviton is split into focused SBT sub-projects. Each module owns a specific slice of the system so that storage backends, protocol surfaces, and UI experiences can evolve independently. Use this page as a map before diving into the detailed module guides.

High-level Layout

• Core (documented separately): pure data types, codecs, algebraic structures, and the Transducer algebra for composable pipeline stages.
• Runtime: storage ports, policies, and operational concerns that wire pure logic to effectful infrastructure.
• Streams: ZIO Stream utilities that implement chunking, hashing, and signal processing primitives — designed to compose with Transducers.
• PDF: typed, bounded-memory PDF ingest backed by zio-pdf's incremental structural scanner.
• Shardcake integration: opt-in session ownership and direct streamed upload locality for multi-node deployments.
• Distributed admission: optional Redis or Valkey leases for cluster-wide service, tenant, and backend transfer fairness.
• Protocol: shared API models and transport-specific servers (gRPC + HTTP).
• Backend adapters: concrete integrations (PostgreSQL, S3, RocksDB) that satisfy runtime ports.
• Frontend: Scala.js remote console embedded in the documentation site. The packaged server also has a separate server-rendered Datastar console.

Quick Links

• Backend adapters
• Runtime module
• Streams utilities
• PDF-aware ingest
• Shardcake upload locality
• Distributed transfer admission
• Protocol stack
• Scala.js documentation frontend

Each page outlines the responsibilities, current implementation status, and remaining engineering work. Apache Tika is deliberately absent from this module list because it is not shipped. See the implementation status ledger before choosing a released artifact or a source-only main feature.