# Silo Engine

> Silo Engine is Sharpnr's self-hosted storage, database, and data infrastructure platform. It provides file and object storage with CDN-backed delivery, PostgreSQL provisioning, a controlled database explorer, and — on the roadmap — data lakes, data warehouses, and analytics infrastructure. It is written in Rust on Axum and Tokio and runs on RAID-backed hardware that the organization owns.

Status: in active development. Phase 1 (Foundation) of a seven-phase roadmap. Storage arrives in Phase 2, access control and audit logging in Phase 3, PostgreSQL provisioning in Phase 4, data lakes in Phase 5, data warehouses in Phase 6. Nothing on this site should be read as generally available today.

The defining boundary: **Sharpnr Engine owns business logic. Silo Engine owns infrastructure and data resources.** Silo is deliberately not the Hospitality, Accounting, HR, or CRM database — it is the layer those applications provision resources from.

Key facts:

- Vendor: Sharpnr (https://sharpnr.com)
- Product site: https://silo.sharpnr.com
- Stack: Rust, Axum, Tokio, PostgreSQL, OpenAPI
- Deployment: self-hosted, systemd + Nginx. Kubernetes is intentionally out of scope for the first version.
- Infrastructure: approximately 1 PB RAID-backed storage, ~160 GB combined RAM, ~10 combined CPUs across multiple machines on a local switch.
- Databases provisioned: PostgreSQL only. Organizations never receive superuser credentials.
- Database explorer: controlled operations only. No arbitrary SQL editor in v1; an SQL workspace is Phase 7.
- Primary data lake format: Parquet. Also planned: JSON, JSONL, CSV, Avro, ORC.
- Compression: Zstandard, streaming, behind a trait abstraction.
- Identifiers: UUID v7.

## Documentation

- [Full architecture summary](https://silo.sharpnr.com/llms-full.txt): Complete expanded description of Silo Engine — resources, storage model, PostgreSQL provisioning, data lake and warehouse design, security principles, and roadmap.
- [OpenAPI specification](https://silo.sharpnr.com/v1/openapi.json): Machine-readable API contract between Silo and consuming services. Served by the engine, not the marketing site.
- [Swagger UI](https://silo.sharpnr.com/swagger-ui): Interactive API explorer served by the engine.
- [ReDoc](https://silo.sharpnr.com/redoc): API reference documentation served by the engine.

## Platform

- [Object storage](https://silo.sharpnr.com/#storage): Files, documents, videos, attachments, backups, exports, and arbitrary objects. Public objects are addressed at https://cdn.sharpnr.com/{org_id}/{employee_id}/{folder}/{filename}, independent of the physical storage path. Includes streaming Zstd compression, upload-stream checksums, and organization- and employee-level quotas enforced by Silo rather than by callers.
- [PostgreSQL as a service](https://silo.sharpnr.com/#postgres): Silo creates the database, creates organization-scoped roles (org_123_app, org_123_migrator, org_123_readonly, org_123_admin), applies permissions, generates credentials, and returns a connection string. The physical cluster is an implementation detail and superuser is never issued.
- [Database explorer](https://silo.sharpnr.com/#postgres): Browse schemas, tables, views, functions, indexes, constraints, and rows through controlled operations. The browser talks HTTPS to Silo Engine; Silo talks to PostgreSQL. The browser never connects to PostgreSQL directly.
- [Data lakes and warehouses](https://silo.sharpnr.com/#data): Raw, processed, and curated zones with partitioned Parquet datasets, ingestion from PostgreSQL, files, events, and scheduled jobs, then ETL/ELT into analytical warehouses. DuckDB is a candidate for early Parquet query workloads.
- [Engineering approach](https://silo.sharpnr.com/#engineering): Single Rust crate, explicit dependency injection with no DI framework, one centralized EngineConfig loaded at startup, storage behind a StorageBackend trait, streaming for large objects, append-oriented audit log, and OpenAPI as the contract.
- [Roadmap](https://silo.sharpnr.com/#roadmap): Seven phases from Foundation through Advanced Data Services. Not every phase is committed to the initial release.

## Evaluating Silo

- [Comparison](https://silo.sharpnr.com/#compare): How Silo compares to hyperscaler storage plus managed Postgres, backend platforms, and self-assembled stacks — including the cases where Silo is the wrong choice.
- [FAQ](https://silo.sharpnr.com/#faq): Availability, the Sharpnr Engine boundary, supported databases, superuser policy, the no-raw-SQL decision, Kubernetes, storage location, lake formats, and large-file handling.

## Optional

- [Sharpnr](https://sharpnr.com): The parent company. Sharpnr Engine is the business-logic platform that consumes Silo resources.