How It Works

From Markdown to Machine-Executable Workflows

From Markdown to Machine-Executable Workflows

The Spry Execution Flow

1

Write Markdown

Create Markdown files with fenced code blocks. Each block can be SQL, Bash, JSON, prompts, or any executable code.

2

Spry Runtime Executes

The Spry CLI reads your Markdown, parses fenced blocks, validates them with TypeScript and Zod, then executes each cell in order.

3

Structured Outputs

Results are captured as structured artifacts: SQL tables, JSON files, embeddings, or vector stores. Everything is versioned and traceable.

4

Deploy & Compose

Use outputs in SQLPage applications, LLM context graphs, or as evidence for compliance. Every step is reproducible and auditable.

Architecture

Spry unifies documentation, data pipelines, and AI orchestration into one transparent system.
TypeScript-native, dependency-light, and plugin-friendly. Perfect for reproducible intelligence engineering.

Markdown Files
Spry CLI
Deno Runtime
Data Outputs (SQLite, JSON)
AI Context (Vector Stores, Prompts)

Key Principles

Simple

Just Markdown. No complex DSLs or configuration files. Write what you know.

Deterministic

Type-safe execution with Zod validation. Every run produces the same results.

Reproducible

Version everything. Trace lineage. Audit with confidence. Evidence by default.