Under the Hood

Tech Stack

Understanding the technology behind Shiro's performance and capabilities.

<100ms
Startup time
~10MB
Binary size
0
Dependencies
4+
AI providers

Why GoLang?

Go's compiled nature enables Shiro to process complex DAG workflows in under 100ms.

Blazing Fast

Compiled binary executes workflows in milliseconds. Zero interpreter overhead, no JVM, no startup delay.

📦

Single Binary

Compile to a static binary with zero dependencies. Drop it anywhere — CI runner, container, or bare metal.

🔀

Built-in Concurrency

Goroutines and channels enable efficient parallel DAG execution without complex threading primitives.

🌍

Cross-Platform

One codebase compiles for Linux, macOS, Windows — AMD64 and ARM64. No environment-specific builds.

Module System Architecture

Four module types give you zero-dependency built-ins and unlimited extensibility.

📦

Built-in Modules

Zero deps

Core functionality compiled into the binary: print, slack.notify, git.diff, ai.generate — always available.

🔌

Subprocess Modules

Extensible

External programs communicating via JSON over stdin/stdout. Binary mode or go-run mode.

🌐

HTTP Modules

Distributed

RESTful API endpoints with load balancing, health checks, and circuit breakers.

🐙

GitHub Integration

Marketplace

Auto-discover and install modules from GitHub repositories tagged shiro-automation-module.

AI Provider Support

First-class support for multiple AI providers — local or cloud.

🧠

Ollama

Local / Private

Run Llama, Mistral, and more on your own hardware. Full privacy, zero cloud dependency.

OpenAI

GPT-4 / GPT-4o

Production-grade AI with enterprise reliability. GPT-4 and latest OpenAI models supported.

Gemini

Google AI

Google's Gemini models via Google AI Studio or Vertex AI integration.

�🔧

Custom Endpoint

Bring your own

Any OpenAI-compatible API: vLLM, LM Studio, custom deployments — same config syntax.

Secure config: Use {env.API_KEY} syntax — API keys never hardcoded in workflow files.

CI Platform Integration

Runs inside your existing runners — no new infrastructure required.

🦊

GitLab CI

Recommended

Native artifact state storage, human-in-loop approvals via manual jobs, and pipeline resumption.

🐙

GitHub Actions

Supported

Full integration for PR reviews, push notifications, and workflow automation.

🛠️

Jenkins

Planned

Compatible as a build step or shell command in any Jenkinsfile pipeline.

☁️

Kubernetes Jobs

Cloud-native

Run as K8s Jobs for cloud-native CI/CD with container isolation and horizontal scalability.

State Storage Backends

Pluggable state backends — from CI-native artifacts to in-memory ephemeral storage.

🗂️

GitLab Artifacts

CI default

Automatically uploads workflow state to GitLab CI artifacts. Downloaded between stages — no external infra.

💾

Filesystem

Local dev

Stores state in the local filesystem. Simple and reliable for local development and testing.

Memory

Fastest

In-memory storage for ephemeral workflows. Maximum speed, zero persistence.