Skip to content

disciplinedware/swiftward

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

1 Commit
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Swiftward

Ship policy changes without fear. Test in production, rollback instantly.

The policy enforcement infrastructure you'd build yourself — deterministic evaluation, A/B testing, HITL workflows — without the 18-month timeline. Self-hosted only.

Try in 5 minutes · Talk to us


The Problem

If you're enforcing policies (content moderation, AI guardrails, fraud detection), you're doing one of:

  1. Hardcoding logic — policy changes require deploys, no safe testing, debugging is guesswork
  2. Using SaaS — data leaves your infrastructure, vendor lock-in, limited customization
  3. Building from scratch — 12-18 months minimum, still missing versioning/testing/audit infrastructure

The core challenge: How do you validate policy changes without breaking production?

Our Approach

1. Deterministic Evaluation

Same event + same policy + same state = same verdict. Always. Essential for compliance audits, debugging, and safe testing.

2. Shadow Mode + A/B Testing

Test new policies in production without affecting users. Split traffic between versions. Compare metrics. Roll back instantly.

3. Stateful Decisions

Track entity state across events — trust scores, violation counts, rate limits. ACID transactions, replay-safe.

4. Complete Audit Trails

Every decision traced with signals computed, rules matched, and state changes applied. Explain any historical decision.

How It Compares

Capability Swiftward OPA/Cedar Classification SaaS Build-It-Yourself
Deterministic Replay Guaranteed Yes No Hard to achieve
A/B Testing Built-in None Limited 6+ months
Shadow Mode Built-in None Some Build yourself
Entity State Built-in External Limited Build yourself
HITL Workflows Built-in None Varies Build yourself
Deployment Self-hosted Self-hosted SaaS Your infra
Time to Value Days Weeks Days 12-18 months

Detailed comparisons →

Try It

Live demos (no build required):

Demo Description Time
Minimal Demo Simplest setup - perfect for first-time users 2 min
Policy Enforcement Demo A/B testing, shadow mode, HITL workflows, LLM classification 5 min (+ ~20 min first run to download ML models)

Both demos use pre-built images - just run docker compose up.

Current Status

  • Stage: Seeking first design partners
  • Looking for: 1-2 design partners to validate approach
  • Maturity: Core works, production hardening in progress

What works: Event processing, policy evaluation, state management, A/B testing, HITL queues, audit trails.

What we're validating: Real-world policy complexity, DSL ergonomics, integration patterns.

Get Started

We're looking for 1-2 design partners to validate Swiftward together. Direct founder access, real use cases, honest feedback.

Email: hello@swiftward.dev — 30-minute scoping call, we'll figure out if there's a fit.

Repository

This is the public examples and documentation repository.

  • Core source code → Available to customers and partners
  • This repo → Live demo, examples, and documentation to evaluate Swiftward
  • License → Source-available, see LICENSE

Documentation

FAQ

Ready for production? Core works, good for pilots. Production hardening in progress.

Different from OPA? OPA is stateless authorization. Swiftward is stateful policy enforcement with A/B testing and HITL.

Performance? Thousands of events/sec single node, scales horizontally.

Full FAQ →

Contact

Email: hello@swiftward.dev
Founder: Konstantin Trunin

Contributors