flagura

README

⚡ Flagura — Sub-Microsecond Feature Flagging & Release Engine

Feature flags that disappear from your critical path.
Sub-microsecond deterministic in-memory evaluation, OpenFeature native provider, gradual rollouts, and instant kill-switch circuit breakers.

Live Demo & Console • Architecture • OpenFeature • API Reference • SDK Integration • Deployment • Runbooks

---

📖 Overview

Flagura is a modern, high-performance feature flag and release management engine engineered for zero-latency in-process flag evaluation. Traditional SaaS feature flagging services introduce remote HTTP network hops (~20–80ms) that slow down user interactions and degrade Core Web Vitals.

Flagura solves this by evaluating rules and percentage rollouts locally in-memory using deterministic 64-bit FNV-1a hashing and atomic rule bitmaps, executing in sub-microsecond time (< 80 nanoseconds) with zero database roundtrips on evaluation hot paths.

---

⚡ Key Highlights

• ⚡ Sub-Microsecond Resolution: In-memory deterministic rule engine resolving flags in 80ns – 4µs without blocking the main event loop.
• 🌐 Native OpenFeature Provider: Drop-in vendor-neutral open-feature/go-sdk integration—switch or adopt Flagura with zero code lock-in.
• 🛡️ Resilient Offline Snapshot Mode: SDK caches rules to disk and boots seamlessly during server outages with zero request failures.
• 🎯 Deterministic Sticky Percentage Bucketing: Pure mathematical FNV-1a 64-bit hashing ensures users consistently land in the exact same rollout bucket across all sessions and platforms.
• 🚨 Instant Master Kill-Switch: 1-click edge circuit breaker to instantly shut down buggy features in production with zero code redeployments.
• 📊 Production Observability: Built-in Prometheus /metrics endpoint and Kubernetes /healthz, /livez, /readyz probes.
• 💎 Clean, Stripe-Grade UI/UX: Docked developer console, interactive switchboard playground, live evaluation sandbox, latency benchmark suite, and immutable audit trails.

---

⚖️ How Flagura Compares

Feature / Metric	⚡ Flagura	LaunchDarkly	Unleash	Flagsmith	GrowthBook
Core Architecture	Go (Native Binary)	Cloud SaaS / Daemon	Node.js / TypeScript	Python (Django)	TypeScript / Next.js
P50 Evaluation Latency	123 nanoseconds	~25 µs (SDK) / ~35ms	~25 µs	~50 µs	~30 µs
Memory Footprint	~35 MB	High (Relay Daemon)	~450 MB – 800 MB	~400 MB – 1 GB	~350 MB – 600 MB
Pricing Model	100% Free & Open Source	$500 – $5,000+/mo	Freemium ($80+/mo)	Freemium ($45+/mo)	Freemium ($20+/mo)
Zero DB I/O on Evaluation	✅ Yes (In-Memory)	⚠️ Requires Proxy	⚠️ Requires Redis/Proxy	❌ Queries DB/Cache	⚠️ Requires Cache
Data Privacy (Self-Hostable)	✅ 100% (Your DB)	❌ 3rd-Party SaaS	✅ Yes	✅ Yes	✅ Yes
Deployment Model	Single Binary / Docker / Vercel	Cloud SaaS Only	Multi-tier (Node + Redis + DB)	Multi-tier (Django + DB)	Multi-tier (Next + DB)

👉 Read full architectural comparison in documentation

---

🏗️ Architecture

Flagura is engineered for maximum execution speed, zero runtime overhead, and simple deployment:

┌────────────────────────────────────────────────────────────────────────┐
│                              FLAGURA ENGINE                            │
├────────────────────────────────────────────────────────────────────────┤
│  • Go 1.22+ Core          ── High-concurrency engine & FNV-1a hashing   │
│  • Direct SQL Store       ── Pure database/sql + lib/pq for Supabase   │
│  • API-First Design       ── Ultra-fast batch & single evaluation API  │
│  • Compiled Views         ── Type-safe compiled HTML views & CSS      │
│  • Reactive UI Components ── Fast, zero-page-reload console components │
└────────────────────────────────────────────────────────────────────────┘

                     ┌──────────────────────────┐
                     │ Client Application / SDK │
                     └─────────────┬────────────┘
                                   │ (In-Memory / HTTP)
                                   ▼
        ┌─────────────────────────────────────────────────────┐
        │               Flagura Evaluation Engine             │
        ├─────────────────────────────────────────────────────┤
        │  1. Check Environment Master Kill-Switch            │
        │  2. Evaluate Specific Attribute Rules (Email/Tier)  │
        │  3. Compute FNV-1a 64-bit Sticky Hash Bucket (0-99) │
        │  4. Resolve Treatment vs Control Group              │
        └──────────────────────────┬──────────────────────────┘
                                   │ (< 80ns execution)
                                   ▼
                     ┌──────────────────────────┐
                     │  Boolean / Variant Value │
                     └──────────────────────────┘

---

🚀 Quickstart

1. Run Locally (In-Memory Edge Mode)

No database installation required to get started. Flagura includes a built-in in-memory edge store pre-seeded with sample flags:

# Clone the repository
git clone https://github.com/dhawalhost/flagura.git
cd flagura

# Run the Go server
go run main.go

Open your browser to:

• Storytelling Landing Page: http://localhost:3000
• Developer Console: http://localhost:3000/dashboard

---

2. Connect to Supabase PostgreSQL

1. Open your Supabase Dashboard -> SQL Editor.

2. Run the migration script in supabase/schema.sql:

   CREATE TABLE IF NOT EXISTS feature_flags (
       id TEXT PRIMARY KEY,
       key TEXT UNIQUE NOT NULL,
       name TEXT NOT NULL,
       description TEXT,
       type TEXT NOT NULL DEFAULT 'boolean',
       tags TEXT[] DEFAULT '{}',
       variants JSONB DEFAULT '[]'::jsonb,
       environments JSONB NOT NULL DEFAULT '{}'::jsonb,
       created_at TIMESTAMPTZ NOT NULL DEFAULT NOW(),
       updated_at TIMESTAMPTZ NOT NULL DEFAULT NOW()
   );
   
   CREATE TABLE IF NOT EXISTS audit_logs (
       id TEXT PRIMARY KEY,
       flag_key TEXT NOT NULL,
       action TEXT NOT NULL,
       environment TEXT NOT NULL,
       actor TEXT NOT NULL,
       details TEXT,
       timestamp TIMESTAMPTZ NOT NULL DEFAULT NOW()
   );
   

3. Set your connection string and launch:

   export DATABASE_URL="postgres://postgres.[PROJECT_REF]:[PASSWORD]@aws-0-[REGION].pooler.supabase.com:6543/postgres?sslmode=require"
   go run main.go
   

---

🔌 REST API Reference

1. Batch / Single Flag Evaluation

Endpoint: POST /api/v1/evaluate

Evaluates requested feature flags against a user context in sub-millisecond time.

Request:

curl -X POST http://localhost:3000/api/v1/evaluate \
  -H "Content-Type: application/json" \
  -d '{
    "flags": ["ai-smart-search", "new-checkout-flow"],
    "context": {
      "user_id": "usr_dhawal_01",
      "email": "dhawal@flagship.dev",
      "country": "US",
      "role": "admin",
      "tier": "enterprise",
      "environment": "production"
    }
  }'

Response:

{
  "results": {
    "ai-smart-search": {
      "flag_key": "ai-smart-search",
      "enabled": true,
      "variant": "treatment",
      "value": true,
      "reason": "TARGETING_RULE_MATCH",
      "matched_rule_id": "rule_staff_whitelist",
      "bucket_val": 42,
      "evaluation_latency_ns": 4820,
      "evaluation_latency_us": 4.82
    },
    "new-checkout-flow": {
      "flag_key": "new-checkout-flow",
      "enabled": true,
      "variant": "true",
      "value": true,
      "reason": "PERCENTAGE_ROLLOUT",
      "bucket_val": 35,
      "evaluation_latency_ns": 120,
      "evaluation_latency_us": 0.12
    }
  },
  "total_flags": 2,
  "environment": "production",
  "total_duration_us": 22.4
}

---

2. Flag Management Endpoints

Method	Endpoint	Description
GET	/healthz / /livez	Kubernetes liveness health probe
GET	/readyz	Kubernetes readiness probe (checks storage)
GET	/metrics	Standard Prometheus metrics exposition
GET	/api/v1/flags	List all feature flags across environments
GET	/api/v1/flags/:key	Retrieve details for a specific flag
POST	/api/v1/flags	Create or update a feature flag configuration
PATCH	/api/v1/flags/:key/toggle	Instant 1-click toggle for master kill-switch
PATCH	/api/v1/flags/:key/rollout	Dynamically update percentage rollout (0–100%)
DELETE	/api/v1/flags/:key	Permanently remove a feature flag
POST	/api/v1/benchmark	Execute live in-process latency stress test
GET	/api/v1/audit-logs	Fetch immutable audit trail history

---

🌐 OpenFeature Go Provider

Flagura includes a first-class, official OpenFeature Provider implementing the CNCF vendor-neutral specification:

go get github.com/dhawalhost/flagura/pkg/openfeature
go get github.com/open-feature/go-sdk

package main

import (
	"context"
	"log"
	"time"

	"github.com/dhawalhost/flagura/pkg/client"
	flaguraOF "github.com/dhawalhost/flagura/pkg/openfeature"
	of "github.com/open-feature/go-sdk/openfeature"
)

func main() {
	// 1. Initialize Flagura in-memory client
	flaguraClient := client.New("https://flagura.dhawalhost.com",
		client.WithLocalEvaluation(30*time.Second),
		client.WithSnapshotFile("/tmp/flagura-cache.json"), // offline resilience
	)
	defer flaguraClient.Close()

	// 2. Register Flagura as your global OpenFeature Provider
	_ = of.SetProviderAndWait(flaguraOF.NewProvider(flaguraClient))

	// 3. Evaluate flags using standard OpenFeature APIs
	ofClient := of.NewClient("checkout-service")
	evalCtx := of.NewEvaluationContext("usr_alex_42", map[string]interface{}{
		"email": "alex@company.com",
		"tier":  "enterprise",
	})

	enabled, _ := ofClient.BooleanValue(context.Background(), "ai-smart-search", false, evalCtx)
	log.Printf("Flag Status: %v", enabled)
}

---

💻 Polyglot SDK Quickstart

1. Official Go Native SDK (pkg/client)

The built-in Go client supports sub-microsecond in-memory evaluations (cached in-process) and remote evaluations:

go get github.com/dhawalhost/flagura/pkg/client

package main

import (
	"context"
	"fmt"
	"time"

	"github.com/dhawalhost/flagura/pkg/client"
)

func main() {
	// Initialize Flagura client with in-memory caching and offline snapshot
	c := client.New("https://flagura.dhawalhost.com",
		client.WithLocalEvaluation(30*time.Second),
		client.WithSnapshotFile("/tmp/flagura-cache.json"),
	)
	defer c.Close()

	// Evaluate flag locally in < 80 nanoseconds
	result, err := c.Evaluate(context.Background(), "ai-smart-search", client.Context{
		UserID:  "usr_dhawal_01",
		Email:   "dhawal@flagura.dev",
		Country: "US",
		Role:    "admin",
	})
	if err != nil {
		panic(err)
	}

	fmt.Printf("Flag: %s | Enabled: %v | Variant: %s (in %d ns)\n",
		result.FlagKey, result.Enabled, result.Variant, result.EvaluationLatencyNs)
}

---

2. TypeScript / JavaScript (Node.js, Next.js, Browser)

Zero-dependency standard fetch helper for full-stack JavaScript:

interface FlagEvaluation {
  flag_key: string;
  enabled: boolean;
  variant: string;
  value: any;
  reason: string;
}

export async function evaluateFlag(
  flagKey: string,
  context: { userId: string; email?: string; country?: string; role?: string; environment?: string },
  endpoint = "https://flagura.dhawalhost.com"
): Promise<FlagEvaluation> {
  const res = await fetch(`${endpoint}/api/v1/evaluate`, {
    method: "POST",
    headers: { "Content-Type": "application/json" },
    body: JSON.stringify({
      flags: [flagKey],
      context: {
        user_id: context.userId,
        email: context.email,
        country: context.country,
        role: context.role,
        environment: context.environment || "production",
      },
    }),
  });

  if (!res.ok) throw new Error(`Flagura API error: ${res.statusText}`);
  const data = await res.json();
  return data.results[flagKey];
}

// Example Usage
const flag = await evaluateFlag("ai-smart-search", {
  userId: "usr_123",
  email: "user@flagura.dev",
});
if (flag.enabled) {
  console.log("Feature active! Variant:", flag.variant);
}

---

3. Python (FastAPI, Django, AI Agents)

import requests

def evaluate_flag(flag_key: str, user_id: str, email: str = "", endpoint: str = "https://flagura.dhawalhost.com") -> bool:
    try:
        response = requests.post(f"{endpoint}/api/v1/evaluate", json={
            "flags": [flag_key],
            "context": {
                "user_id": user_id,
                "email": email,
                "environment": "production"
            }
        }, timeout=2.0)
        response.raise_for_status()
        data = response.json()
        return data.get("results", {}).get(flag_key, {}).get("enabled", False)
    except Exception as e:
        print(f"[WARN] Flag evaluation fallback to false: {e}")
        return False

# Example Usage
if evaluate_flag("ai-smart-search", user_id="usr_dhawal_01", email="dhawal@flagura.dev"):
    print("AI Smart Search is enabled!")

---

4. cURL / REST API

curl -X POST https://flagura.dhawalhost.com/api/v1/evaluate \
  -H "Content-Type: application/json" \
  -d '{
    "flags": ["ai-smart-search"],
    "context": {
      "user_id": "usr_dhawal_01",
      "email": "dhawal@flagura.dev",
      "environment": "production"
    }
  }'

---

🚢 Deployment Options

Option A: Vercel Serverless (Zero Config)

1. Push your code to GitHub.
2. Import the project into Vercel.
3. Under Settings -> Environment Variables, add:
   • DATABASE_URL: postgres://... (Your Supabase connection string)
4. Click Deploy. Vercel compiles the serverless handler via api/index.go and vercel.json.

Option B: Docker Container

# Build Docker image
docker build -t flagura:latest .

# Run container
docker run -p 3000:3000 -e DATABASE_URL="postgres://..." flagura:latest

---

🧪 Testing & Verification

Run the comprehensive unit and integration test suite:

# Run all Go tests
go test -v ./...

# Recompile Templ components
templ generate

# Build production binary
go build -o bin/flagura main.go

---

📚 Operations & Runbooks

Comprehensive SRE, emergency response, and operational runbooks are available in the docs/runbooks/ directory:

• 🚀 Deployment & Release Management: Standalone binary, Docker, Vercel serverless, and rollback procedures.
• 🌐 Supabase & Vercel Integration Guide: Step-by-step cloud setup, connection pooling, and continuous deployment.
• 🚨 Incident Response & Emergency Operations: 1-click kill switch, database outage fallback, latency debugging, and audit forensics.
• 🗄️ Database Operations & Disaster Recovery: Supabase PostgreSQL schema, connection pool tuning, and PITR backup/recovery.
• 🛡️ Security & Access Management: RBAC provisioning, credential rotation, and automated vulnerability triage.

---

📄 License

This project is licensed under the MIT License.