Valdrics

Optimize Cloud Value, Not Just Cost

Value + Matrix — A FinOps engine that continuously optimizes cloud value
by eliminating waste, controlling cost, and reducing unnecessary overhead.

The Problem · The Solution · Features · Get Started · Roadmap

---

💸 The Problem

"We're spending $47,000/month on AWS... and I can't explain where 30% of it goes."
— Every engineering manager at some point

Here's what the industry knows but rarely admits:

• $164 billion in cloud spend was wasted in 2024. (Flexera State of the Cloud Report)
• 30-35% of cloud resources are idle, orphaned, or oversized. (Gartner)
• The average company has no idea what a developer spin-up costs until the monthly invoice arrives.

And it's not just money. Every idle EC2 instance, every orphan EBS volume, every forgotten load balancer is burning electricity and emitting CO₂ for absolutely nothing.

Most FinOps tools give you dashboards.
Dashboards give you graphs.
Graphs give you... more questions.

You don't need more graphs. You need actionable value insights.

---

🛡️ The Solution

Valdrics transforms raw cost data into actionable value intelligence.

It connects to your cloud, uncovers waste, explains spend behavior, and gives you exactly what to do—with receipts.

Here's how it works:

1. Connect → One-click AWS IAM role setup. Read-only. Zero secrets stored.
2. Scan → Our 11 zombie-detection plugins sweep your account every day.
3. Reason → The LLM brain (GPT-4o, Claude 3.5, Groq, Gemini) analyzes context, not just metrics.
4. Act → Get Slack alerts, approve remediations, and watch your bill shrink.

Tip

Zero API Costs for Your AWS Account: Valdrics uses AWS CUR and Resource Explorer 2 instead of expensive APIs like Cost Explorer ($0.01/request). Your AWS bill from Valdrics scans is ~$0.00/month.

              ┌──────────────────┐
              │   Your Cloud ☁️  │
              └────────┬─────────┘
                       ▼
              ┌──────────────────┐
              │  🔌 Valdrics Core │
              │    (FastAPI)     │
              └────────┬─────────┘
                       ▼
   ┌───────────────────┴───────────────────┐
   │           Zombie Detection            │
   │  EC2 · EBS · S3 · RDS · NAT · ELB    │
   │  SageMaker · Redshift · ECR · EIPs   │
   └───────────────────┬───────────────────┘
                       ▼
              ┌──────────────────┐
              │   🧠 LLM Brain   │
              │  (Multi-Model)   │
              └────────┬─────────┘
                       ▼
   ┌───────────────────┴───────────────────┐
   │          Slack Alerts + Dashboard     │
   │         (Approve / Reject / Act)      │
   └───────────────────────────────────────┘

---

✨ Features

🧟 Deep Zombie Detection

Not just "idle EC2." We find everything:

Category	What We Hunt	Precision Signals
Compute	Idle EC2, Azure VMs, GCP Instances	GPU Hunting (P/G/Nvidia), Owner Attribution
Storage	Orphan EBS, Managed Disks, Snapshots	Creator Attribution, Age-based decay
Network	Unallocated IPs, Orphan LBs, NAT GWs	Association Tracking
Data	Idle RDS, Redshift, GCP SQL	Connection Activity
Registry	Stale ECR, ACR, GCR Images	Pull Frequency

11 detection plugins + Multi-Cloud Parity (AWS, Azure, GCP).

---

🧠 AI That Actually Thinks

Other tools use static rules: "CPU < 10% for 7 days = zombie."

Valdrics asks: "Why did RDS costs spike 47% on Tuesday?"
And answers: "Because Staging-DB-04 was left running after the load test. Estimated waste: $312/month."

Powered by your choice of:

• OpenAI (GPT-4o, GPT-4o-mini)
• Anthropic (Claude 3.5 Sonnet)
• Google (Gemini 2.0 Flash)
• Groq (Llama 3.3 70B — fast and cheap)

Bring Your Own Key (BYOK) supported. Keep your API costs in your own account.

---

🌿 GreenOps Native

Every wasted dollar has a carbon cost. Valdrics calculates it.

Total CO₂ this month:      42.7 kg
Equivalent to:             105 miles driven  🚗
Trees needed to offset:    1.9 trees  🌳
Carbon efficiency:         89 gCO₂e per $1 spent

Region recommendations included. Move to us-west-2 and cut emissions by 94%.

---

🔔 Slack-First Alerts

Your engineering team lives in Slack. So does Valdrics.

• Anomaly alerts when costs spike unexpectedly
• Daily digests with top savings opportunities
• Leaderboards — "Who saved the most this week?"
• One-click approve/reject for remediations

---

🔁 Workflow Automation (GitHub/GitLab/CI)

Trigger runbooks directly from policy/remediation events:

• policy.block / policy.escalate / remediation.completed
• Native dispatch to GitHub Actions and GitLab CI
• Generic CI webhook fallback
• Deterministic evidence links embedded in event payloads

See configuration details in docs/integrations/workflow_automation.md.

---

🛡️ Enterprise-Grade Security

We're paranoid, so you don't have to be:

• Zero-Trust Architecture — We assume IAM roles via STS. No long-lived credentials.
• Read-Only by Default — Our CloudFormation/Terraform templates grant only Describe* and Get* permissions.
• Human-in-the-Loop — The AI recommends; you approve the action.
• GitOps-First Remediation — Generate professional Terraform plans (state rm and removed blocks) to decommission resources via your existing CI/CD.
• Audit Trail — Every remediation request is logged with who requested, who approved, and when.

---

🚀 Get Started

Prerequisites

• Docker & Docker Compose
• An AWS account with:
  • AWS CUR configured to deliver Parquet reports to S3
  • Resource Explorer 2 enabled
• Cost Explorer is optional (Valdrics ingestion path is CUR + Resource Explorer 2)
• An LLM API key (OpenAI, Anthropic, Google, or Groq)

Runtime Dependency Policy (Prod/Staging)

• tiktoken is required for accurate token accounting and LLM budget enforcement.
• If SENTRY_DSN is configured, sentry-sdk is required.
• prophet is required by default in staging/production.
• Temporary break-glass fallback is allowed only with:
  • FORECASTER_ALLOW_HOLT_WINTERS_FALLBACK=true
  • FORECASTER_BREAK_GLASS_REASON (auditable justification)
  • FORECASTER_BREAK_GLASS_EXPIRES_AT (ISO-8601 UTC expiry inside the configured max break-glass window)

1. Clone & Configure

git clone https://github.com/Valdrics/valdrics.git
cd valdrics
cp .env.example .env

For fast local sqlite development, generate the local runtime profile and bootstrap the current ORM schema instead of replaying the historical Alembic chain:

make env-dev
make bootstrap-local-db

.env.dev is local-only, runs with TESTING=false, and must not be used in staging/production.

For the Postgres/Redis docker path, edit .env and add:

DATABASE_URL=postgresql+asyncpg://...
OPENAI_API_KEY=sk-...  # or GROQ_API_KEY, etc.
SUPABASE_JWT_SECRET=your-jwt-secret

2. Start the Stack

Fast local sqlite path:

make dev

If .env.dev exists, make dev auto-loads it and bootstraps the local sqlite schema before starting the API.

Full dockerized Postgres/Redis path:

docker-compose up -d

3. Open the Dashboard

• API Docs: http://localhost:8000/docs
• Dashboard: http://localhost:5174

4. Connect Your AWS Account

The dashboard will guide you through deploying our read-only IAM role via CloudFormation or Terraform. Takes 60 seconds.

---

📊 Tech Stack

Layer	Technology
Backend	Python 3.12, FastAPI, Pydantic v2, SQLAlchemy (async)
Frontend	SvelteKit (Svelte 5 Runes), TailwindCSS v4, Shadcn-Svelte
Database	PostgreSQL (managed or self-hosted), Supabase-compatible auth flows
LLM	LangChain, OpenAI, Anthropic, Google Genai, Groq
Infra	Docker, Kubernetes (Helm), GitHub Actions, Prometheus
Observability	OpenTelemetry, Grafana Dashboards, Prometheus Metrics
GreenOps	CodeCarbon integration

---

🏗️ Production Infrastructure

Valdrics includes production-ready infrastructure components:

Kubernetes Deployment

# Deploy with Helm
helm install valdrics helm/valdrics/ \
  --set image.tag=latest \
  --set existingSecrets.name=valdrics-secrets

# Or customize values
helm install valdrics helm/valdrics/ -f my-values.yaml

Pre-configured Components

Component	Location	Description
Helm Chart	helm/valdrics/	Full K8s deployment (HPA, Ingress, Service)
Grafana Dashboards	grafana/dashboards/	API Overview + FinOps metrics
Load Tests	loadtest/	k6 + Locust performance tests
SBOM Generation	.github/workflows/sbom.yml	CycloneDX + vulnerability scanning

CI/CD Pipeline

• ✅ Linting: Ruff + MyPy
• ✅ Testing: Pytest with coverage
• ✅ Security: Bandit (SAST), Trivy (containers), TruffleHog (secrets)
• ✅ GreenOps: CodeCarbon emissions tracking
• ✅ E2E: Playwright browser tests

---

🗺️ Roadmap

We're in active development. Here's where we are:

✅ Done

• Multi-tenant AWS onboarding (CloudFormation + Terraform)
• 11 zombie detection plugins
• Multi-LLM analysis (OpenAI, Claude, Gemini, Groq)
• Carbon footprint calculator with regional intensity
• Slack integration (alerts, digests, leaderboards)
• SvelteKit dashboard with dark mode
• Human-in-the-loop remediation workflow
• Azure & GCP support (Adapters complete)
• ML-based forecasting (Prophet-integrated)
• ActiveOps (Autonomous Remediation Engine)

🔨 In Progress

• FinOps-as-Code (GitHub Action to preview cost changes on PRs)
• Real-time WebSocket updates
• Production deployment profile hardening and capacity validation

🔮 Coming Soon

• ClickHouse migration (for 100M+ scaling)
• Stripe billing & usage metering
• Virtual tagging (LLM infers team ownership)

---

📜 License

Valdrics is source available under the Business Source License 1.1 (BUSL-1.1, commonly called BSL 1.1).

• ✅ Free for internal use — Run it on your own infrastructure.
• ❌ No competing hosted service — You cannot offer Valdrics itself as a managed or hosted service to third parties without a separate commercial agreement.
• 🗓️ Freedom date: Converts to Apache 2.0 on January 5, 2030.

This repository is public on GitHub, but it is not OSI open source today. Public source availability does not change the license terms.

See LICENSE for full terms.

Additional policy docs:

• Licensing FAQ
• Commercial Licensing
• Trademark Policy
• Contributor License Agreement (CLA)
• Open-Core Boundary
• Tenancy ADR
• Pricing Metric Model

---

🤝 Contributing

We welcome contributions! Please read our CONTRIBUTING.md before opening a PR.

---

💖 Support

If Valdrics saved your team $1,000 this month, consider sponsoring the project:

---

Built with obsessive attention to detail by AbdulGoniyy Dare.
Because your cloud bill shouldn't keep you up at night.