ai-infra-cost-estimator 1.0.0

Estimate real-world AI + cloud infrastructure costs for LLM applications

🚀 AI Infra Cost Estimator

Estimate real-world AI + cloud costs before scaling breaks you.

---

🎯 Problem

Most AI startups underestimate costs because LLM usage, concurrency, and infrastructure scaling are tightly coupled.

A small growth in requests can cause non-linear cost jumps that catch founders off guard.

This tool estimates:

• 💰 LLM API costs (tokens, caching impact)
• 🖥️ Infrastructure costs (compute, pods, scaling)
• 📈 Scaling thresholds (when costs jump)
• 💡 Optimization recommendations

Before you deploy.

---

✨ Features

• Multi-model support: OpenAI, Anthropic, Google, Meta, Mistral
• Infrastructure estimation: Pod/container scaling calculations
• Caching analysis: See how caching affects your costs
• Scaling breakpoints: Know when you'll need more resources
• Cost projections: See costs at 1.5x, 2x, 3x, 5x growth
• Optimization recommendations: Actionable cost-saving suggestions
• Multiple output formats: Markdown reports & JSON for automation

---

📊 Example Output

### Monthly Cost Summary

LLM Cost:     $1,820
Infra Cost:   $640
────────────────────────────
Total:        $2,460

Cost per Request: $0.0082
Yearly Estimate:  $29,520

### Scaling Alerts
- At 18,000 req/day → scale to 2 pods
- At 45,000 req/day → cost doubles to $4,920/month

### Recommendations
1. 🟠 Improve cache hit ratio to 40%
   Potential Savings: $520/month
   
2. 🟡 Consider switching to gpt-3.5-turbo
   Potential Savings: $890/month

---

🚀 Quick Start

Installation

Option 1: Install via pip (Recommended)

pip install ai-infra-cost-estimator

Option 2: Install from source

# Clone the repository
git clone https://github.com/MindTheInfraAI/AI_Infra_Cost_Estimator.git
cd AI_Infra_Cost_Estimator

# Install the package
pip install -e .

Basic Usage

# Run with a config file
ai-cost-estimator run config.yaml

# Output as JSON
ai-cost-estimator run config.yaml --format json

# Save report to file
ai-cost-estimator run config.yaml --output report.md

Create Your Configuration

# Generate sample config
ai-cost-estimator init my-config.yaml

Edit my-config.yaml:

requests_per_day: 10000
avg_input_tokens: 800
avg_output_tokens: 400
model: gpt-4o-mini
region: us-east-1
cache_hit_ratio: 0.2
concurrency_limit: 50

---

📖 CLI Commands

Command	Description
run <config>	Run cost estimation
run <config> --format json	Output as JSON
run <config> -o report.md	Save to file
list-models	Show available LLM models
list-regions	Show available cloud regions
list-instances	Show available instance types
compare <config>	Compare costs across all models
init [filename]	Create sample configuration

Examples

# List all supported models with pricing
ai-cost-estimator list-models

# Compare your workload across all models
ai-cost-estimator compare examples/startup.yaml

# Use a specific example config
ai-cost-estimator run examples/enterprise.yaml

---

📁 Configuration Reference

# Required Parameters
requests_per_day: 10000      # Daily API request volume
avg_input_tokens: 800        # Average tokens per input/prompt
avg_output_tokens: 400       # Average tokens per response
model: gpt-4o-mini           # LLM model name

# Optional Parameters (with defaults)
region: us-east-1            # Cloud region
cache_hit_ratio: 0.0         # Cache effectiveness (0.0 - 1.0)
concurrency_limit: 50        # Max concurrent requests per pod
avg_latency_ms: 500          # Average request latency
headroom_percent: 20.0       # Extra capacity buffer
instance_type: auto          # small, medium, large, xlarge, gpu_t4, gpu_a100

---

🤖 Supported Models

⚠️ Important Disclaimer: The pricing shown below are estimated values for reference purposes only. Actual costs may vary based on provider pricing changes, volume discounts, and regional variations. Please verify current pricing from the official provider documentation before making business decisions.

Model	Provider	Input/1K	Output/1K
gpt-4o	OpenAI	$0.0025	$0.01
gpt-4o-mini	OpenAI	$0.00015	$0.0006
gpt-4-turbo	OpenAI	$0.01	$0.03
gpt-3.5-turbo	OpenAI	$0.0005	$0.0015
claude-3-5-sonnet	Anthropic	$0.003	$0.015
claude-3-opus	Anthropic	$0.015	$0.075
claude-3-haiku	Anthropic	$0.00025	$0.00125
gemini-1.5-pro	Google	$0.00125	$0.005
gemini-1.5-flash	Google	$0.000075	$0.0003
llama-3.1-70b	Meta	$0.00079	$0.00079
llama-3.1-8b	Meta	$0.00018	$0.00018
mistral-large	Mistral	$0.002	$0.006
mistral-small	Mistral	$0.0002	$0.0006

Always check official pricing pages for current rates.

---

🌍 Supported Regions

Region	Location	Price Multiplier
us-east-1	N. Virginia	1.0x
us-west-2	Oregon	1.0x
eu-west-1	Ireland	1.1x
eu-central-1	Frankfurt	1.15x
ap-south-1	Mumbai	0.9x
ap-southeast-1	Singapore	1.05x
ap-northeast-1	Tokyo	1.2x

---

📂 Project Structure

ai-infra-cost-estimator/
├── ai_infra_cost_estimator/  # Main package (pip installable)
│   ├── __init__.py
│   ├── cli.py               # Command-line interface
│   ├── calculator/
│   │   ├── __init__.py
│   │   ├── llm.py           # LLM cost calculations
│   │   ├── infra.py         # Infrastructure cost calculations
│   │   └── scaling.py       # Scaling analysis & recommendations
│   ├── report/
│   │   ├── __init__.py
│   │   ├── markdown.py      # Markdown report generator
│   │   └── json_report.py   # JSON report generator
│   └── pricing/
│       ├── __init__.py
│       └── models.json      # Model & infra pricing data
├── examples/
│   ├── startup.yaml         # Startup configuration
│   ├── enterprise.yaml      # Enterprise configuration
│   ├── budget.yaml          # Budget-conscious config
│   ├── chatbot.yaml         # Chatbot application config
│   └── code_assistant.yaml  # Code assistant config
├── pyproject.toml           # Package configuration
├── MANIFEST.in
├── README.md
└── LICENSE

---

🧮 How It Works

LLM Cost Calculation

effective_requests = requests_per_day × (1 - cache_hit_ratio)
daily_tokens = effective_requests × (input_tokens + output_tokens)
monthly_cost = (daily_tokens / 1000) × price_per_1k × 30

Infrastructure Calculation

requests_per_second = requests_per_day / 86400
required_concurrency = rps × avg_latency_seconds × overhead_multiplier
required_pods = ceil(required_concurrency / concurrency_limit)
monthly_cost = pods × cost_per_hour × 720

Scaling Breakpoints

The analyzer finds:

• When you need to add pods
• When costs double
• Impact of cache improvements
• Cheaper model alternatives

---

👥 Who This Is For

• AI SaaS Founders - Understand costs before launch
• Platform Engineers - Plan infrastructure scaling
• Cloud Architects - Optimize deployment strategy
• Indie Hackers - Build within budget constraints

---

🔧 Programmatic Usage

from ai_infra_cost_estimator import (
    LLMCostCalculator,
    InfraCostCalculator,
    ScalingAnalyzer,
    MarkdownReportGenerator
)

# Calculate LLM costs
llm_calc = LLMCostCalculator()
llm_result = llm_calc.calculate(
    requests_per_day=10000,
    avg_input_tokens=800,
    avg_output_tokens=400,
    model="gpt-4o-mini",
    cache_hit_ratio=0.2
)

print(f"Monthly LLM Cost: ${llm_result.monthly_total_cost:.2f}")

# Calculate infrastructure costs
infra_calc = InfraCostCalculator()
infra_result = infra_calc.calculate(
    requests_per_day=10000,
    avg_latency_ms=500,
    concurrency_limit=50,
    region="us-east-1"
)

print(f"Monthly Infra Cost: ${infra_result.total_monthly_cost:.2f}")

# Get scaling analysis
analyzer = ScalingAnalyzer()
analysis = analyzer.analyze(
    requests_per_day=10000,
    avg_input_tokens=800,
    avg_output_tokens=400,
    model="gpt-4o-mini",
    cache_hit_ratio=0.2
)

for alert in analysis.scaling_alerts:
    print(f"⚠️ {alert}")

for rec in analysis.recommendations[:3]:
    print(f"💡 {rec.title}: Save ${rec.potential_savings:.2f}/month")

---

🛣️ Roadmap

• v1.1: GPU cost estimation for self-hosted models
• v1.2: Multi-region deployment cost comparison
• v1.3: API endpoint for integration
• v2.0: Real-time cloud integration (AWS, GCP, Azure)

---

📝 Assumptions

These estimates are based on:

• Stateless inference workloads
• HTTP-based serving
• Average token sizes (actual may vary)
• Public API pricing (enterprise pricing may differ)
• No model fine-tuning costs

Clear assumptions = trust.

---

🤝 Contributing

Contributions are welcome! Please:

1. Fork the repository
2. Create a feature branch
3. Make your changes
4. Submit a pull request

---

📄 License

MIT License - see LICENSE for details.

---

💬 Support

• Issues: GitHub Issues
• Discussions: GitHub Discussions

---

⚠️ Disclaimer

This tool provides estimated costs for planning and budgeting purposes only.

• All model pricing values are approximations and may not reflect current actual pricing
• LLM providers frequently update their pricing — always verify with official documentation
• Infrastructure costs vary by cloud provider, region, and specific configurations
• Enterprise agreements, volume discounts, and promotional pricing are not factored in
• Please recheck and validate all estimates before making financial or business decisions

Official Pricing Pages:

• OpenAI Pricing
• Anthropic Pricing
• Google AI Pricing
• Mistral Pricing

---

Built for builders who want to understand their AI costs before they scale.