Cost Analytics

As AI adoption grows, understanding costs becomes just as important as understanding performance.

Organizations need visibility into:

• Which AI flows generate the highest costs
• Which models contribute most to spending
• Which teams consume the largest budgets
• How spending changes over time
• Where optimization opportunities exist

Karyam provides cost visibility across AI flows, agents, models, and providers.

---

Where to Find Cost Analytics

AI flowRuns

Individual execution costs can be viewed from:

OPERATE
 └── Runs
      └── Info
           └── Usage

Example:

Cost
──────────────
Input Cost      $0.0124
Output Cost     $0.0042
System Cost     $0.0011

---

RAG Runs

Knowledge ingestion and retrieval costs are available from:

AI Infra
 └── RAG Runs
      └── Info
           └── Usage

This includes:

• Embedding costs
• Retrieval costs
• Completion costs

---

Cost Metrics

Input Cost

Cost generated by prompt tokens sent to the model.

Example:

Input Cost
$0.0124

---

Output Cost

Cost generated by tokens returned by the model.

Example:

Output Cost
$0.0042

---

System Cost

Additional costs associated with execution.

Examples include:

• Platform overhead
• Infrastructure services
• Future premium capabilities

Example:

System Cost
$0.0011

---

Dashboard Cost Analytics

Karyam provides organization-wide financial visibility through dashboard widgets.

Top Expensive AI flows

Identify AI flows generating the highest spending.

Useful for:

• Cost optimization
• AI flowredesign
• Model tuning

---

Monthly Cost Trend by Model

Monitor how spending changes over time for each model.

Examples:

• GPT-4o
• Claude Sonnet
• Gemini
• Bedrock Models

---

Monthly Total Model Cost Overview

Provides a consolidated cost view across all providers and models.

Useful for:

• Budget forecasting
• Provider analysis
• Executive reporting

---

Daily Token Cost

Track daily spending trends.

Useful for:

• Detecting spikes
• Budget monitoring
• Usage forecasting

---

Model Pricing Configuration

Model costs are calculated using pricing configured directly on each model.

Navigate to:

AI Infra
 └── Models

Each model supports:

Input Token Cost ($)

Cost per 1 million input tokens.

Example:

0.15

---

Output Token Cost ($)

Cost per 1 million output tokens.

Example:

0.60

---

Spending Cap ($)

Defines the maximum spend allowed for the model.

Example:

100

---

Unlimited Spending

Setting:

Spending Cap ($)
0

means:

Unlimited spending

---

Why Spending Caps Matter

Spending caps help organizations:

• Prevent runaway costs
• Control experimentation
• Enforce budgets
• Improve governance

---

Example Cost Breakdown

Model:
gpt-4o-mini

Input Cost:
$0.0037

Output Cost:
$0.0010

System Cost:
$0.0004

Total Cost:
$0.0051

---

Cost Optimization Best Practices

Choose The Right Model

Not every task requires the largest model.

---

Monitor Expensive AI flows

Regularly review AI flows appearing in:

• Top Expensive AI flows
• Monthly Cost Trends

---

Configure Spending Caps

Protect against unexpected spending by setting model limits.

---

Review Provider Usage

Monitor spending distribution across providers to optimize costs and reduce dependency.

---

Related Features

Cost Analytics works alongside:

• Token Usage
• Agent Logs
• AI flow Logs
• RAG Logs

Together they provide complete operational and financial visibility into AI systems.

---

Next Steps

Continue with:

➡️ Debugging

Learn how Karyam helps diagnose failures across agents, AI flows, retrieval systems, and integrations.

---