> ## Documentation Index
> Fetch the complete documentation index at: https://cascadeflow-docs-readme-hermes-callout.mintlify.site/llms.txt
> Use this file to discover all available pages before exploring further.

# KPI-Weighted Routing

> Configure quality, cost, latency, and energy weights to encode business priorities into model routing decisions.

Inject business priorities into every model decision using KPI weights.

## Quality-First (Premium Workload)

```python theme={null}
import cascadeflow

cascadeflow.init(mode="enforce")

with cascadeflow.run(
    budget=2.00,
    kpi_weights={"quality": 0.8, "cost": 0.1, "latency": 0.1},
    kpi_targets={"quality": 0.9}
) as session:
    # Routes to highest-quality models within budget
    result = await agent.run("Draft a legal contract clause")
    print(session.summary())
```

## Cost-First (High-Volume Batch)

```python theme={null}
with cascadeflow.run(
    budget=5.00,
    kpi_weights={"cost": 0.7, "quality": 0.2, "latency": 0.1}
) as session:
    # Routes to cheapest models that meet quality floor
    for query in batch_queries:
        result = await agent.run(query)
    print(f"Total cost: ${session.summary()['cost_total']:.4f}")
```

## Latency-First (Real-Time)

```python theme={null}
with cascadeflow.run(
    kpi_weights={"latency": 0.7, "quality": 0.2, "cost": 0.1},
    max_latency_ms=2000.0
) as session:
    # Routes to fastest models, hard cap at 2 seconds
    result = await agent.run("Quick classification task")
```

## Energy-Aware (Carbon-Conscious)

```python theme={null}
with cascadeflow.run(
    kpi_weights={"quality": 0.4, "energy": 0.3, "cost": 0.3},
    max_energy=100.0
) as session:
    # Balances quality with energy efficiency
    result = await agent.run("Summarize this report")
    print(f"Energy used: {session.summary()['energy_used']:.1f} units")
```

## Per-Agent Profiles

```python theme={null}
@cascadeflow.agent(
    budget=0.10,
    kpi_weights={"cost": 0.9, "quality": 0.1}
)
async def triage_agent(query: str):
    """Quick classification — prioritize cost."""
    return await llm.complete(query)

@cascadeflow.agent(
    budget=2.00,
    kpi_weights={"quality": 0.9, "cost": 0.1},
    kpi_targets={"quality": 0.95}
)
async def analysis_agent(query: str):
    """Deep analysis — prioritize quality."""
    return await llm.complete(query)
```

## Quality Priors

The harness uses built-in quality priors for scoring:

| Model         | Quality Prior | Latency Prior |
| ------------- | ------------- | ------------- |
| o1            | 0.95          | 0.40          |
| gpt-4o        | 0.90          | 0.72          |
| gpt-4-turbo   | 0.88          | 0.66          |
| gpt-5-mini    | 0.86          | 0.84          |
| gpt-4o-mini   | 0.75          | 0.93          |
| gpt-3.5-turbo | 0.65          | 1.00          |