Skip to main content

​
Presets

Create a CascadeAgent with a single function call using preset profiles that configure models, quality thresholds, and routing strategies.

​
Preset Functions

from cascadeflow.utils.presets import (
    get_cost_optimized_agent,
    get_balanced_agent,
    get_speed_optimized_agent,
    get_quality_optimized_agent,
    get_development_agent,
    auto_agent,
)

​
Cost Optimized

Maximizes savings by using the cheapest models first with aggressive cascading. 
agent = get_cost_optimized_agent(verbose=True)

​
Balanced

Default tradeoff between cost, quality, and speed. 
agent = get_balanced_agent()

​
Speed Optimized

Prioritizes low latency — prefers fast models and direct routing. 
agent = get_speed_optimized_agent()

​
Quality Optimized

Prioritizes response quality — higher thresholds, more willing to escalate to verifier. 
agent = get_quality_optimized_agent()

​
Development

Uses free/local models for development and testing. 
agent = get_development_agent(verbose=True)

​
Auto Agent

Create from a profile name string. 
agent = auto_agent("cost_optimized")
agent = auto_agent("balanced")
agent = auto_agent("speed_optimized")
agent = auto_agent("quality_optimized")
agent = auto_agent("development")

​
From Profile

CascadeAgent.from_profile() is the class method equivalent: 
from cascadeflow import CascadeAgent

agent = CascadeAgent.from_profile("balanced", verbose=True)

​
From Environment

Auto-detect available providers from environment variables: 
agent = CascadeAgent.from_env(verbose=True)
# Detects OPENAI_API_KEY, ANTHROPIC_API_KEY, GROQ_API_KEY, etc.