Skip to main content

​
Thinking Module

The thinking module provides configurable thinking budgets for LLM reasoning, allowing control over token and time budgets for extended thinking.

​
Installation

pip install praisonaiagents

​
Features

  • Token budgets for extended thinking
  • Time budgets for reasoning
  • Adaptive budget allocation
  • Budget tracking and reporting

​
Quick Start

from praisonaiagents import Agent
from praisonaiagents.thinking import ThinkingBudget

# Create a thinking budget
budget = ThinkingBudget(
    max_tokens=16000,
    max_time_seconds=60,
    adaptive=True
)

# Apply to agent
agent = Agent(
    name="Reasoner",
    instructions="You are a reasoning assistant.",
)
# Set thinking budget via property
agent.thinking_budget = budget

​
Classes

​
ThinkingBudget

Configure thinking budgets for agents. 
from praisonaiagents.thinking import ThinkingBudget

budget = ThinkingBudget(
    max_tokens=16000,
    max_time_seconds=60,
    adaptive=True
)

​
Constructor

ParameterTypeDefaultDescription
max_tokensint8000Maximum thinking tokens
max_time_secondsint30Maximum thinking time
adaptiveboolFalseAdapt budget based on task

​
ThinkingConfig

Configuration for thinking behavior. 
from praisonaiagents.thinking import ThinkingConfig

config = ThinkingConfig(
    enabled=True,
    budget=budget,
    show_thinking=False
)

​
Attributes

AttributeTypeDescription
enabledboolWhether thinking is enabled
budgetThinkingBudgetBudget configuration
show_thinkingboolShow thinking in output

​
ThinkingUsage

Track thinking usage. 
from praisonaiagents.thinking import ThinkingUsage

usage = ThinkingUsage(
    tokens_used=5000,
    time_seconds=15.5
)

​
Attributes

AttributeTypeDescription
tokens_usedintTokens used for thinking
time_secondsfloatTime spent thinking
budget_remainingintRemaining token budget

​
ThinkingTracker

Track thinking across multiple calls. 
from praisonaiagents.thinking import ThinkingTracker

tracker = ThinkingTracker()
tracker.record(usage)
print(tracker.total_tokens)
print(tracker.total_time)

​
Usage Examples

​
Basic Thinking Budget

from praisonaiagents import Agent
from praisonaiagents.thinking import ThinkingBudget

agent = Agent(
    name="Analyst",
    instructions="You are an analytical assistant.",
)
agent.thinking_budget = ThinkingBudget(max_tokens=16000)

# Agent will use extended thinking for complex tasks
result = agent.chat("Analyze the implications of quantum computing on cryptography")

​
Adaptive Budget

from praisonaiagents.thinking import ThinkingBudget

# Budget adapts based on task complexity
budget = ThinkingBudget(
    max_tokens=32000,
    adaptive=True  # Automatically adjusts based on task
)

agent = Agent(
    name="Researcher",
    instructions="You are a research assistant.",
)
agent.thinking_budget = budget

​
Time-Limited Thinking

from praisonaiagents.thinking import ThinkingBudget

# Limit thinking time for faster responses
budget = ThinkingBudget(
    max_tokens=16000,
    max_time_seconds=30  # Stop thinking after 30 seconds
)

​
Track Usage

from praisonaiagents.thinking import ThinkingTracker

tracker = ThinkingTracker()

# After agent runs
usage = agent.get_thinking_usage()
tracker.record(usage)

print(f"Total thinking tokens: {tracker.total_tokens}")
print(f"Total thinking time: {tracker.total_time}s")

​
Best Practices

  1. Set appropriate budgets - Balance quality vs. cost
  2. Use adaptive mode - Let the system optimize for task complexity
  3. Monitor usage - Track thinking costs over time
  4. Time limits - Set time limits for time-sensitive applications