"""
Pattern 025: Action Sequence Planning
Description:
The Action Sequence Planning pattern enables agents to decompose complex goals
into ordered sequences of actions with preconditions, effects, and dependencies.
The agent analyzes the goal, identifies required actions, determines ordering
constraints, validates feasibility, and executes the plan step-by-step with
monitoring and replanning capabilities.
Components:
- Action Library: Catalog of available actions with specifications
- Planner: Generates action sequences to achieve goals
- Dependency Analyzer: Identifies action dependencies and constraints
- Feasibility Checker: Validates plan viability
- Executor: Executes plans with monitoring and error handling
Use Cases:
- Multi-step task automation
- Workflow orchestration
- Robot task planning
- Complex problem decomposition
- Project planning and execution
LangChain Implementation:
Uses LLM for action selection and ordering, implements STRIPS-like
planning with preconditions and effects, handles dynamic replanning.
Key Features:
- Action dependency analysis
- Precondition and effect modeling
- Parallel action detection
- Plan validation
- Dynamic replanning on failures
"""
import os
from typing import List, Dict, Any, Optional, Set
from dataclasses import dataclass, field
from enum import Enum
from datetime import datetime
from dotenv import load_dotenv
from langchain_core.prompts import ChatPromptTemplate
from langchain_core.output_parsers import StrOutputParser
from langchain_openai import ChatOpenAI
load_dotenv()
class ActionStatus(Enum):
"""Status of action execution."""
PENDING = "pending"
READY = "ready"
EXECUTING = "executing"
COMPLETED = "completed"
FAILED = "failed"
SKIPPED = "skipped"
@dataclass
class ActionDefinition:
"""Definition of an available action."""
name: str
description: str
preconditions: List[str]
effects: List[str]
duration: float = 1.0 # Estimated duration
def to_dict(self) -> Dict[str, Any]:
"""Convert to dictionary for LLM context."""
return {
"name": self.name,
"description": self.description,
"preconditions": self.preconditions,
"effects": self.effects
}
@dataclass
class PlannedAction:
"""An action in a plan."""
action: ActionDefinition
step_number: int
dependencies: List[int] = field(default_factory=list)
status: ActionStatus = ActionStatus.PENDING
result: Optional[str] = None
error: Optional[str] = None
start_time: Optional[datetime] = None
end_time: Optional[datetime] = None
@dataclass
class ActionPlan:
"""Complete action plan."""
goal: str
actions: List[PlannedAction]
initial_state: Set[str]
final_state: Set[str]
timestamp: datetime = field(default_factory=datetime.now)
def get_ready_actions(self) -> List[PlannedAction]:
"""Get actions that are ready to execute."""
ready = []
for action in self.actions:
if action.status == ActionStatus.PENDING:
# Check if all dependencies completed
dependencies_met = all(
self.actions[dep].status == ActionStatus.COMPLETED
for dep in action.dependencies
)
if dependencies_met:
ready.append(action)
return ready
def is_complete(self) -> bool:
"""Check if plan is complete."""
return all(
action.status in [ActionStatus.COMPLETED, ActionStatus.SKIPPED]
for action in self.actions
)
class ActionLibrary:
"""
Library of available actions.
"""
def __init__(self):
"""Initialize action library with common actions."""
self.actions: Dict[str, ActionDefinition] = {}
self._register_builtin_actions()
def _register_builtin_actions(self):
"""Register built-in actions."""
# Data collection actions
self.register_action(ActionDefinition(
name="gather_requirements",
description="Collect and document project requirements",
preconditions=["project_defined"],
effects=["requirements_gathered"]
))
self.register_action(ActionDefinition(
name="research_topic",
description="Research and gather information on a topic",
preconditions=[],
effects=["research_completed", "information_gathered"]
))
# Analysis actions
self.register_action(ActionDefinition(
name="analyze_data",
description="Analyze collected data or information",
preconditions=["data_available"],
effects=["analysis_completed", "insights_generated"]
))
self.register_action(ActionDefinition(
name="identify_risks",
description="Identify and assess potential risks",
preconditions=["requirements_gathered"],
effects=["risks_identified"]
))
# Design actions
self.register_action(ActionDefinition(
name="create_design",
description="Create system or solution design",
preconditions=["requirements_gathered", "research_completed"],
effects=["design_created"]
))
self.register_action(ActionDefinition(
name="review_design",
description="Review and validate design",
preconditions=["design_created"],
effects=["design_reviewed"]
))
# Implementation actions
self.register_action(ActionDefinition(
name="implement_solution",
description="Implement the designed solution",
preconditions=["design_reviewed"],
effects=["solution_implemented"]
))
self.register_action(ActionDefinition(
name="write_documentation",
description="Write technical documentation",
preconditions=["solution_implemented"],
effects=["documentation_written"]
))
# Testing actions
self.register_action(ActionDefinition(
name="create_test_plan",
description="Create comprehensive test plan",
preconditions=["design_created"],
effects=["test_plan_created"]
))
self.register_action(ActionDefinition(
name="execute_tests",
description="Execute test cases and validate solution",
preconditions=["solution_implemented", "test_plan_created"],
effects=["tests_executed", "quality_validated"]
))
# Deployment actions
self.register_action(ActionDefinition(
name="prepare_deployment",
description="Prepare solution for deployment",
preconditions=["tests_executed", "documentation_written"],
effects=["deployment_ready"]
))
self.register_action(ActionDefinition(
name="deploy_solution",
description="Deploy solution to production",
preconditions=["deployment_ready"],
effects=["solution_deployed"]
))
def register_action(self, action: ActionDefinition):
"""Register a new action."""
self.actions[action.name] = action
def get_action(self, name: str) -> Optional[ActionDefinition]:
"""Get action by name."""
return self.actions.get(name)
def get_all_actions_description(self) -> str:
"""Get formatted description of all actions."""
descriptions = []
for action in self.actions.values():
precond = ", ".join(action.preconditions) if action.preconditions else "none"
effects = ", ".join(action.effects)
descriptions.append(
f"- {action.name}: {action.description}\n"
f" Preconditions: {precond}\n"
f" Effects: {effects}"
)
return "\n\n".join(descriptions)
class ActionPlanner:
"""
Plans action sequences to achieve goals.
"""
def __init__(
self,
action_library: ActionLibrary,
model: str = "gpt-3.5-turbo"
):
"""
Initialize action planner.
Args:
action_library: Library of available actions
model: LLM model to use
"""
self.action_library = action_library
self.llm = ChatOpenAI(model=model, temperature=0.3)
def create_plan(
self,
goal: str,
initial_state: Optional[Set[str]] = None
) -> ActionPlan:
"""
Create action plan to achieve goal.
Args:
goal: Goal to achieve
initial_state: Initial world state
Returns:
Action plan
"""
if initial_state is None:
initial_state = set()
actions_desc = self.action_library.get_all_actions_description()
prompt = ChatPromptTemplate.from_messages([
("system", """You are an expert action planner. Create a sequence of actions
to achieve the given goal, respecting preconditions and dependencies.
Available Actions:
{actions_description}
Your response should list actions in order, one per line, using exact action names.
Only use actions from the available list."""),
("user", """Goal: {goal}
Initial State: {initial_state}
Create an action plan listing actions in execution order:""")
])
chain = prompt | self.llm | StrOutputParser()
initial_state_str = ", ".join(initial_state) if initial_state else "none"
response = chain.invoke({
"actions_description": actions_desc,
"goal": goal,
"initial_state": initial_state_str
})
# Parse action names
action_names = self._parse_action_names(response)
# Build plan with dependencies
planned_actions = self._build_plan_with_dependencies(
action_names,
initial_state
)
# Calculate final state
final_state = self._calculate_final_state(planned_actions, initial_state)
return ActionPlan(
goal=goal,
actions=planned_actions,
initial_state=initial_state,
final_state=final_state
)
def _parse_action_names(self, response: str) -> List[str]:
"""Parse action names from LLM response."""
lines = response.strip().split("\n")
action_names = []
for line in lines:
line = line.strip()
# Remove numbering, bullets, etc.
for prefix in ["1.", "2.", "3.", "4.", "5.", "6.", "7.", "8.", "9.", "-", "*", "•"]:
if line.startswith(prefix):
line = line[len(prefix):].strip()
# Check if this is a valid action name
if self.action_library.get_action(line):
action_names.append(line)
return action_names
def _build_plan_with_dependencies(
self,
action_names: List[str],
initial_state: Set[str]
) -> List[PlannedAction]:
"""Build plan with dependency analysis."""
planned_actions = []
current_state = initial_state.copy()
for step_num, action_name in enumerate(action_names):
action_def = self.action_library.get_action(action_name)
if not action_def:
continue
# Find dependencies (actions that provide required preconditions)
dependencies = []
for precond in action_def.preconditions:
if precond not in current_state:
# Find which previous action provides this
for i, prev_action in enumerate(planned_actions):
if precond in prev_action.action.effects:
dependencies.append(i)
break
planned_action = PlannedAction(
action=action_def,
step_number=step_num,
dependencies=dependencies
)
planned_actions.append(planned_action)
# Update state with effects
current_state.update(action_def.effects)
return planned_actions
def _calculate_final_state(
self,
planned_actions: List[PlannedAction],
initial_state: Set[str]
) -> Set[str]:
"""Calculate final state after all actions."""
final_state = initial_state.copy()
for planned_action in planned_actions:
final_state.update(planned_action.action.effects)
return final_state
class PlanExecutor:
"""
Executes action plans.
"""
def __init__(self):
"""Initialize plan executor."""
pass
def execute_plan(self, plan: ActionPlan) -> Dict[str, Any]:
"""
Execute action plan.
Args:
plan: Action plan to execute
Returns:
Execution results
"""
print(f"\n[Executor] Executing plan for goal: {plan.goal}")
print(f"[Executor] Total actions: {len(plan.actions)}\n")
completed = 0
failed = 0
while not plan.is_complete():
# Get ready actions
ready_actions = plan.get_ready_actions()
if not ready_actions:
# Check if we have pending actions that can't proceed
pending = [a for a in plan.actions if a.status == ActionStatus.PENDING]
if pending:
print("[Executor] ✗ Deadlock: Actions have unmet dependencies")
for action in pending:
action.status = ActionStatus.FAILED
action.error = "Unmet dependencies"
failed += 1
break
# Execute ready actions
for action in ready_actions:
self._execute_action(action)
if action.status == ActionStatus.COMPLETED:
completed += 1
elif action.status == ActionStatus.FAILED:
failed += 1
success_rate = completed / len(plan.actions) * 100 if plan.actions else 0
return {
"goal": plan.goal,
"total_actions": len(plan.actions),
"completed": completed,
"failed": failed,
"success_rate": success_rate,
"plan": plan
}
def _execute_action(self, action: PlannedAction):
"""Execute a single action."""
action.status = ActionStatus.EXECUTING
action.start_time = datetime.now()
print(f"[Step {action.step_number + 1}] Executing: {action.action.name}")
print(f" Description: {action.action.description}")
if action.dependencies:
deps_str = ", ".join([str(d + 1) for d in action.dependencies])
print(f" Dependencies: Steps {deps_str}")
# Simulate action execution (in real system, would call actual function)
try:
# Simulate success
action.result = f"Successfully completed {action.action.name}"
action.status = ActionStatus.COMPLETED
print(f" ✓ Completed")
print(f" Effects: {', '.join(action.action.effects)}")
except Exception as e:
action.error = str(e)
action.status = ActionStatus.FAILED
print(f" ✗ Failed: {action.error}")
action.end_time = datetime.now()
print()
class ActionSequencePlanningAgent:
"""
Agent that plans and executes action sequences.
"""
def __init__(self, model: str = "gpt-3.5-turbo"):
"""
Initialize action sequence planning agent.
Args:
model: LLM model to use
"""
self.action_library = ActionLibrary()
self.planner = ActionPlanner(self.action_library, model)
self.executor = PlanExecutor()
def achieve_goal(
self,
goal: str,
initial_state: Optional[Set[str]] = None
) -> Dict[str, Any]:
"""
Plan and execute actions to achieve goal.
Args:
goal: Goal to achieve
initial_state: Initial world state
Returns:
Execution results
"""
print(f"\n{'=' * 80}")
print(f"GOAL: {goal}")
print("=" * 80)
# Create plan
print("\n[Agent] Creating action plan...")
plan = self.planner.create_plan(goal, initial_state)
print(f"\n[Agent] Plan created with {len(plan.actions)} actions")
print("\nACTION PLAN:")
print("-" * 80)
for action in plan.actions:
deps = ""
if action.dependencies:
deps = f" (depends on: {[d+1 for d in action.dependencies]})"
print(f"{action.step_number + 1}. {action.action.name}{deps}")
print(f" {action.action.description}")
print("-" * 80)
print(f"\nInitial state: {', '.join(plan.initial_state) if plan.initial_state else 'none'}")
print(f"Expected final state: {', '.join(plan.final_state)}")
# Execute plan
print(f"\n{'=' * 80}")
print("EXECUTION")
print("=" * 80)
result = self.executor.execute_plan(plan)
# Summary
print("=" * 80)
print("EXECUTION SUMMARY")
print("=" * 80)
print(f"Goal: {result['goal']}")
print(f"Total actions: {result['total_actions']}")
print(f"Completed: {result['completed']}")
print(f"Failed: {result['failed']}")
print(f"Success rate: {result['success_rate']:.0f}%")
return result
def demonstrate_action_sequence_planning():
"""Demonstrate the Action Sequence Planning pattern."""
print("=" * 80)
print("ACTION SEQUENCE PLANNING PATTERN DEMONSTRATION")
print("=" * 80)
agent = ActionSequencePlanningAgent()
# Show available actions
print("\n" + "=" * 80)
print("AVAILABLE ACTIONS")
print("=" * 80)
print(agent.action_library.get_all_actions_description())
# Test 1: Software project workflow
print("\n\n" + "=" * 80)
print("TEST 1: Software Development Project")
print("=" * 80)
goal1 = "Complete a software development project from requirements to deployment"
initial_state1 = {"project_defined"}
result1 = agent.achieve_goal(goal1, initial_state1)
# Test 2: Research and analysis task
print("\n\n" + "=" * 80)
print("TEST 2: Research and Analysis Task")
print("=" * 80)
goal2 = "Research a topic, analyze findings, and create a report"
initial_state2 = set()
result2 = agent.achieve_goal(goal2, initial_state2)
# Test 3: Design and implementation
print("\n\n" + "=" * 80)
print("TEST 3: Design and Implementation")
print("=" * 80)
goal3 = "Design, implement, and test a new feature"
initial_state3 = {"requirements_gathered", "research_completed"}
result3 = agent.achieve_goal(goal3, initial_state3)
# Overall summary
print("\n\n" + "=" * 80)
print("PATTERN SUMMARY")
print("=" * 80)
print("""
The Action Sequence Planning pattern demonstrates:
1. **Goal Decomposition**: Breaks complex goals into action sequences
2. **Dependency Analysis**: Identifies action dependencies and ordering
3. **Precondition Checking**: Ensures actions are feasible
4. **Effect Modeling**: Tracks state changes from actions
5. **Sequential Execution**: Executes plan respecting dependencies
Key Benefits:
- **Structured Approach**: Systematic goal achievement
- **Dependency Management**: Proper action ordering
- **State Tracking**: Monitors world state changes
- **Validation**: Ensures plan feasibility
- **Transparency**: Clear action sequences and reasoning
Use Cases:
- Multi-step task automation
- Workflow orchestration
- Project planning and execution
- Robot task planning
- Business process automation
Planning Principles:
- Actions have clear preconditions and effects
- Dependencies are automatically inferred
- Plans respect causal relationships
- State evolves through action effects
- Parallel execution possible when no dependencies
This pattern enables agents to tackle complex, multi-step goals
through structured planning and systematic execution.
""")
if __name__ == "__main__":
demonstrate_action_sequence_planning()
Agentic Design Pattern: Action Sequence Planning
Pattern 025: Action Sequence Planning – Line-by-Line Explanation
Overview
This document provides a detailed line-by-line explanation of the Action Sequence Planning pattern implementation.
Pattern Description
Pattern 025: Action Sequence Planning
Description: The Action Sequence Planning pattern enables agents to decompose complex goals into ordered sequences of actions with preconditions, effects, and dependencies. The agent analyzes the goal, identifies required actions, determines ordering constraints, validates feasibility, and executes the plan step-by-step with monitoring and replanning capabilities.
Components: – Action Library: Catalog of available actions with specifications – Planner: Generates action sequences to achieve goals – Dependency Analyzer: Identifies action dependencies and constraints – Feasibility Checker: Validates plan viability – Executor: Executes plans with monitoring and error handling
Use Cases: – Multi-step task automation – Workflow orchestration – Robot task planning – Complex problem decomposition – Project planning and execution
LangChain Implementation: Uses LLM for action selection and ordering, implements STRIPS-like planning with preconditions and effects, handles dynamic replanning.
Key Features: – Action dependency analysis – Precondition and effect modeling – Parallel action detection – Plan validation – Dynamic replanning on failures
File Statistics
- Total Lines: 652
- Classes: 8
- Functions: 1
- Constants: 0
Module Documentation
"""
Pattern 025: Action Sequence Planning
Description:
The Action Sequence Planning pattern enables agents to decompose complex goals
into ordered sequences of actions with preconditions, effects, and dependencies.
The agent analyzes the goal, identifies required actions, determines ordering
constraints, validates feasibility, and executes the plan step-by-step with
monitoring and replanning capabilities.
Components:
- Action Library: Catalog of available actions with specifications
- Planner: Generates action sequences to achieve goals
- Dependency Analyzer: Identifies action dependencies and constraints
- Feasibility Checker: Validates plan viability
- Executor: Executes plans with monitoring and error handling
Use Cases:
- Multi-step task automation
- Workflow orchestration
- Robot task planning
- Complex problem decomposition
- Project planning and execution
LangChain Implementation:
Uses LLM for action selection and ordering, implements STRIPS-like
planning with preconditions and effects, handles dynamic replanning.
Key Features:
- Action dependency analysis
- Precondition and effect modeling
- Parallel action detection
- Plan validation
- Dynamic replanning on failures
"""
Purpose: This module implements the Action Sequence Planning pattern, providing a structured approach to agent design and implementation.
Import Statements
Standard Library Imports
import dataclasses.dataclass
import dataclasses.field
import datetime.datetime
import enum.Enum
import langchain_core.output_parsers.StrOutputParser
import langchain_core.prompts.ChatPromptTemplate
import os
import typing.Any
import typing.Dict
import typing.List
import typing.Optional
import typing.Set
Purpose: Core Python functionality:
– dataclasses.dataclass: Simplified class definitions with @dataclass
– dataclasses.field: Simplified class definitions with @dataclass
– datetime.datetime: Date and time handling
– enum.Enum: Enumeration support
– langchain_core.output_parsers.StrOutputParser: Regular expression operations
– langchain_core.prompts.ChatPromptTemplate: Regular expression operations
– os: Environment variable access and system operations
– typing.Any: Type hints for better code documentation (List, Dict, Any, Optional)
– typing.Dict: Type hints for better code documentation (List, Dict, Any, Optional)
– typing.List: Type hints for better code documentation (List, Dict, Any, Optional)
– typing.Optional: Type hints for better code documentation (List, Dict, Any, Optional)
– typing.Set: Type hints for better code documentation (List, Dict, Any, Optional)
LangChain Imports
from langchain_openai.ChatOpenAI
Purpose: LangChain framework components: – LLM Interface: Wrapper for OpenAI models – Chains: Composable building blocks for complex workflows
Other Imports
from dotenv.load_dotenv
Purpose: Additional dependencies for specialized functionality.
Classes
Class: ActionStatus
Lines 51-58
Inherits from: Enum
Purpose:
Status of action execution.
Class: ActionDefinition
Lines 62-77
Purpose:
Definition of an available action.
Methods:
to_dict(self)
Lines 70-77
Purpose:
Convert to dictionary for LLM context.
Parameters:
Returns: <ast.Subscript object at 0x000001A06D60F5D0>
Class: PlannedAction
Lines 81-90
Purpose:
An action in a plan.
Class: ActionPlan
Lines 94-121
Purpose:
Complete action plan.
Methods:
get_ready_actions(self)
Lines 102-114
Purpose:
Get actions that are ready to execute.
Parameters:
Returns: <ast.Subscript object at 0x000001A06D645450>
is_complete(self)
Lines 116-121
Purpose:
Check if plan is complete.
Parameters:
Returns: bool
Class: ActionLibrary
Lines 124-246
Purpose:
Library of available actions.
Methods:
__init__(self)
Lines 129-132
Purpose:
Initialize action library with common actions.
Parameters:
_register_builtin_actions(self)
Lines 134-225
Purpose:
Register built-in actions.
Parameters:
register_action(self, action)
Lines 227-229
Purpose:
Register a new action.
Parameters:
– action: Action
get_action(self, name)
Lines 231-233
Purpose:
Get action by name.
Parameters:
– name: Name
Returns: <ast.Subscript object at 0x000001A06D7512D0>
get_all_actions_description(self)
Lines 235-246
Purpose:
Get formatted description of all actions.
Parameters:
Returns: str
Class: ActionPlanner
Lines 249-400
Purpose:
Plans action sequences to achieve goals.
Methods:
__init__(self, action_library, model)
Lines 254-267
Purpose:
Initialize action planner.
Args:
action_library: Library of available actions
model: LLM model to use
Parameters:
– action_library: Action library
– model: Language model instance
create_plan(self, goal, initial_state)
Lines 269-332
Purpose:
Create action plan to achieve goal.
Args:
goal: Goal to achieve
initial_state: Initial world state
Returns:
Action plan
Parameters:
– goal: Goal
– initial_state: Initial state
Returns: ActionPlan
_parse_action_names(self, response)
Lines 334-350
Purpose:
Parse action names from LLM response.
Parameters:
– response: Response
Returns: <ast.Subscript object at 0x000001A06D648550>
_build_plan_with_dependencies(self, action_names, initial_state)
Lines 352-387
Purpose:
Build plan with dependency analysis.
Parameters:
– action_names: Action names
– initial_state: Initial state
Returns: <ast.Subscript object at 0x000001A06D63C510>
_calculate_final_state(self, planned_actions, initial_state)
Lines 389-400
Purpose:
Calculate final state after all actions.
Parameters:
– planned_actions: Planned actions
– initial_state: Initial state
Returns: <ast.Subscript object at 0x000001A06D6B2710>
Class: PlanExecutor
Lines 403-489
Purpose:
Executes action plans.
Methods:
__init__(self)
Lines 408-410
Purpose:
Initialize plan executor.
Parameters:
execute_plan(self, plan)
Lines 412-461
Purpose:
Execute action plan.
Args:
plan: Action plan to execute
Returns:
Execution results
Parameters:
– plan: Plan
Returns: <ast.Subscript object at 0x000001A06D64FF10>
_execute_action(self, action)
Lines 463-489
Purpose:
Execute a single action.
Parameters:
– action: Action
Class: ActionSequencePlanningAgent
Lines 492-563
Purpose:
Agent that plans and executes action sequences.
Methods:
__init__(self, model)
Lines 497-506
Purpose:
Initialize action sequence planning agent.
Args:
model: LLM model to use
Parameters:
– model: Language model instance
achieve_goal(self, goal, initial_state)
Lines 508-563
Purpose:
Plan and execute actions to achieve goal.
Args:
goal: Goal to achieve
initial_state: Initial world state
Returns:
Execution results
Parameters:
– goal: Goal
– initial_state: Initial state
Returns: <ast.Subscript object at 0x000001A06D6A8650>
Functions
demonstrate_action_sequence_planning()
Lines 566-647
Purpose:
Demonstrate the Action Sequence Planning pattern.
Key Concepts Summary
Pattern Overview
The Action Sequence Planning pattern provides a structured approach to building AI agents with specific capabilities and behaviors. This implementation leverages LangChain’s composable components to create a robust and maintainable solution.
Core Components
- Agent Class: Main implementation of the pattern logic
- Prompt Engineering: Carefully crafted prompts to guide LLM behavior
- Tool Integration: Optional tools for extended functionality
- Memory Management: State handling across interactions
- Output Parsing: Structured response processing
When to Use
✅ Good for: – Scenarios requiring this specific pattern’s strengths – Applications needing structured agent behavior – Systems where explainability is important – Production environments requiring reliability
❌ Not ideal for: – Simple single-prompt tasks – Scenarios requiring different pattern characteristics – Resource-constrained environments (if computationally intensive)
Best Practices Demonstrated
1. Environment Configuration
- ✅ Use
.envfiles for API keys - ✅ Never hardcode credentials
- ✅ Validate environment variables on startup
2. Error Handling
- ✅ Graceful degradation on failures
- ✅ Informative error messages
- ✅ Logging for debugging
3. Prompt Engineering
- ✅ Clear, specific instructions
- ✅ Structured output formats
- ✅ Few-shot examples when beneficial
4. Code Organization
- ✅ Separation of concerns (class-based design)
- ✅ Type hints for clarity
- ✅ Docstrings for documentation
5. Performance Optimization
- ✅ Appropriate temperature settings
- ✅ Token budget management
- ✅ Caching where applicable
LangChain Components Used
- ChatOpenAI: OpenAI model interface
- Chains: Sequential component composition
Why LangChain? – Composable abstractions for LLM applications – Rich ecosystem of integrations – Production-ready patterns – Active community and development
Execution Flow
1. Initialize agent with configuration
↓
2. Load environment and dependencies
↓
3. Prepare prompt with input data
↓
4. Invoke LLM through LangChain
↓
5. Parse and validate response
↓
6. Return formatted result
Key Decision Points: – Model selection (temperature, model type) – Prompt structure (few-shot vs zero-shot) – Error handling strategy – Output format validation
Performance Considerations
Computational Cost
- LLM Calls: Primary cost factor (based on tokens)
- Latency: Depends on model and input size
- Memory: Minimal for most patterns
Optimization Strategies
- Prompt Optimization: Minimize tokens while maintaining clarity
- Caching: Cache repeated queries/responses
- Model Selection: Use smaller models when appropriate
- Batch Processing: Group similar requests
- Streaming: For long responses, stream tokens
Cost Estimates (Approximate)
- GPT-3.5-turbo: ~$0.002 per request
- GPT-4: ~$0.03 per request
- GPT-4-turbo: ~$0.01 per request
Actual costs vary based on input/output token counts
Common Pitfalls to Avoid
1. Prompt Design Issues
❌ Vague or ambiguous instructions
✅ Clear, specific, testable prompts
2. Error Handling
❌ Ignoring API errors or rate limits
✅ Implement retry logic and graceful degradation
3. Token Management
❌ Exceeding model token limits
✅ Truncate inputs, chunk processing
4. Security
❌ Exposing API keys in code
✅ Use environment variables
5. Testing
❌ Only testing happy paths
✅ Test edge cases, failures, malformed inputs
6. Cost Management
❌ Unmonitored API usage
✅ Implement budgets, alerts, and monitoring
Further Improvements
Production Enhancements
- Monitoring: Add logging, metrics, tracing
- Validation: Input/output schema validation
- Rate Limiting: Implement request throttling
- Caching: Redis or in-memory cache for responses
- Async Support: Non-blocking operations for scalability
Feature Extensions
- Multi-Model Support: Switch between providers
- Fallback Models: Graceful degradation strategy
- Custom Tools: Domain-specific function calling
- Advanced Memory: Vector stores for long-term memory
- Evaluation: Automated testing and benchmarking
Code Quality
- Testing: Unit tests, integration tests
- Documentation: API docs, usage examples
- Type Safety: Stricter type hints
- Configuration: Externalized config files
Related Patterns
- Pattern 024
- Pattern 026
- Other Tool Use Patterns (020-025)
Pattern Families: – Reasoning Patterns (001-011) – Multi-Agent Patterns (012-019) – Tool Use Patterns (020-025) – Memory Patterns (026-031) – And many more…
Summary
The Action Sequence Planning pattern provides a powerful approach to building AI agents with specific capabilities. This implementation demonstrates:
✅ Key Strengths: – Clean, maintainable code structure – LangChain integration for composability – Type hints and documentation – Production-ready patterns
🎯 Best Use Cases: – Applications requiring this pattern’s characteristics – Systems needing structured agent behavior – Production environments requiring reliability
📚 Learning Resources: – LangChain Documentation: https://python.langchain.com/ – Pattern Documentation: See related patterns – OpenAI API Reference: https://platform.openai.com/docs/
File: 025_action_sequence_planning.py
Total Lines: 652
Pattern Category: AI Agent Design Pattern #025
Generated from actual implementation code
For educational and reference purposes
Leave a Reply