Conditional routing in graphs in LangChain - Performance & Optimization

Performance: Conditional routing in graphs

MEDIUM IMPACT

This affects the speed of decision-making and response time in graph-based workflows by controlling which nodes execute based on conditions.

Routing user queries through graph nodes based on conditions

LangChain

if condition:
  result = nodeA.run()
else:
  result = nodeB.run()
# Only one node runs based on condition

Runs only the necessary node, reducing CPU usage and speeding up response.

📈 Performance GainSingle node execution, cutting processing time roughly in half compared to bad pattern.

Routing user queries through graph nodes based on conditions

LangChain

graph.add_edge('start', 'nodeA')
graph.add_edge('start', 'nodeB')
# Both nodes run regardless of condition
resultA = nodeA.run()
resultB = nodeB.run()
# Condition checked after both run

Both nodes execute even if only one is needed, wasting CPU and increasing response time.

📉 Performance CostTriggers multiple unnecessary node executions, increasing processing time and delaying response.

Performance Comparison

Pattern	Node Executions	Conditional Checks	Processing Time	Verdict
Run all nodes then check condition	All nodes run	After execution	High due to wasted runs	[X] Bad
Check condition before running nodes	Only needed node runs	Before execution	Lower, efficient processing	[OK] Good

Rendering Pipeline

Conditional routing affects the execution flow in the graph processing pipeline, determining which nodes run and when results propagate.

→Decision Evaluation

→Node Execution

→Result Propagation

⚠️ BottleneckNode Execution stage is most expensive when unnecessary nodes run.

Core Web Vital Affected

INP

This affects the speed of decision-making and response time in graph-based workflows by controlling which nodes execute based on conditions.

Optimization Tips

1Evaluate conditions before executing graph nodes to avoid wasted work.

2Minimize the number of nodes executed per request to speed up response.

3Use early exits in routing logic to reduce CPU and memory usage.

Performance Quiz - 3 Questions

What is the main performance benefit of checking conditions before running graph nodes?

A Reduces unnecessary node executions and speeds up response

B Increases the number of nodes executed

C Delays the conditional evaluation

D Adds more CPU overhead

DevTools: Performance

How to check: Record a session while triggering graph routing; look for node execution times and conditional evaluation order.

What to look for: Check if unnecessary nodes run (wasted CPU) and if conditional checks happen before node execution.

Practice

1. What is the main purpose of conditional routing in Langchain graphs?

easy

A. To randomly select a node without any rules

B. To execute all nodes in parallel regardless of conditions

C. To stop the graph execution immediately

D. To choose the next node based on specific conditions

Solution

1. Step 1: Understand conditional routing concept

   Conditional routing means selecting the next step based on rules or conditions.

2. Step 2: Match purpose with options

   Only To choose the next node based on specific conditions describes choosing the next node based on conditions, which fits the concept.

3. Final Answer:

   To choose the next node based on specific conditions -> Option D

4. Quick Check:

   Conditional routing = choose next node by condition [OK]

Hint: Think: routing means choosing path by rules [OK]

Common Mistakes:

• Confusing routing with parallel execution
• Assuming routing stops the graph
• Thinking routing is random

2. Which of the following is the correct way to define a condition function for routing in Langchain?

easy

A. def condition(context): return context['value'] > 10

B. condition = context => context.value > 10

C. def condition(): return context['value'] > 10

D. condition(context): return context.value > 10

Solution

1. Step 1: Check function syntax in Python

   Python functions require 'def' keyword, a parameter list, and a return statement.

2. Step 2: Validate each option

   def condition(context): return context['value'] > 10 correctly defines a function with one parameter and returns a boolean. condition = context => context.value > 10 uses JavaScript syntax. def condition(): return context['value'] > 10 misses the parameter. condition(context): return context.value > 10 misses 'def' keyword.

3. Final Answer:

   def condition(context): return context['value'] > 10 -> Option A

4. Quick Check:

   Python function with parameter and return = def condition(context): return context['value'] > 10 [OK]

Hint: Remember Python function syntax: def name(params): return value [OK]

Common Mistakes:

• Using JavaScript arrow function syntax in Python
• Omitting function parameters
• Missing 'def' keyword

3. Given this routing setup in Langchain graph:

conditions = [
  lambda ctx: ctx['score'] > 80,
  lambda ctx: ctx['score'] > 50
]
routes = ['high', 'medium', 'low']
context = {'score': 65}

Which route will be chosen?

medium

A. "high"

B. "medium"

C. "low"

D. Error due to missing condition

Solution

1. Step 1: Evaluate conditions in order with context

   First condition: score > 80? 65 > 80 is False. Second condition: score > 50? 65 > 50 is True.

2. Step 2: Match true condition to route

   Second condition matches, so route at index 1 is chosen, which is "medium".

3. Final Answer:

   "medium" -> Option B

4. Quick Check:

   First true condition index = route chosen [OK]

Hint: Check conditions top to bottom, pick first true route [OK]

Common Mistakes:

• Choosing 'high' because 65 > 50 but ignoring order
• Picking 'low' when conditions match
• Assuming error if not all conditions true

4. Identify the error in this Langchain routing code snippet:

def route_condition(context):
  if context['value'] > 10:
    return True
  elif context['value'] < 5:
    return False

routes = ['path1', 'path2']
# Routing uses route_condition

medium

A. Routes list should have three paths

B. The function uses wrong comparison operators

C. The function does not return a value for all cases

D. The function should return strings, not booleans

Solution

1. Step 1: Check function return paths

   The function returns True if value > 10, False if value < 5, but returns nothing if value is between 5 and 10.

2. Step 2: Understand routing condition requirements

   Routing conditions must return a boolean for every input to decide path. Missing return causes errors or unexpected behavior.

3. Final Answer:

   The function does not return a value for all cases -> Option C

4. Quick Check:

   All code paths must return a boolean [OK]

Hint: Ensure all if/else paths return a value [OK]

Common Mistakes:

• Ignoring missing return in some cases
• Thinking routes count must match conditions exactly
• Returning wrong data types

5. You want to route a Langchain graph node based on user input where:
- If input contains "urgent", go to 'priority' node
- If input length > 100, go to 'long' node
- Otherwise, go to 'normal' node

Which conditional routing setup correctly implements this logic?

hard

A. conditions = [lambda ctx: 'urgent' in ctx['input'], lambda ctx: len(ctx['input']) > 100]
routes = ['priority', 'long', 'normal']

B. conditions = [lambda ctx: len(ctx['input']) > 100, lambda ctx: 'urgent' in ctx['input']]
routes = ['long', 'priority', 'normal']

C. conditions = [lambda ctx: 'urgent' in ctx['input'] and len(ctx['input']) > 100]
routes = ['priority', 'long', 'normal']

D. conditions = [lambda ctx: 'urgent' not in ctx['input'], lambda ctx: len(ctx['input']) <= 100]
routes = ['normal', 'long', 'priority']

Solution

1. Step 1: Match conditions to requirements in order

   First condition checks if 'urgent' is in input, matching priority route. Second checks input length > 100 for long route.

2. Step 2: Confirm routes order matches conditions plus default

   Routes list has 'priority', 'long', then 'normal' as default if no condition matches.

3. Final Answer:

   conditions = [lambda ctx: 'urgent' in ctx['input'], lambda ctx: len(ctx['input']) > 100] routes = ['priority', 'long', 'normal'] -> Option A

4. Quick Check:

   Order and logic match requirements [OK]

Hint: Order conditions by priority, add default route last [OK]

Common Mistakes:

• Swapping condition order and routes
• Combining conditions incorrectly
• Using negated conditions that break logic