Langgraph Architecture

LangGraph Architecture Decisions

When to Use LangGraph

Use LangGraph When You Need:

• Stateful conversations - Multi-turn interactions with memory
• Human-in-the-loop - Approval gates, corrections, interventions
• Complex control flow - Loops, branches, conditional routing
• Multi-agent coordination - Multiple LLMs working together
• Persistence - Resume from checkpoints, time travel debugging
• Streaming - Real-time token streaming, progress updates
• Reliability - Retries, error recovery, durability guarantees

Consider Alternatives When:

State Schema Decisions

TypedDict vs Pydantic

Recommendation: Use TypedDict for most cases. Use Pydantic when you need validation or complex nested structures.

Reducer Selection

State Size Considerations

# SMALL STATE (< 1MB) - Put in state
class State(TypedDict):
    messages: Annotated[list, add_messages]
    context: str

# LARGE DATA - Use Store
class State(TypedDict):
    messages: Annotated[list, add_messages]
    document_ref: str  # Reference to store

def node(state, *, store: BaseStore):
    doc = store.get(namespace, state["document_ref"])
    # Process without bloating checkpoints

Graph Structure Decisions

Single Graph vs Subgraphs

Single Graph when:

• All nodes share the same state schema
• Simple linear or branching flow
• < 10 nodes

Subgraphs when:

• Different state schemas needed
• Reusable components across graphs
• Team separation of concerns
• Complex hierarchical workflows

Conditional Edges vs Command

# Conditional Edge - when routing is the focus
def router(state) -> Literal["a", "b"]:
    return "a" if condition else "b"
builder.add_conditional_edges("node", router)

# Command - when combining routing with updates
def node(state) -> Command:
    return Command(goto="next", update={"step": state["step"] + 1})

Static vs Dynamic Routing

Static Edges (add_edge):

• Fixed flow known at build time
• Clearer graph visualization
• Easier to reason about

Dynamic Routing (add_conditional_edges, Command, Send):

• Runtime decisions based on state
• Agent-driven navigation
• Fan-out patterns

Persistence Strategy

Checkpointer Selection

Checkpointing Scope

# Full persistence (default)
graph = builder.compile(checkpointer=checkpointer)

# Subgraph options
subgraph = sub_builder.compile(
    checkpointer=None,   # Inherit from parent
    checkpointer=True,   # Independent checkpointing
    checkpointer=False,  # No checkpointing (runs atomically)
)

When to Disable Checkpointing

• Short-lived subgraphs that should be atomic
• Subgraphs with incompatible state schemas
• Performance-critical paths without need for resume

Multi-Agent Architecture

Supervisor Pattern

Best for:

• Clear hierarchy
• Centralized decision making
• Different agent specializations

          ┌─────────────┐
          │  Supervisor │
          └──────┬──────┘
    ┌────────┬───┴───┬────────┐
    ▼        ▼       ▼        ▼
┌──────┐ ┌──────┐ ┌──────┐ ┌──────┐
│Agent1│ │Agent2│ │Agent3│ │Agent4│
└──────┘ └──────┘ └──────┘ └──────┘

Peer-to-Peer Pattern

Best for:

• Collaborative agents
• No clear hierarchy
• Flexible communication

┌──────┐     ┌──────┐
│Agent1│◄───►│Agent2│
└──┬───┘     └───┬──┘
   │             │
   ▼             ▼
┌──────┐     ┌──────┐
│Agent3│◄───►│Agent4│
└──────┘     └──────┘

Handoff Pattern

Best for:

• Sequential specialization
• Clear stage transitions
• Different capabilities per stage

┌────────┐    ┌────────┐    ┌────────┐
│Research│───►│Planning│───►│Execute │
└────────┘    └────────┘    └────────┘

Streaming Strategy

Stream Mode Selection

Subgraph Streaming

# Stream from subgraphs
async for chunk in graph.astream(
    input,
    stream_mode="updates",
    subgraphs=True  # Include subgraph events
):
    namespace, data = chunk  # namespace indicates depth

Human-in-the-Loop Design

Interrupt Placement

Resume Patterns

# Simple resume (same thread)
graph.invoke(None, config)

# Resume with value
graph.invoke(Command(resume="approved"), config)

# Resume specific interrupt
graph.invoke(Command(resume={interrupt_id: value}), config)

# Modify state and resume
graph.update_state(config, {"field": "new_value"})
graph.invoke(None, config)

Gates (sequenced)

Complete in order before treating a LangGraph design as locked in. Each step has an objective pass condition (artifact or explicit “none”), not an honor-system “we considered it.”

1. Alternatives — Pass: For the workload, either (a) at least one row from Consider Alternatives When was evaluated and rejected with a one-line reason, or (b) the use case clearly matches Use LangGraph When You Need and does not fit a “consider alternative” row. 2. State contract — Pass: Every state field has an assigned reducer (or default/LastValue) documented in the same place as the schema; large payloads are references or Store-backed, not inlined blobs (see State Size Considerations). 3. Checkpointer — Pass: The saver type is chosen for the target environment per Checkpointer Selection (e.g. production is not InMemorySaver unless explicitly test-only). 4. Loops and flaky nodes — Pass: recursion_limit (or equivalent) is set for any graph that can cycle; per-node RetryPolicy or a documented “no retries” choice exists for external calls (see Retry Configuration).

Error Handling Strategy

Retry Configuration

# Per-node retry
RetryPolicy(
    initial_interval=0.5,
    backoff_factor=2.0,
    max_interval=60.0,
    max_attempts=3,
    retry_on=lambda e: isinstance(e, (APIError, TimeoutError))
)

# Multiple policies (first match wins)
builder.add_node("node", fn, retry_policy=[
    RetryPolicy(retry_on=RateLimitError, max_attempts=5),
    RetryPolicy(retry_on=Exception, max_attempts=2),
])

Fallback Patterns

def node_with_fallback(state):
    try:
        return primary_operation(state)
    except PrimaryError:
        return fallback_operation(state)

# Or use conditional edges for complex fallback routing
def route_on_error(state) -> Literal["retry", "fallback", "__end__"]:
    if state.get("error") and state["attempts"] < 3:
        return "retry"
    elif state.get("error"):
        return "fallback"
    return END

Scaling Considerations

Horizontal Scaling

• Use PostgresSaver for shared state
• Consider LangGraph Platform for managed infrastructure
• Use stores for large data outside checkpoints

Performance Optimization

1. Minimize state size - Use references for large data 2. Parallel nodes - Fan out when possible 3. Cache expensive operations - Use CachePolicy 4. Async everywhere - Use ainvoke, astream

Resource Limits

# Set recursion limit
config = {"recursion_limit": 50}
graph.invoke(input, config)

# Track remaining steps in state
class State(TypedDict):
    remaining_steps: RemainingSteps

def check_budget(state):
    if state["remaining_steps"] < 5:
        return "wrap_up"
    return "continue"

Decision Checklist

After Gates (sequenced), before implementing:

1. [ ] Is LangGraph the right tool? (vs simpler alternatives) 2. [ ] State schema defined with appropriate reducers? 3. [ ] Persistence strategy chosen? (dev vs prod checkpointer) 4. [ ] Streaming needs identified? 5. [ ] Human-in-the-loop points defined? 6. [ ] Error handling and retry strategy? 7. [ ] Multi-agent coordination pattern? (if applicable) 8. [ ] Resource limits configured?