Skip to content

Fast-LangGraph

High-performance Rust accelerators for LangGraph applications

CI PyPI License: MIT

Fast-LangGraph provides drop-in components that deliver up to 700x speedups for checkpoint operations and 10-50x speedups for state management in your LangGraph applications.

Why Fast-LangGraph?

LangGraph is excellent for building AI agents, but production workloads often encounter performance bottlenecks:

Problem Impact Fast-LangGraph Solution
Checkpoint serialization Python's deepcopy is slow for complex state Rust-based serialization (up to 737x faster)
State management at scale High-frequency updates accumulate overhead Optimized state operations (13-46x faster)
Repeated LLM calls Identical prompts waste API costs Built-in response caching (10x+ speedup)

Fast-LangGraph solves these by reimplementing critical paths in Rust while maintaining full API compatibility.

Performance at a Glance

Operation Speedup Use Case
Checkpoint Serialization (250KB) 737x Large agent state
Sustained State Updates 46x Long-running graphs
End-to-End Graph Execution 2.8x Production workloads
LLM Response Caching 10x Repeated prompts

Two Ways to Accelerate

1. Automatic Acceleration (Zero Code Changes)

Enable transparent acceleration with a single environment variable:

export FAST_LANGGRAPH_AUTO_PATCH=1
python your_app.py

Or with a single function call:

import fast_langgraph
fast_langgraph.shim.patch_langgraph()

# Your existing LangGraph code runs 2-3x faster automatically

2. Manual Acceleration (Maximum Performance)

For the largest speedups, use Rust components directly:

from fast_langgraph import (
    RustSQLiteCheckpointer,  # 5-6x faster checkpointing
    cached,                   # LLM response caching
    langgraph_state_update,   # Fast state merging
)

Quick Example

from fast_langgraph import RustSQLiteCheckpointer, cached
from langgraph.graph import StateGraph

# Cache LLM responses
@cached(max_size=1000)
def call_llm(prompt):
    return llm.invoke(prompt)

# Use fast checkpointing
checkpointer = RustSQLiteCheckpointer("state.db")
graph = graph.compile(checkpointer=checkpointer)

# Run with automatic state persistence
result = graph.invoke(
    {"messages": [HumanMessage(content="Hello")]},
    config={"configurable": {"thread_id": "user-123"}}
)

Requirements

  • Python 3.9+
  • Works with any LangGraph version

Project Metadata

Next Steps