需求:continue a conversation, resume from a interrupt, recover from a failure, remember information across conversations LangGraph提供了两个互补的持久化系统:

Checkpointer将线程的graph state持久化为checkpoints 适用场景:short-term memory,thread -scoped memory

Stores:将程序规定的数据持久化到graph state之外适用场景:long-term memory,across-thread memory

Quickstart

编译你的 graph with a checkpointer ,a store ,or both

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
from langgraph.checkpoint.memory import InMemorySaver
from langgraph.store.memory import InMemoryStore

checkpointer = InMemorySaver()
store = InMemoryStore()

graph = builder.compile(checkpointer=checkpointer, store=store)

result = graph.invoke(
    {"messages": [{"role": "user", "content": "Hi, my name is Bob."}]},
    {"configurable": {"thread_id": "thread-1"}},
)

Checkpointer vs store

	Checkpointer	Store
Persists	Graph state snapshots	Application-defined key-value data
Scope	A single thread	Across threads
Memory type	Short-term, thread-scoped memory	Long-term, cross-thread memory
Use for	Conversation continuity, human-in-the-loop, time travel, and fault tolerance	User preferences, facts, and shared knowledge
Access pattern	Pass a `thread_id` in graph config	Read and write items from nodes or application code
Full guide	Checkpointers	Stores

Troubleshooting common issues

PostgresSaver:`thread_id`too long

使用(Async)PostgresSaver,thread_id存储在一个长度受限的列中,如果长度超出限制,则会报错

解决办法:将thread_id保持在255字符之内

MemorySaver does not persist between restarts

MemorySaver和InMemorySaver 把checkpoints保存在RAM中,重启时,所有的checkpoints会丢失

解决方法:生产环境中使用可持久化的checkpointer

Checkpoints growing unboundedly

checkpoints无休止增长可能导致延迟和存储的开销

解决方法:定期清理旧的checkpoints或者设置保留策略

State access from parent graph to subgraph

子图的state更新时,父图不会立马看到这些更新:因为子图管理自己的checkpoint namespace

解决方法:需要跨图边界的数据,使用Store共享状态,或者配置你的子图谢图父checkpoints

Quickstart#

Checkpointer vs store#

Troubleshooting common issues#

PostgresSaver:thread_idtoo long#

MemorySaver does not persist between restarts#

Checkpoints growing unboundedly#

State access from parent graph to subgraph#