LangGraph SQLite Checkpoint
Implementation of LangGraph CheckpointSaver that uses SQLite DB (both sync and async, via aiosqlite
)
Usage
from langgraph.checkpoint.sqlite import SqliteSaver
write_config = {"configurable": {"thread_id": "1", "checkpoint_ns": ""}}
read_config = {"configurable": {"thread_id": "1"}}
with SqliteSaver.from_conn_string(":memory:") as checkpointer:
checkpoint = {
"v": 1,
"ts": "2024-07-31T20:14:19.804150+00:00",
"id": "1ef4f797-8335-6428-8001-8a1503f9b875",
"channel_values": {
"my_key": "meow",
"node": "node"
},
"channel_versions": {
"__start__": 2,
"my_key": 3,
"start:node": 3,
"node": 3
},
"versions_seen": {
"__input__": {},
"__start__": {
"__start__": 1
},
"node": {
"start:node": 2
}
},
"pending_sends": [],
}
checkpointer.put(write_config, checkpoint, {}, {})
checkpointer.get(read_config)
list(checkpointer.list(read_config))
Async
from langgraph.checkpoint.sqlite.aio import AsyncSqliteSaver
async with AsyncSqliteSaver.from_conn_string(":memory:") as checkpointer:
checkpoint = {
"v": 1,
"ts": "2024-07-31T20:14:19.804150+00:00",
"id": "1ef4f797-8335-6428-8001-8a1503f9b875",
"channel_values": {
"my_key": "meow",
"node": "node"
},
"channel_versions": {
"__start__": 2,
"my_key": 3,
"start:node": 3,
"node": 3
},
"versions_seen": {
"__input__": {},
"__start__": {
"__start__": 1
},
"node": {
"start:node": 2
}
},
"pending_sends": [],
}
await checkpointer.aput(write_config, checkpoint, {}, {})
await checkpointer.aget(read_config)
[c async for c in checkpointer.alist(read_config)]