Can someone provide a complete list of CheckpointSavers, including community.

[managedcoder]

Looking at the currently available checkpointers at https://langchain-ai.github.io/langgraph/reference/checkpoints/ shows that only Sqlite and Postgres CheckpointSavers are currently available.

Is there a list of community contributed checkpointers? I am looking for an Azure Storage CheckpointSaver (Blob or Table) and I saw the recent discussion thread about an Azure Table CheckpointSaver but the author had a question 3 weeks ago that went unanswered so I can't use what he's got going on that thread.

Is there detailed doc on implementing a CheckpointSaver? I'd consider writing one myself but based on the question in the above discussion post and given the limited doc here: https://langchain-ai.github.io/langgraph/reference/checkpoints/ I don't feel like I'd have confidence I'd have covered all the bases of being a checkpointer. That doc seems to be more about consumption of existing checkpointers, (i.e., how to use them) not the finer points on how to build one yourself on some, as of yet, unsupported persistence layer.

I did find this interesting doc (https://pypi.org/project/langgraph-checkpoint/) on Checkpoints that is outside of the LangGraph doc that discusses key concepts, including light discussion of pending checkpoint writes (what they are and when they come into play, but not how to handle them if you should see those). It also discusses the why's and what's for of Serde. Having access to doc that specifically discuss how to implement a full-fledged Checkpointer would give me more confidence that the Checkpointer I might create is covering all the expected behavior it might see when being push around in the wild by LangGraph. Trying to decern that from looking at the code for the Postgres implementation is a heavy lift. Specific doc would be much better.

[vbarda]

@managedcoder thanks for the questions! we do not currently have an endorsed list of community-contributed checkpointer libraries beyond the official Sqlite and Postgres libraries ones you already mentioned. we also provide reference implementations for Redis and MongoDB.

for more background information on checkpointing you can also check out this conceptual guide: https://langchain-ai.github.io/langgraph/concepts/persistence/

lastly, we do plan to add a doc on how to implement your own checkpointer, as well as a test suite that you can use to validate that your checkpointer implementation is working correctly. stay tuned!

[managedcoder]

Very helpful, thank you! Would it be possible for someone to take a look at and respond to the post about an Azure Table CheckpointSaver here: #1673 That might unblock us since it looks like he is close to having something that would work.

[phpmac]

@managedcoder thanks for the questions! we do not currently have an endorsed list of community-contributed checkpointer libraries beyond the official Sqlite and Postgres libraries ones you already mentioned. we also provide reference implementations for Redis and MongoDB.

for more background information on checkpointing you can also check out this conceptual guide: https://langchain-ai.github.io/langgraph/concepts/persistence/

lastly, we do plan to add a doc on how to implement your own checkpointer, as well as a test suite that you can use to validate that your checkpointer implementation is working correctly. stay tuned!

Are you the technical director of the company? I see that you often reply. I also have a question. I want to ask you how to turn off this debugging information. I swear I have looked for a lot of information, but I can't see how to turn it off. It is turned on by default. It is not written in py, but in TS.

options {
  tags: [],
  metadata: {
    langgraph_step: 5,
    langgraph_node: 'agent',
    langgraph_triggers: [ 'tools' ],
    langgraph_path: [ '__pregel_pull', 'agent' ],
    langgraph_checkpoint_ns: 'math_expert:dee19283-f4c1-594a-adc0-03780de8ba69|agent:f14f5aec-0778-57be-90e3-da67b2cdc3a6',
    __pregel_resuming: false,
    __pregel_task_id: 'dee19283-f4c1-594a-adc0-03780de8ba69',
    checkpoint_ns: 'math_expert:dee19283-f4c1-594a-adc0-03780de8ba69'
  },
  recursionLimit: 25,
  runId: undefined,
  configurable: {
    __pregel_resuming: false,
    __pregel_task_id: 'f14f5aec-0778-57be-90e3-da67b2cdc3a6',
    __pregel_send: [Function: bound writer],
    __pregel_read: [Function: __pregel_read],
    __pregel_checkpointer: undefined,
    checkpoint_map: {
      '': '1f005b9f-afe6-64f0-fffe-67a43b221a4e',
      'math_expert:dee19283-f4c1-594a-adc0-03780de8ba69': '1f005b9f-bee1-6c20-fffe-8e0f4c1965de'
    },
    __pregel_scratchpad: {
      callCounter: 0,
      interruptCounter: -1,
      resume: [],
      nullResume: undefined,
      subgraphCounter: 0,
      currentTaskInput: [Object],
      consumeNullResume: [Function: consumeNullResume]
    },
    __pregel_previous: undefined,
    checkpoint_id: undefined,
    checkpoint_ns: 'math_expert:dee19283-f4c1-594a-adc0-03780de8ba69|agent:f14f5aec-0778-57be-90e3-da67b2cdc3a6',
    __pregel_call: [Function: bound call]
  },
  runName: 'agent',
  tools: [
    { type: 'function', function: [Object] },
    { type: 'function', function: [Object] }
  ]
}
invoke result {
  generations: [ [ [Object] ] ],
  llmOutput: {
    tokenUsage: { completionTokens: 77, promptTokens: 469, totalTokens: 546 }
  }
}

make langgraph great again

[lfnovo]

Here is my SurrealDB checkpointer if anybody is looking for something like this.

[Prudhvi0717]

@vbarda

lastly, we do plan to add a doc on how to implement your own checkpointer, as well as a test suite that you can use to validate that your checkpointer implementation is working correctly. stay tuned!

Is this available yet?