Are LangGraph and LangChain compatible with serverless architectures? Are there any limitations or potential issues that don’t exist in traditional server environments?
Some additional considerations:
Do they rely on long-lived processes or persistent memory between invocations?
Are there cold start issues or performance penalties with certain serverless providers?
How well do they handle concurrent executions or state management in a stateless environment?
Are there specific deployment best practices or architectural patterns recommended for serverless setups?
I can share my experience. I tried with AWS Lambda and I can confirm that the default in memory checkpoints won’t work across multiple invocations. Even if we mitigate cold start issues using provisioned throughput, we can’t rely on a warm lambda function to persists states.
I think, an external checkpointer like PostgresSaver might help in this case. Would love to hear some thoughts!
Yeah we are also adding external checkpointer postgress for maintaining short term/thread memory. In general I think in memory is only for dev purposes