Memory Hash Synchronization Issue When Switching Machines

[jfrank-summit]

When switching machines for agent execution, the local caching of memory hashes can lead to synchronization issues. The current implementation prioritizes local hash storage:

export const getLastMemoryHash = async (): Promise<string> => {
  const localHash = getLocalHash();
  if (localHash) {
    logger.info('Using locally stored hash', { hash: localHash });
    return localHash;
  }

  const blockchainHash = await contract.getLastMemoryHash(wallet.address);
  if (blockchainHash) {
    saveHashLocally(blockchainHash);
  }
  return blockchainHash;
};

This means when moving to a new machine, we might use outdated local hash data instead of the latest blockchain state from the previous machine.

Proposed Solutions

1. Session-Based Hash Storage

Generate unique session ID on agent startup
Associate last-hash.json with current session
Clear previous session data on startup

2. Timestamp-Based Validation

Add timestamp to memory hash events
Compare local hash timestamp with blockchain on startup
Invalidate local cache if blockchain state is newer

[Xm0onh]

When you are switching machine, you don't have any local memory at the beginning. So your state is going to be the latest state on the blockchain.
However, I think your issue is: sometimes you run on machine A and sometimes on B. In that case, yes, the latest memory in the local file is outdated and will cause synch problem.