[BUG] S3SessionManager with empty prefix produces leading slash in S3 keys causing session restore failure on MinIO and S3-compatible backends

[AadarshBhalerao]

Checks

• I have updated to the lastest minor and patch version of Strands
• I have checked the documentation and this is not expected behavior
• I have searched ./issues and there are no duplicates of my issue

Strands Version

1.22.0, 1.30.0

Python Version

3.12.3

Operating System

Linux (WSL2, Ubuntu)

Installation Method

pip

Steps to Reproduce

1. Create an S3SessionManager with default prefix="" (no prefix argument)
2. Create an Agent with the session manager and send a message
3. Stop the process
4. Create a new S3SessionManager with the same session_id
5. Create a new Agent with the new session manager
6. Observe that agent.messages is empty — session was not restored

Expected Behavior

When an Agent is recreated with the same session_id, all previous messages
are restored from S3 and agent.messages contains the full conversation history.

Actual Behavior

agent.messages is always empty on restore. The agent starts fresh every time,
overwriting existing messages from index 0.

Additional Context

Root cause: get_session_path with empty prefix (default) produces keys with
a leading slash e.g. /session/agents/agent_default/agent.json

MinIO and some S3-compatible backends strip the leading slash on write,
storing the key as session_/... but the read uses /session_/...
which never matches. read_agent() returns None, initialize() takes the
new-session branch and starts from scratch.

Verified by writing a key with leading slash to MinIO and listing with
both /prefix and prefix, only the one without leading slash finds the object.

Possible Solution

def _get_session_path(self, session_id: str) -> str:
    """
	Get session S3 prefix.

    Args:
        session_id: ID for the session.

    Raises:
        ValueError: If session id contains a path separator.
    """
    session_id = _identifier.validate(session_id, _identifier.Identifier.SESSION)

    if self.prefix:
		return f"{self.prefix}/{SESSION_PREFIX}{session_id}/"
	return f"{SESSION_PREFIX}{session_id}/"

Related Issues

No response

[giulio-leone]

Submitted a fix in #1901.

Root cause: _get_session_path always uses f"{self.prefix}/{SESSION_PREFIX}{session_id}/", producing /session_<id>/... when prefix is empty. MinIO and S3-compatible backends strip the leading slash on write but not on read, so read_agent() returns None and sessions are never restored.

Fix: Check if prefix is non-empty before prepending the slash separator. Added 3 regression tests including a full create/read roundtrip with empty prefix.

[AadarshBhalerao]

Hi maintainers (@cagataycali @pgrayy),

Just adding context here since a few PRs have been opened for this issue.

I originally opened this bug report (#1863) after reproducing the problem with S3 session restoration when prefix="". While raising the issue I also included the root cause analysis and a proposed fix in the report.

Since the investigation and proposed solution were documented in the original issue, I have also opened a PR implementing the fix. If possible, I would appreciate it being considered for merge since the analysis and debugging for the issue were part of the original report.

Thank you for reviewing!

[gvelesandro]

@AadarshBhalerao I'm researching failures where agent session persistence looks configured correctly but the restore path still comes back empty. Your report is useful because the same session id should have recovered full history from S3/MinIO, yet a path mismatch made the agent start from scratch and overwrite prior state.

If you're open to it, I'd value one short written postmortem here: https://www.agentsneedcontext.com/agent-failure-postmortem

I'm looking for five specifics: what the agent was trying to do, what session or message history went missing, where that context should have lived, what workaround you used, and what it cost. No pitch, just research.