bug: incorrect pagination in Memory Session Manager `list_messages`

Open
#346 1 comment 0 reactions 0 assignees View on GitHub

Maintainers usually reply within 1 day

Nobody has claimed this yet.

Assessment

Difficulty
3/5
Estimated time
1-2 days
Newbie friendliness
58/100
Issue type
Bug
Clarity
Mostly clear
Activity status
Quiet
Tech stack
python
Domain
api, backend

Research direction

Start in session_manager.py at list_messages and inspect how memory_client.list_events applies max_results and how events_to_messages discards state events. Reproduce the case with shared actor IDs and session or agent state events, then add unit or integration coverage; done means pagination returns the requested messages or exposes a nextToken when more messages exist.

Written by the indexing model from the issue text.

Description

bug

Problem

Before PR #244

State events and conversational events were naturally separated by actorId:

# create_session used a prefixed actorId
create_event(actorId="session_my-session-id", payload=[{"blob": ...}])

# create_agent used a prefixed actorId
create_event(actorId="agent_my-agent-id", payload=[{"blob": ...}])

# create_message used the regular actorId
create_event(actorId="my-actor-id", payload=[{"conversational": ...}])

So when list_messages called list_events with actorId="my-actor-id", it only got conversational events back. State events lived under different actorIds and were invisible to this query. max_results counted only
what mattered.

After PR #244

Everything now uses the same actorId, with metadata distinguishing state events:

# create_session — same actorId, metadata marks it as state
create_event(
    actorId="my-actor-id",  # ← same as messages now
    metadata={"stateType": {"stringValue": "SESSION"}},
    payload=[{"blob": ...}],
)


# create_agent — same actorId, metadata marks it as state
create_event(
    actorId="my-actor-id",  # ← same as messages now
    metadata={"stateType": {"stringValue": "AGENT"}, "agentId": {"stringValue": "..."}},
    payload=[{"blob": ...}],
)

# create_message — same actorId, no metadata
create_event(actorId="my-actor-id", payload=[{"conversational": ...}])
Potential Issues

list_messages in session_manager.py fetches events without any metadata filter:

events = self.memory_client.list_events(
    memory_id=self.config.memory_id,
    actor_id=self.config.actor_id,
    session_id=session_id,
    max_results=max_results,  # ← this now includes state events in the count
)

Since all events share the same actorId now, this query returns both conversational AND state events. The API returned max_results events total, some of which were state events. Those get discarded by the converter, so the final list is shorter than what the caller asked for.

Concrete example

A session with 1 session state event, 2 agent state events, and 20 conversational messages. Customer calls list_messages(limit=10):

  1. max_results = limit + offset = 10
  2. list_events returns 10 events: 3 state + 7 conversational
  3. events_to_messages discards the 3 state events → 7 messages
  4. Customer gets 7 messages, expected 10

Customer does not get a pagination token, so they assume 7 is all that exists, and doesn't know there are more.

Acceptance Criteria
  • customer should be given nextToken, or should always get maxResults exactly if more results exist.
  • unit tests and/or integration verifying fix of the bug.
Dominant language
Python
Stars
786
Forks
163
Avg merge
4h
Merged PRs (30d)
22

Getting set up

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

More from aws/bedrock-agentcore-sdk-python

All issues in aws/bedrock-agentcore-sdk-python

Similar issues

More Python issues

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.