fix(memory): Improve pagination behavior in get_last_k_turns() and list_messages() by kevmyung · Pull Request #209 · aws/bedrock-agentcore-sdk-python

kevmyung · 2026-01-09T00:35:23Z

Summary

get_last_k_turns() in both MemoryClient and MemorySessionManager now automatically paginates until k turns are found when max_results is not specified
list_messages() in AgentCoreMemorySessionManager adds fetch_all parameter to retrieve all messages instead of being limited to 100
Maintains backward compatibility: existing code with explicit max_results parameter behaves the same

Problem

Previously, get_last_k_turns(k=200) would only fetch 100 events (the default max_results) regardless of the k value requested. This caused issues when users needed more than 100 events worth of conversation turns.

Similarly, list_messages(limit=None) was hardcoded to return only 100 messages, making it impossible to retrieve all messages in a session.

Changes

`client.py` & `session.py`

Changed max_results default from 100 to None
When max_results=None, automatically paginates until k turns are found
When max_results is explicitly provided, respects that limit (backward compatible)

`strands/session_manager.py`

Added fetch_all: bool = False parameter to list_messages()
When fetch_all=True, fetches up to 10,000 messages

Tests

Added test_get_last_k_turns_auto_pagination - verifies pagination continues until k turns found
Added test_get_last_k_turns_explicit_max_results - verifies backward compatibility
Added test_list_messages_fetch_all - verifies fetch_all parameter works correctly

Test plan

Unit tests pass (pytest tests/bedrock_agentcore/memory/)
Manual integration test with actual Amazon AgentCore Memory

Closes #206

aidandaly24 · 2026-01-12T14:36:21Z

-            # Process events to group into turns
            turns = []
            current_turn = []
+            next_token = None


suggestion (non-blocking): This pagination logic (lines 805-857) is nearly identical to client.py lines 1108-1157. Consider extracting into a shared helper to reduce duplication and maintenance burden. Future bug fixes would need to be applied to both locations otherwise.

aidandaly24 · 2026-01-12T14:39:32Z

+        """Test get_last_k_turns auto-calculates max_results based on k."""
+        with patch("boto3.Session") as mock_boto_client:
+            mock_client_instance = MagicMock()
+            mock_boto_client.return_value = mock_client_instance


issue: This test mocks manager.list_events, but the new implementation in session.py no longer calls list_events() - it directly calls self._data_plane_client.list_events(). This mock won't intercept the actual calls.

Should mock manager._data_plane_client.list_events instead, similar to how test_client.py mocks client.gmdp_client.list_events.

aidandaly24 · 2026-01-12T14:42:01Z

        try:
-            max_results = (limit + offset) if limit else 100
+            if fetch_all:
+                max_results = 10000


nit: The value 10000 is a magic number. Consider extracting to a named constant (e.g., MAX_FETCH_ALL_RESULTS = 10000) for clarity and maintainability.

aidandaly24 · 2026-01-12T14:43:58Z

@@ -786,52 +786,82 @@
        k: int = 5,
        branch_name: Optional[str] = None,
        include_parent_branches: bool = False,


thought: client.py uses include_branches while session.py uses include_parent_branches. This inconsistency predates this PR but is worth noting for future cleanup.

…ist_messages() - get_last_k_turns(): Auto-calculate max_results based on k (max(100, k*3)) - list_messages(): Add fetch_all parameter to fetch all messages (up to 10000) - Backward compatible: default behavior unchanged

- Extract shared pagination logic into pagination.py helper - Fix test mocks to use _data_plane_client.list_events - Add MAX_FETCH_ALL_RESULTS constant (10000) in strands session_manager - Rename include_branches to include_parent_branches in client.py for consistency - Add comprehensive tests for pagination helper

jariy17 · 2026-01-12T22:40:39Z

+        agent_id: str,
+        limit: Optional[int] = None,
+        offset: int = 0,
+        fetch_all: bool = False,


For fetch_all, we aren't really fetching all issue, just whatever MAX_FETCH_ALL_RESULTS is. Therefore, we should just remove this field.

jariy17 · 2026-01-12T22:46:28Z

            raise SessionException(f"Session ID mismatch: expected {self.config.session_id}, got {session_id}")

        try:
-            max_results = (limit + offset) if limit else 100


Let's keep this line but have MAX_FETCH_ALL_RESULTS which is 10000

jariy17 · 2026-01-12T22:47:36Z

+logger = logging.getLogger(__name__)
+
+
+def paginate_turns(


I don't really seeing anyone using this function besides MemoryClient and Session class. I would either move into either class or both classes.

jariy17 · 2026-01-12T22:57:30Z

        branch_name: Optional[str] = None,
-        include_branches: bool = False,
-        max_results: int = 100,
+        include_parent_branches: bool = False,


If include_branches is already commited in the code, renaming this will introduce a breaking change

- Remove fetch_all parameter from list_messages (misleading name) - Use MAX_FETCH_ALL_RESULTS (10000) as default when no limit specified - Remove pagination.py module, inline logic into client.py and session.py - Revert include_branches rename to avoid breaking change

codecov-commenter · 2026-01-13T14:54:35Z

Codecov Report

❌ Patch coverage is 81.81818% with 12 lines in your changes missing coverage. Please review.
⚠️ Please upload report for BASE (main@0df9757). Learn more about missing BASE report.

Files with missing lines	Patch %	Lines
src/bedrock_agentcore/memory/client.py	77.77%	2 Missing and 4 partials ⚠️
src/bedrock_agentcore/memory/session.py	83.78%	2 Missing and 4 partials ⚠️

Additional details and impacted files

@@           Coverage Diff           @@
##             main     #209   +/-   ##
=======================================
  Coverage        ?   90.50%           
=======================================
  Files           ?       35           
  Lines           ?     3412           
  Branches        ?      507           
=======================================
  Hits            ?     3088           
  Misses          ?      179           
  Partials        ?      145

Flag	Coverage Δ
unittests	`90.50% <81.81%> (?)`

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

Apply automatic formatting fixes identified by ruff format pre-commit hook. Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>

aidandaly24

LGTM

kevmyung temporarily deployed to manual-approval January 9, 2026 00:35 — with GitHub Actions Inactive

aidandaly24 reviewed Jan 12, 2026

View reviewed changes

aidandaly24 temporarily deployed to manual-approval January 12, 2026 21:18 — with GitHub Actions Inactive

aidandaly24 force-pushed the fix/pagination-behavior branch from 798e504 to 6d7f19e Compare January 12, 2026 21:52

aidandaly24 temporarily deployed to manual-approval January 12, 2026 21:52 — with GitHub Actions Inactive

kevmyung and others added 2 commits January 12, 2026 17:05

aidandaly24 force-pushed the fix/pagination-behavior branch from 6d7f19e to 355da39 Compare January 12, 2026 22:08

aidandaly24 temporarily deployed to manual-approval January 12, 2026 22:08 — with GitHub Actions Inactive

jariy17 reviewed Jan 12, 2026

View reviewed changes

aidandaly24 temporarily deployed to manual-approval January 12, 2026 23:36 — with GitHub Actions Inactive

jariy17 previously approved these changes Jan 13, 2026

View reviewed changes

fix: Apply ruff formatting to pagination changes

5315b0b

Apply automatic formatting fixes identified by ruff format pre-commit hook. Co-Authored-By: Claude Sonnet 4.5 <noreply@anthropic.com>

jariy17 dismissed their stale review via 5315b0b January 13, 2026 21:28

jariy17 approved these changes Jan 13, 2026

View reviewed changes

aidandaly24 approved these changes Jan 13, 2026

View reviewed changes

jariy17 temporarily deployed to manual-approval January 13, 2026 21:51 — with GitHub Actions Inactive

aidandaly24 merged commit 2b047ff into aws:main Jan 13, 2026
19 checks passed

ecoonomykim mentioned this pull request Feb 5, 2026

Bug Report: get_last_k_turns() returns messages in reversed order within each turn #253

Open

Navigation Menu

Uh oh!

fix(memory): Improve pagination behavior in get_last_k_turns() and list_messages() - #209

fix(memory): Improve pagination behavior in get_last_k_turns() and list_messages()#209
aidandaly24 merged 4 commits into
aws:mainfrom
kevmyung:fix/pagination-behavior

kevmyung commented Jan 9, 2026

Uh oh!

aidandaly24 Jan 12, 2026

Uh oh!

aidandaly24 Jan 12, 2026

Uh oh!

aidandaly24 Jan 12, 2026

Uh oh!

aidandaly24 Jan 12, 2026

Uh oh!

jariy17 Jan 12, 2026

Uh oh!

jariy17 Jan 12, 2026

Uh oh!

jariy17 Jan 12, 2026

Uh oh!

jariy17 Jan 12, 2026

Uh oh!

codecov-commenter commented Jan 13, 2026 •

edited

Loading

Uh oh!

aidandaly24 left a comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

		logger = logging.getLogger(__name__)


		def paginate_turns(

Uh oh!

Conversation

kevmyung commented Jan 9, 2026

Summary

Problem

Changes

client.py & session.py

strands/session_manager.py

Tests

Test plan

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

codecov-commenter commented Jan 13, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Report

Uh oh!

aidandaly24 left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

`client.py` & `session.py`

`strands/session_manager.py`

codecov-commenter commented Jan 13, 2026 •

edited

Loading