feat(strands-memory): add converter injection and optional restored-tool filtering by dharamendrak · Pull Request #288 · aws/bedrock-agentcore-sdk-python

dharamendrak · 2026-03-01T06:35:28Z

Issue #, if available:
#254

Description of changes:

Added injectable message converter support in AgentCoreMemorySessionManager to remove hardcoded converter dependency and enable custom STM canonical formats.
Introduced MemoryConverter protocol with:
- message_to_payload(...)
- events_to_messages(...)
- exceeds_conversational_limit(...)
Updated session manager constructor to accept optional converter and default to BedrockConverseConverter for backward compatibility.
Replaced hardcoded converter calls in write/read paths with self.converter.
Added converter modules under integrations/strands/converters/:
- bedrock.py (default Strands-native shape adapter)
- openai.py (OpenAI canonical format converter)
- protocol.py (shared interface and size-limit utility)
Added optional config flag filter_restored_tool_context (default False) to strip historical toolUse/toolResult content blocks during restore when enabled.
Added tests for:
- OpenAI converter behavior
- session manager converter injection and restore filtering
Updated comments/docstrings to clarify wording as “Strands-native message shape”.

By submitting this pull request, I confirm that you can use, modify, copy, and redistribute this contribution, under the terms of your choice.

…ool filtering

…converters and tests

… and enhance OpenAI converter handling - Added AutoConverseConverter to facilitate automatic converter selection based on agent model. - Updated AgentCoreMemorySessionManager to support auto converter mode. - Enhanced OpenAI converter to preserve reasoning content and handle oversized payloads. - Updated tests to validate new auto converter functionality and reasoning content preservation.

- Changed module paths for OpenAI, Anthropic, and Gemini models in test files to align with the updated project structure. - Ensured consistency across tests for accurate model identification.

afarntrog · 2026-03-04T20:26:32Z

src/bedrock_agentcore/memory/integrations/strands/converters/auto.py

@@ -0,0 +1,85 @@
+"""Automatic converter selection for mixed-model sessions."""


I'm not convinced this is necessary. Although it may be possible that someone will have mixed messages, I see it as a really rare use case so I'm not convinced it's worth the tech debt.

I agree this is case is rare

afarntrog · 2026-03-04T21:01:47Z

src/bedrock_agentcore/memory/integrations/strands/converters/bedrock.py

@@ -0,0 +1,7 @@
+"""Strands-native message-shape converter adapter."""


I think this should not be necessary as strands is natively bedrock. We can just leave the converter as None

afarntrog · 2026-03-04T21:02:42Z

src/bedrock_agentcore/memory/integrations/strands/converters/anthropic.py

@@ -0,0 +1,200 @@
+"""Anthropic-format converter for AgentCore Memory.


I would be in favor of progressively adding these converters only as we need them. I think we can leave this one out for now.

We can incrementally add it

afarntrog · 2026-03-04T21:06:44Z

src/bedrock_agentcore/memory/integrations/strands/session_manager.py

+            self.converter = AutoConverseConverter
+            AutoConverseConverter.set_write_converter(BedrockConverseConverter)
+        else:
+            self.converter = converter or BedrockConverseConverter


nit: consider just leaving it as None for bedrock (native) format

@afarntrog agree

afarntrog · 2026-03-04T21:07:38Z

src/bedrock_agentcore/memory/integrations/strands/session_manager.py

               Defaults to None.
            **kwargs (Any): Additional keyword arguments.
        """
+        self._auto_converter_enabled = converter == "auto"


I think this _auto_converter_enabled is over engineering for now. consider yagni

I agree, thanks @afarntrog

feat(strands-memory): add converter injection and optional restored-t…

b3ffc3c

…ool filtering

dharamendrak requested a review from a team March 1, 2026 06:35

dharamendrak requested a deployment to manual-approval March 1, 2026 06:35 — with GitHub Actions Waiting

fix: update conversational payload size limit to 100000 chars across …

98f6241

…converters and tests

dharamendrak requested a deployment to manual-approval March 1, 2026 06:45 — with GitHub Actions Waiting

dharamendrak requested a deployment to manual-approval March 2, 2026 00:01 — with GitHub Actions Waiting

fix: update model module paths in tests to reflect new structure

2223eb0

- Changed module paths for OpenAI, Anthropic, and Gemini models in test files to align with the updated project structure. - Ensured consistency across tests for accurate model identification.

dharamendrak requested a deployment to manual-approval March 2, 2026 00:53 — with GitHub Actions Waiting

afarntrog reviewed Mar 4, 2026

View reviewed changes

Remove nit and kept only openai converter

bbbc7c0

dharamendrak requested a deployment to manual-approval March 5, 2026 02:19 — with GitHub Actions Waiting

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

feat(strands-memory): add converter injection and optional restored-tool filtering#288

feat(strands-memory): add converter injection and optional restored-tool filtering#288
dharamendrak wants to merge 5 commits intoaws:mainfrom
dharamendrak:feat/strands-converter-injection

dharamendrak commented Mar 1, 2026

Uh oh!

afarntrog Mar 4, 2026

Uh oh!

dharamendrak Mar 5, 2026

Uh oh!

afarntrog Mar 4, 2026

Uh oh!

dharamendrak Mar 5, 2026

Uh oh!

afarntrog Mar 4, 2026

Uh oh!

dharamendrak Mar 5, 2026

Uh oh!

afarntrog Mar 4, 2026

Uh oh!

dharamendrak Mar 5, 2026

Uh oh!

afarntrog Mar 4, 2026

Uh oh!

dharamendrak Mar 4, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

		@@ -0,0 +1,85 @@
		"""Automatic converter selection for mixed-model sessions."""

		@@ -0,0 +1,7 @@
		"""Strands-native message-shape converter adapter."""

		@@ -0,0 +1,200 @@
		"""Anthropic-format converter for AgentCore Memory.

Conversation

dharamendrak commented Mar 1, 2026

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!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants