fix(templates): retrieve SUMMARIZATION summaries with an actor-scoped namespace (#665)

[aidandaly24]

Description

#665

Agents created/extended with --strategies SUMMARIZATION write per-session conversation summaries correctly, but the generated session.py retrieves them with a per-session namespace, so a new session never sees prior sessions' summaries. Cross-session summary recall silently does nothing; the feature appears write-only to users.

Fix

Re-apply PR #1299's one-line change in all three Strands session.py templates: change the SUMMARIZATION retrieval line at line 31 from f"/summaries/{actor_id}/{session_id}" to f"/summaries/{actor_id}". Run npm run test:update-snapshots to refresh the 3 affected entries in assets.snapshot.test.ts.snap (lines 2264, 3286, 6554). Update docs/memory.md:95 (manual-wiring snippet) and the strategy table at docs/memory.md:160 to /summaries/{actorId} for SUMMARIZATION. Leave the write-side namespaceTemplate (memory.ts:26) and EPISODIC ({session_id} kept intentionally; the issue is scoped to SUMMARIZATION) unchanged. Add a regression test asserting the generated SUMMARIZATION retrieval namespace contains no {session_id}, so a future squash-merge cannot silently revert it again. No SDK or CDK change needed (SDK prefix-retrieval is correct; CDK owns nothing in this path).

This fix was produced by an automated triage+fix run and validated locally (build + unit suite + symptom reproduction). Opened as a draft for CI and human review; will be bug-bashed before it's marked ready.

Related Issue

Closes #665

Documentation PR

N/A — bug fix; no devguide change required.

Type of Change

• Bug fix
• New feature
• Breaking change
• Documentation update
• Other (please describe):

Testing

How have you tested the change? (full validation in PR comments after bug bash)

• I ran npm run test:unit and npm run test:integ
• I ran npm run typecheck
• I ran npm run lint
• If I modified src/assets/, I ran npm run test:update-snapshots and committed the updated snapshots

Checklist

• I have read the CONTRIBUTING document
• I have added any necessary tests that prove my fix is effective or my feature works
• I have updated the documentation accordingly
• I have added an appropriate example to the documentation to outline the feature, or no new docs are needed
• My changes generate no new warnings
• Any dependent changes have been merged and published

---

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

[agentcore-devx-automation]

Claude Security Review: no high-confidence findings. (run)

[github-actions]

Package Tarball

aws-agentcore-0.21.0.tgz

How to install

gh release download pr-1660-tarball --repo aws/agentcore-cli --pattern "*.tgz" --dir /tmp/pr-tarball
npm install -g /tmp/pr-tarball/aws-agentcore-0.21.0.tgz

[agentcore-cli-automation]

LGTM. The fix correctly mirrors PR #1299: retrieval moves to the actor-scoped prefix /summaries/{actor_id} so the SDK's hierarchical prefix match surfaces summaries from prior sessions, while writes remain session-scoped via DEFAULT_STRATEGY_NAMESPACE_TEMPLATES.SUMMARIZATION (memory.ts:23) — so storage organization is preserved and only the read scope is widened. All three template flavors (http, agui, a2a) are updated consistently, the snapshots regenerate cleanly, and the docs (manual-wiring snippet + strategy table) are kept in sync.

The new regression test is well-scoped: it reads the real template files and renders them through the same global Handlebars instance the production renderer uses (registering the includes helper via the side-effect import of render.ts), with no excessive mocking. The negative assertion against /summaries/{actor_id}/{session_id} is exactly what's needed to prevent another silent revert.

No telemetry needed (template bug fix, not a new feature). The api-types.ts:158 comment example still references /summaries/{actorId}/{sessionId}, but that's the write/namespaceTemplate namespace which is intentionally unchanged, so the comment remains accurate.

[agentcore-devx-automation]

Claude Security Review: no high-confidence findings. (run)

[agentcore-cli-automation]

Reviewed independently. Confirmed:

• All three Strands flavor templates (http, agui, a2a) and the corresponding snapshots are updated consistently to actor-scoped retrieval (/summaries/{actor_id}), while the write/storage namespaceTemplate at src/schema/schemas/primitives/memory.ts:26 remains session-scoped — preserving storage organization and only widening the read prefix, exactly as PR fix: scope summaries retrieval to actor for cross-session memory recall #1299 did.
• The new regression test at src/assets/__tests__/summarization-namespace.test.ts reads the real on-disk templates and renders them through Handlebars (registering the includes helper via the side-effect import of render.ts) with no excessive mocking. The negative assertion against /summaries/{actor_id}/{session_id} directly guards against a future silent revert.
• Docs (docs/memory.md:95 manual-wiring snippet and the strategy table at :160) are kept in sync. The api-types.ts:147 comment example correctly refers to the write-side namespaceTemplates so it remains accurate.
• No telemetry needed — this is a template bug fix, not a new feature.

All the concerns I would have flagged have already been covered in the existing approving review. LGTM to merge.