feat(memory-tools): accept client supplied memory ids in CreateMemoryTool

[nkanu17]

Closes #24

CreateMemoryTool.run_async now accepts an optional application-level id. On the managed Redis Agent Memory backend, the ID is combined with the resolved namespace and user to derive a collision-resistant, managed-safe record ID. A retry with the same application ID in the same scope upserts instead of creating a duplicate, while equal IDs in different user or namespace scopes remain isolated. When absent, the existing stable_memory_id fallback is unchanged.

Design decisions:

• The parameter is deliberately not added to the LLM tool declaration; models must not invent memory IDs. Direct callers pass it programmatically for idempotent retries.
• The self-hosted opensource-agent-memory backend cannot honor client IDs because add_memory_tool generates its own ID. The tool logs the limitation without exposing the supplied ID and proceeds with a server-generated ID.
• The branch is rebased onto the latest main, including invocation-user resolution through tool_context.

Tests cover scoped ID derivation, raw-ID collision resistance, namespace/user isolation including tool_context users, numeric zero, stable fallback behavior, same-ID idempotency, and redacted self-hosted warnings.

make check passes fully: format, lint, mypy, 142 passed, 9 skipped.

[Copilot]

Pull request overview

Adds support for application-supplied memory IDs to CreateMemoryTool.run_async to enable idempotent retries for managed Redis Agent Memory writes, while explicitly not exposing the parameter to the LLM tool declaration. For the self-hosted backend, the tool warns that client IDs are not supported and proceeds with the server-generated ID.

Changes:

• Accept optional id in CreateMemoryTool.run_async, sanitize it for the managed backend, and use it as the record ID to support upserts on retries.
• Log a warning when id is provided for the self-hosted opensource-agent-memory backend (which cannot honor client IDs).
• Add tests for client ID passthrough, sanitization, stable-id fallback, idempotency behavior, and the self-hosted warning path; document the behavior in the memory concepts docs.

Reviewed changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 1 comment.

File	Description
src/adk_redis/tools/memory/create.py	Adds optional client-supplied id handling, managed-ID sanitization, and self-hosted warning behavior.
tests/tools/test_memory_tools.py	Adds coverage for client ID behavior (managed) and warning behavior (self-hosted).
docs/concepts/memory.md	Documents the optional application-level id and the differing backend behaviors.

Suppressed comments (1)

src/adk_redis/tools/memory/create.py:205

• client_memory_id is checked using truthiness (if client_memory_id:), which treats values like 0 as absent and falls back to stable_memory_id. If callers derive IDs from numeric task IDs, this breaks the idempotency contract. Use an explicit None check so any provided ID value is honored (or rejected by sanitization).

      if client_memory_id:
        memory_id = sanitize_managed_identifier(str(client_memory_id))
      else:

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 3d20ca6e20

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[Copilot]

Pull request overview

Copilot reviewed 3 out of 3 changed files in this pull request and generated no new comments.

Suppressed comments (1)

src/adk_redis/tools/memory/create.py:177

• The warning for the self-hosted backend logs the raw client-supplied memory id. Since this is an application-level id (often derived from queue/task identifiers), logging it at WARNING can unintentionally leak sensitive identifiers into logs. Consider logging the limitation without echoing the provided id (or log a redacted/hash form).

          logger.warning(
              "Client-supplied memory id %r is not supported by the "
              "opensource-agent-memory backend: add_memory_tool generates "
              "its own memory ID. Proceeding without the client id.",
              client_memory_id,

[Copilot]

Pull request overview

Copilot reviewed 3 out of 3 changed files in this pull request and generated no new comments.

[Copilot]

Pull request overview

Copilot reviewed 3 out of 3 changed files in this pull request and generated no new comments.