[eric7h]

Problem

InMemoryLTMBackend (LongTermMemory(backend="local")) ignores user_id. save_memory writes every user's events into one shared VectorStoreIndex, and search_memory searches that shared index, so one user's memories are returned to another user. The other backends (Redis, OpenSearch, ...) build a separate index per veadk-ltm/{index}/{user_id}.

Reproduction

backend.save_memory(user_id="alice", event_strings=["alice likes tea"])
backend.save_memory(user_id="bob", event_strings=["bob likes coffee"])
backend.search_memory(user_id="carol", query="likes", top_k=5)
# main: ['alice likes tea']   expected: []

Fix

Keep one in-memory VectorStoreIndex per user_id. Searching for a user with no saved memories returns [].

Verification

• New tests/memory/long_term_memory_backends/test_in_memory_ltm_backend.py (uses llama_index MockEmbedding, no network): both tests fail on main and pass with the fix.
• uv run --extra dev pre-commit run --all-files: pass.
• uv run --extra dev pytest -n 4 -m "not codex_smoke and not piagent_smoke": 8054 passed. The same 24 failures and 2 collection errors (Codex runtime / self-host sandbox tests that need local tooling) occur on unmodified main in this environment.

Impact

No public API change. Only the local backend's behavior changes: memories are now scoped per user, like the other backends.

This PR was prepared with an AI coding agent; I authorized the submission.