Refactor memory types, working memory (breaking changes)

[abrookins]

Starting to tease out a set of more useful abstractions now that we're using this with an agent ourselves.

[Copilot]

Pull Request Overview

This PR refactors the memory subsystem by replacing the old SessionMemory/LongTermMemory abstractions with unified WorkingMemory and MemoryRecord types, and updates all related modules, routes, and tests accordingly.

• Introduce agent_memory_server/working_memory.py for get/set/delete of session-scoped working memory with TTL and optional summarization.
• Replace LongTermMemory with MemoryRecord and use ULID everywhere instead of nanoid; update filters and search index to support new fields.
• Update API endpoints, MCP tools, client library, and tests to adopt the new models and behaviors.

Reviewed Changes

Copilot reviewed 27 out of 27 changed files in this pull request and generated 1 comment.

Show a summary per file

File	Description
agent_memory_server/working_memory.py	New working memory module: JSON (de)serialization, TTL, summarization, promotion
agent_memory_server/models.py	Remove SessionMemory/LongTermMemory; add WorkingMemory, MemoryRecord, expanded fields and response/request types
agent_memory_server/utils/redis.py	Extended search index schema to tag id and track persisted_at and extracted_from
tests/test_messages.py	Tests updated to use WorkingMemory/MemoryRecord and ULID
tests/test_memory_compaction.py	Compaction and indexing tests now reference MemoryRecord, use ULID, and correct pipeline usage

Comments suppressed due to low confidence (2)

agent_memory_server/working_memory.py:69

• In get_working_memory, datetime.fromtimestamp is called without timezone information, yielding a naive datetime. Consider using datetime.fromtimestamp(..., UTC) or datetime.fromtimestamp(...).replace(tzinfo=UTC) so returned WorkingMemory timestamps are timezone-aware.

last_accessed=datetime.fromtimestamp(working_memory_data.get("last_accessed", int(time.time())))

agent_memory_server/working_memory.py:127

• [nitpick] Using default=str for JSON serialization can produce non-ISO datetime strings, which may not round-trip cleanly. Consider converting timestamps to integers or using datetime.isoformat() explicitly to ensure consistent format.

await redis_client.setex(key, working_memory.ttl_seconds, json.dumps(data, default=str),)