Omnibus of changes

[abrookins]

This PR introduces background task management via Docket, adds a first version of the memory compaction system to deduplicate and merge similar memories, and improves client-side interaction with the Redis Memory Server through a new asynchronous HTTP client. These changes lay the groundwork for scalable background processing and more efficient memory storage.

What changed:

• Adds Docket integration for scheduling background tasks.
• Implements memory compaction, including exact duplicate detection (hash-based) and semantic merging (vector search + LLM-assisted merging).
• Refactors API endpoints to support async background task submission.
• Adds a full client (MemoryAPIClient) for interacting with the server.
• Updates logging configuration to prevent reconfiguration issues.
• Refactors Redis connection handling to be async.
• Adds detailed documentation on running background workers and memory compaction testing.

[Copilot]

Pull Request Overview

This PR is an omnibus change that refactors the codebase, moving functionality between modules, deprecating unused utilities, and integrating Docket for background task management while updating API and client logic. Key changes include:

• Removal and refactoring of agent_memory_server/utils.py with related updates in init.py.
• Integration of Docket-based background tasks into API endpoints and dependency injection.
• Numerous updates and cleanups in API endpoints, logging, and client wrappers.

Reviewed Changes

Copilot reviewed 38 out of 38 changed files in this pull request and generated 2 comments.

Show a summary per file

File	Description
agent_memory_server/utils/init.py	Adds re-exports for Keys and _redis_pool after removal of utils.py.
agent_memory_server/utils.py	Entire file removed, consolidating functionality into other modules.
agent_memory_server/summarization.py	Updates import and redis connection handling; minor adjustments to summarization parameters.
agent_memory_server/models/*	New and updated model definitions including addition of memory_hash for deduplication.
agent_memory_server/messages.py	Added multiple debug print statements and refactored message and metadata handling.
agent_memory_server/mcp.py	Adjusted imports and input patterns for task-based functions.
agent_memory_server/main.py	Refined startup logic with explicit port parsing and Docket task initialization.
agent_memory_server/logging.py	Added a safeguard for repeated logger configuration.
agent_memory_server/llms.py	Changed parameter naming in create_chat_completion for clarity and updated client caching logic.
agent_memory_server/dependencies.py	Introduced a custom DocketBackgroundTasks class and updated dependency injection.
Other files	Updates to docket tasks, configuration, compaction documentation, API client and API endpoints.

Comments suppressed due to low confidence (1)

agent_memory_server/llms.py:307

• [nitpick] The parameter was renamed from 'progressive_prompt' to 'prompt', but inline comments or documentation may still reference the old name; please update the comments and docstrings to reflect the new parameter name.

async def create_chat_completion(self, model: str, prompt: str, response_format: dict[str, str] | None = None, functions: list[dict[str, Any]] | None = None, function_call: dict[str, str] | None = None) -> ChatResponse:

[Copilot]

Debug print statements are present and may leak sensitive information; consider replacing these with proper logger.debug calls or removing them before production.

Suggested change

	print(f"DEBUG: Listing sessions with key: {sessions_key}")
	# Debug: List all keys in Redis
	all_keys = await redis.keys("*")
	print(f"DEBUG: All keys in Redis: {all_keys}")
	# Debug: Check if the sessions key exists
	exists = await redis.exists(sessions_key)
	print(f"DEBUG: Does {sessions_key} exist? {exists}")
	# Debug: Try to get all sessions directly
	all_sessions = await redis.zrange(sessions_key, 0, -1)
	print(f"DEBUG: Direct zrange of {sessions_key}: {all_sessions}")
	async with redis.pipeline() as pipe:
	pipe.zcard(sessions_key)
	pipe.zrange(sessions_key, start, end)
	total, session_ids = await pipe.execute()
	print(f"DEBUG: Found {total} sessions: {session_ids}")
	logger.debug(f"Listing sessions with key: {sessions_key}")
	# Debug: List all keys in Redis
	all_keys = await redis.keys("*")
	logger.debug(f"All keys in Redis: {all_keys}")
	# Debug: Check if the sessions key exists
	exists = await redis.exists(sessions_key)
	logger.debug(f"Does {sessions_key} exist? {exists}")
	# Debug: Try to get all sessions directly
	all_sessions = await redis.zrange(sessions_key, 0, -1)
	logger.debug(f"Direct zrange of {sessions_key}: {all_sessions}")
	async with redis.pipeline() as pipe:
	pipe.zcard(sessions_key)
	pipe.zrange(sessions_key, start, end)
	total, session_ids = await pipe.execute()
	logger.debug(f"Found {total} sessions: {session_ids}")

[Copilot]

[nitpick] Consider using the logger for debug output instead of print statements to maintain consistency with the rest of the project and avoid undesired console output in production.

Suggested change

	print(f"Using port from command line: {port}")
	except (ValueError, IndexError):
	# If conversion fails or index out of bounds, use default
	print(f"Invalid port argument, using default: {port}")
	else:
	print(f"No port argument provided, using default: {port}")
	# Explicitly unset the PORT environment variable if it exists
	if "PORT" in os.environ:
	port_val = os.environ.pop("PORT")
	print(f"Removed environment variable PORT={port_val}")
	logger.info(f"Using port from command line: {port}")
	except (ValueError, IndexError):
	# If conversion fails or index out of bounds, use default
	logger.warning(f"Invalid port argument, using default: {port}")
	else:
	logger.info(f"No port argument provided, using default: {port}")
	# Explicitly unset the PORT environment variable if it exists
	if "PORT" in os.environ:
	port_val = os.environ.pop("PORT")
	logger.info(f"Removed environment variable PORT={port_val}")