Index discrete memories in long-term storage, add a new memory prompt REST API endpoint, and YAML config

[abrookins]

Adds a few features:

1. Allow storing settings in a YAML file instead of just environment variables (still available as well).
2. Add an option that enables extracting discrete memories from messages for long-term storage, using an LLM, in addition to the current strategy of indexing the messages themselves.
3. Move the "memory prompt" from the MCP implementation into a core REST service, so it's available in both places. Clients can use this to send input text and get back a set of messages that include any relevant context found in memory, both the messages in the current session (if one is stored) and any relevant memories found via semantic search.

[Copilot]

Pull Request Overview

This PR centralizes configuration in YAML, enhances memory handling, and promotes reuse of the memory prompt across services.

• Load settings from a YAML file (with env-var overrides) via Settings and load_yaml_settings.
• Add discrete memory extraction toggle during long-term indexing.
• Refactor the “hydrate_memory_prompt” tool into a core /memory-prompt API with MemoryPromptRequest/Response.

Reviewed Changes

Copilot reviewed 13 out of 13 changed files in this pull request and generated 3 comments.

Show a summary per file

File	Description
tests/test_models.py	Rename SearchPayload → SearchRequest
tests/test_mcp.py	Update tool name to "memory_prompt" and response parsing
tests/test_extraction.py	Switch topic extraction calls to extract_topics_bertopic and add integration tests
tests/test_api.py	Add validation of /memory-prompt responses and combine session & long-term memory
agent_memory_server/test_config.py	Add tests for YAML loading and env-var overrides
agent_memory_server/config.py	Implement load_yaml_settings, support YAML + env-vars
agent_memory_server/models.py	Add new request/response models and rename payloads
agent_memory_server/mcp.py	Rename MCP tool, delegate to core memory prompt API
agent_memory_server/long_term_memory.py	Invoke discrete memory extraction conditionally
agent_memory_server/extraction.py	Rename NER extraction to extract_topics_bertopic and unify path logic
agent_memory_server/api.py	Create /memory-prompt endpoint and helper for window sizing
agent_memory_server/client/api.py	Update client methods for new request models
.env.example	Rename MAX_WINDOW_SIZE → WINDOW_SIZE

Comments suppressed due to low confidence (1)

agent_memory_server/api.py:42

• get_model_config is not imported or defined in this module, leading to a NameError when model_name is provided. Add from agent_memory_server.api import get_model_config (or the correct import path) at the top.

model_config = get_model_config(model_name)