Module auto_capture 

Post-stream auto-capture of memories from conversation turns.

After each assistant turn, this module inspects the latest messages and calls add_memories_from_conversation to let the LLM-backed memory extractor identify and persist notable information.

This is intentionally fire-and-forget: errors are logged but not propagated so that the main streaming path is never blocked.

Structs

ConversationMessage

Represents a single message in a conversation, compatible with the add_memories_from_conversation wire format.

Functions

capture_from_stream_end

Extract and persist memories from recent conversation messages.