Module process_turn 

Unified process_turn orchestration. Unified process_turn orchestration.

This module provides a single MenteDb::process_turn() method that implements the full conversation-turn pipeline used by MCP servers, cloud APIs, and any other client. The steps are:

1. Embed query + retrieve context (speculative cache → hybrid search)
2. Check pain signals
3. Store episodic memory (with write inference)
4. Detect actions, proactive recall, corrections, sentiment
5. Detect phantoms + stream contradiction check
6. Update trajectory + ghost memories
7. Update speculative cache
8. Extract facts + link edges
9. Auto-maintenance (decay / archival / consolidation on intervals)

LLM-powered features (entity resolution, topic canonicalization, contradiction verification, memory extraction) are not included here — they are client-specific and should be layered on top of the ProcessTurnResult by the caller.

Structs

DetectedAction

A detected action from the conversation.

PainWarning

A pain warning matched against the current context.

ProactiveRecall

A proactively recalled memory relevant to detected actions.

ProcessTurnInput

Input for a single conversation turn.

ProcessTurnResult

Result of processing a conversation turn.