Module conversational 

Conversational Memory Search (Phase N5).

User-facing sub-50ms hybrid search path built on top of RecallEngine.

Features:

• Query complexity classification: detects sensitive queries, tool-call intent, and multi-turn patterns to adjust search behavior
• LRU query cache: caches recent query results for instant re-query
• Performance metrics: tracks latency percentiles and cache hit rates
• Galaxy filtering: restrict search to specific memory galaxies
• Snippet extraction: returns truncated content snippets for UI display

Architecture

User Query
    │
    ▼
┌──────────────────┐     ┌─────────────────┐
│ QueryClassifier  │────▶│  Cache Lookup   │
└──────────────────┘     └────────┬────────┘
                                  │ miss
                                  ▼
                         ┌─────────────────┐
                         │  RecallEngine   │
                         │  hybrid_search  │
                         └────────┬────────┘
                                  │
                                  ▼
                         ┌─────────────────┐
                         │  Result Builder │
                         │  + Snippets     │
                         └────────┬────────┘
                                  │
                                  ▼
                         ┌─────────────────┐
                         │  Cache Store    │
                         └─────────────────┘

Environment Variables

Variable	Default	Description
WM_CONVERSATIONAL_CACHE_SIZE	128	Max cached queries
WM_CONVERSATIONAL_SNIPPET_LEN	200	Max snippet length (chars)
WM_CONVERSATIONAL_DEFAULT_LIMIT	10	Default result limit

Structs

ConversationalConfig

Configuration for conversational memory search.

ConversationalResult

A conversational search result with snippet and metadata.

ConversationalSearch

User-facing conversational memory search engine.

QueryClassification

Classification of a conversational search query.

SearchMetrics

Performance metrics for conversational search.