search-memory

name	search-memory
description	Search memory store when past insights would improve response. Recognize when user's stored breakthroughs, decisions, or solutions are relevant. Search proactively based on context, not just explicit requests.

Search Memory

When to Search (Autonomous Recognition)

Strong signals:

• Continuity: Current topic connects to prior work
• Pattern match: Problem resembles past solved issue
• Decision context: "Why/how we chose X" implies documented rationale
• Recurring theme: Topic discussed in past sessions
• Implicit recall: "that approach", "like before"

Contextual signals:

• Complex debugging (may match past root causes)
• Architecture discussion (choices may be documented)
• Domain-specific question (conventions likely stored)

Skip when:

• Fundamentally new topic
• Generic syntax questions
• Fresh perspective explicitly requested

Tool Usage

Use nmem CLI with --json flag for programmatic search:

# Basic search
nmem --json m search "3-7 core concepts"

# With filters
nmem --json m search "API design" --importance 0.8

# With labels (multiple labels use AND logic)
nmem --json m search "authentication" -l backend -l security

# With time filter
nmem --json m search "meeting notes" -t week

Only add --space "<space name>" when the user or environment explicitly provided that existing Mem space (for example NMEM_SPACE). Never infer a space from the current folder, git repository, branch, or project name.

Query: Extract semantic core, preserve terminology, multi-language aware

Filters:

• --importance MIN: Minimum importance score (0.0-1.0)
• -l, --label LABEL: Filter by label (can specify multiple)
• -t, --time RANGE: Time filter (today, week, month, year)
• -n NUM: Limit number of results (default: 10)

JSON Response: Parse memories array, check score field for relevance

Use thread search when the user is really asking about a prior conversation, previous session, or exact discussion:

nmem --json t search "query" --limit 5

If a memory result includes source_thread or thread search finds the likely conversation, inspect it progressively instead of loading the whole thread at once:

nmem --json t show <thread_id> --limit 8 --offset 0 --content-limit 1200

Increase --offset only when more messages are actually needed.

Scores: 0.6-1.0 direct | 0.3-0.6 related | <0.3 skip

Examples:

# Search with importance filter
nmem --json m search "database optimization" --importance 0.7

# Search with multiple labels
nmem --json m search "React patterns" -l frontend -l react

# Search recent memories
nmem --json m search "bug fix" -t week -n 5

Response

Found: Synthesize, cite when helpful None: State clearly, suggest distilling if current discussion valuable

Troubleshooting

If nmem is not in PATH: pip install nmem-cli, or on Arch Linux yay -S nmem-cli / paru -S nmem-cli

For remote servers: run nmem config client set url https://... and nmem config client set api-key ... once on this machine.

Run /status to check server connection.