Search Memory

AI-powered semantic search across your personal knowledge base using Nowledge Mem.

When to Use

Strong signals to search:

• 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

Prerequisites

nmem CLI – Choose one option:

Option 1: uvx (Recommended)

# Install uv if needed
curl -LsSf https://astral.sh/uv/install.sh | sh

# Run nmem directly (auto-downloads)
uvx --from nmem-cli nmem --version

Option 2: pip

pip install nmem-cli
nmem --version

Ensure Nowledge Mem server is running at http://localhost:14242

Usage

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

# Basic search
nmem --json m search "your query here"

# With importance filter
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

# Limit results
nmem --json m search "debugging tips" -n 5

Query Guidelines

• Extract semantic core from user’s request
• Preserve domain terminology
• Multi-language aware (works with any language)
• Use 3-7 core concepts for best results

Available Filters

Available unit types: fact, preference, decision, plan, procedure, learning, context, event.

Understanding Results

Parse the memories array from JSON response. Check score field:

• 0.6-1.0: Directly relevant – include in response
• 0.3-0.6: Related context – may be useful
• < 0.3: Skip – not relevant enough

Results may include a source_thread field linking the memory to the conversation it was distilled from. Use nmem --json t show <thread_id> to fetch the full conversation for deeper context.

Response Guidelines

Found relevant memories: Synthesize insights, cite when helpful

No results: State clearly, suggest distilling current discussion if valuable

Examples

# Search for React patterns
nmem --json m search "React hooks patterns" -l frontend

# Find debugging solutions
nmem --json m search "memory leak debugging" --importance 0.6

# Recent project decisions
nmem --json m search "architecture decision" -t month -n 10

Links

• Documentation
• Nowledge Mem
• Discord Community