Managing Memories

Memories are the fundamental unit in MemNexus. Each memory stores a piece of information with metadata like topics, importance, and timestamps.

Creating memories

Interactive mode

mx memories create --interactive

This walks you through each field with validation and defaults.

Direct mode

mx memories create \
  --content "Deployed v1.26.2 of core-api. Fixed OpenAPI validation issues." \
  --topics "deployment,core-api,openapi" \
  --importance 0.8

With a conversation

Group related memories into conversations for temporal context:

# Start a new conversation
mx memories create \
  --conversation-id "NEW" \
  --content "Starting work on API rate limiting feature." \
  --topics "core-api,rate-limiting,in-progress"

# Continue the same conversation (use the ID from the output)
mx memories create \
  --conversation-id "conv_abc123" \
  --content "Added rate limit middleware. 100 req/min default." \
  --topics "core-api,rate-limiting,in-progress"

Listing memories

mx memories list

Pagination

mx memories list --page 0 --limit 50
mx memories list --page 1 --limit 50

Output formats

mx memories list --format table   # Human-readable (default)
mx memories list --format json    # Machine-readable
mx memories list --format yaml    # Structured

Getting a memory

mx memories get <memory-id>

Batch get

Fetch multiple memories in one call:

mx memories get <id1> <id2> <id3>

Or pipe search results directly:

mx memories search --query "deployment" --id-only | mx memories get --stdin

Updating memories

mx memories update <memory-id> \
  --content "Updated content here" \
  --importance 0.9

Deleting memories

mx memories delete <memory-id>

Skip the confirmation prompt:

mx memories delete <memory-id> --force

Working with conversations

List conversations

mx conversations list
mx conversations list --recent 24h
mx conversations list --recent 7d

View a conversation timeline

mx conversations timeline <conv-id>

This shows all memories in chronological order — useful for reconstructing what happened during a work session.

Working with topics

List topics

mx topics list

Discover related topics

mx topics discover-related --topic-id <id>

Find similar topics

mx topics find-similar --topic-id <id> --threshold 0.7

Cluster topics

mx topics cluster --algorithm kmeans --num-clusters 5

Working with facts

Facts are structured knowledge stored as subject-predicate-object triples:

mx facts create \
  --subject "MemNexus" \
  --predicate "built_with" \
  --object "Node.js" \
  --confidence 1.0

Search facts

mx facts search --query "Node.js"

Scripting tips

Export high-importance memories

mx memories list --format json --limit 1000 | \
  jq '.data[] | select(.importance > 0.8)'

JSON format for automation

Always use --format json in scripts. The table format is for humans only.

#!/bin/bash
set -e

# Search and process results
mx memories search --query "deployment" --format json | \
  jq -r '.data[].memory.id'

Next steps

• Search — Find memories by meaning or keywords
• Core Concepts — Understand the data model