Skip to main content
The longmem stats command displays statistics about your AI’s memory database.

​
Syntax

longmem stats
This command takes no options or arguments.

​
Requirements

The daemon must be running to fetch statistics. If the daemon is stopped, you’ll see: 
$ longmem stats
Error: Daemon not running. Start with: longmem start

​
Output

$ longmem stats
LongMem Statistics
===================
Sessions: 42
Observations: 1,234
User observations: 156
Concepts: 89
Pending compressions: 0

​
Statistics Explained

​
Sessions

Number of distinct coding sessions tracked. A session represents one continuous period of AI interaction, typically:
  • One terminal session
  • One IDE window
  • One conversation thread
Sessions are created with session.start and ended with session.end. What it means:
  • Low count (0-10): New installation or infrequent use
  • Medium count (10-100): Regular use over weeks
  • High count (100+): Heavy use over months

​
Observations

Total number of individual observations stored, including:
  • Tool uses (file reads, edits, commands)
  • AI responses
  • User messages
  • Hook events
This is the raw, uncompressed event stream. What it means:
  • Typically grows by 100-500 per active session
  • High count indicates lots of AI activity
  • Observations older than 7 days may be compressed into concepts

​
User Observations

Subset of observations that are direct user messages (not tool uses or AI responses). What it means:
  • Shows how much the user is directing the AI
  • Typically 10-30% of total observations
  • High ratio indicates user-driven interactions
  • Low ratio indicates autonomous AI work

​
Concepts

Number of compressed memory concepts. These are synthesized summaries created by the compression system from related observations. Examples:
  • “User prefers TypeScript over JavaScript”
  • “Project uses Tailwind for styling”
  • “Debugging CORS issue with API endpoint”
What it means:
  • Grows slowly (1-5 per session after compression)
  • Each concept represents multiple related observations
  • More concepts = richer long-term memory

​
Pending Compressions

Number of observations waiting to be compressed into concepts. What it means:
  • 0 - All caught up, compression is working
  • 1-10 - Normal during active use
  • 10-50 - Queue building up, compression is lagging
  • >50 - Compression is failing or circuit breaker is open
When to worry:
  • If pending count keeps growing over hours/days
  • Check longmem logs for compression errors
  • Verify API keys are configured correctly

​
Understanding Your Memory Growth

​
Healthy Memory

Sessions: 25
Observations: 3,450
User observations: 520
Concepts: 67
Pending compressions: 2
  • Observations per session: ~138 (normal)
  • User ratio: 15% (AI-heavy, normal for coding)
  • Concepts: ~3 per session (good compression)
  • Pending: Very low (compression working)

​
Memory Issues

Sessions: 10
Observations: 8,500
User observations: 340
Concepts: 3
Pending compressions: 127
  • Observations per session: 850 (very high)
  • Concepts: Very low (compression not working)
  • Pending: High (backlog building up)
What to do:
  1. Check logs: longmem logs
  2. Look for compression errors
  3. Verify API configuration
  4. Consider restarting daemon

​
Querying Stats Directly

You can also fetch stats via HTTP: 
curl http://127.0.0.1:5939/stats
Returns JSON: 
{
  "totalSessions": 42,
  "totalObservations": 1234,
  "totalUserObservations": 156,
  "totalConcepts": 89,
  "pendingCompressions": 0
}

​
Stats in Scripts

You can parse stats in shell scripts: 
# Get pending count
PENDING=$(longmem stats | grep "Pending" | awk '{print $3}')

if [ "$PENDING" -gt 50 ]; then
  echo "Warning: Compression queue is backed up"
fi
Or use the JSON API: 
# Get pending count from API
PENDING=$(curl -s http://127.0.0.1:5939/stats | jq -r '.pendingCompressions')

if [ "$PENDING" -gt 50 ]; then
  echo "Warning: Compression queue is backed up"
fi

​
Exit Codes

  • 0 - Stats retrieved successfully
  • 1 - Daemon not running or error fetching stats

​
Troubleshooting

​
”Daemon not running” error

# Check status
longmem status

# Start if needed
longmem start

# Try again
longmem stats

​
Stats show “N/A”

Sessions: N/A
Observations: N/A
This means the /stats endpoint returned incomplete data. Check: 
# View daemon logs
longmem logs

# Check for database errors
grep -i error ~/.longmem/logs/daemon.log

# Verify database exists
ls -lh ~/.longmem/longmem.db

​
Stats request times out

If stats take >5 seconds, the request times out. This can happen if:
  • Database is very large (>1GB)
  • Database is locked
  • Daemon is under heavy load
Fix: 
# Restart daemon
longmem stop
longmem start

# Try again
longmem stats

Build docs developers (and LLMs) love

Get started for freeTalk to us