daemon-loop

Daemon Loop Quickstart

Run evolver as a persistent daemon — continuous evolution cycles with lifecycle management.

Prerequisites

• Python 3.12+ and uv installed
• A git repository (or use EVOLVER_NO_PARENT_GIT=1 for isolated mode)
• .env file with your LLM API key (see hello-world)

1. Run the daemon loop (foreground)

cd /path/to/your/project
uv run evolver --loop

This runs continuous evolution cycles in the foreground. Each cycle:

• Collects session signals from memory/
• Queries the Hub for matching genes/capsules
• Enriches context with cognitive recall
• Runs autopoiesis self-check
• Selects the best gene/capsule for the current situation
• Dispatches a GEP prompt for LLM execution

Press Ctrl+C for graceful shutdown (one cycle completes before exit).

Control the daemon cycle interval

# Run with 60-second interval between cycles (default: 30s)
EVOLVER_LOOP_INTERVAL_MS=60000 uv run evolver --loop

# Run with review mode — pause after each prompt for human approval
uv run evolver --loop --review

# Run a single cycle then exit
uv run evolver run

2. Run as a background daemon

# Start the daemon in the background
uv run evolver start

# Check daemon status
uv run evolver status

# Tail the daemon log
uv run evolver log --lines 50

# Run a health check
uv run evolver check

# Watch health continuously (supervisor mode)
uv run evolver watch

# Stop the daemon
uv run evolver stop

# Restart
uv run evolver restart

How it works

The start command spawns evolver as a detached background process:

• PID file: memory/evolver_loop.pid
• Log file: logs/evolution.log (configurable via EVOLVER_LOGS_DIR)
• The daemon uses the same .env file and environment variables as the foreground mode
• uv run is automatically used if the project has a pyproject.toml

Configure the loop command

The daemon uses EVOLVER_LOOP_COMMAND to determine how to spawn itself:

# Default (auto-detected via uv runtime)
EVOLVER_LAUNCHER=uv uv run evolver start

# Explicit python path
EVOLVER_LAUNCHER=python uv run evolver start

# Custom command
EVOLVER_LOOP_COMMAND="python -m evolver --loop" uv run evolver start

3. Operate the daemon

Apply pending mutations (solidify)

# After a cycle produces a gene, apply it
uv run evolver solidify

# Review before applying
uv run evolver review

Check the asset call log

# Show recent Hub asset interactions
uv run evolver asset-log

# Filter by action
uv run evolver asset-log --action asset_reuse --last 20

# JSON output
uv run evolver asset-log --json

Sync with Hub

# Fetch tasks and download assets from Hub
uv run evolver sync

# Dry-run mode
uv run evolver sync --dry-run

# Force re-sync
uv run evolver sync --force

4. Configure for production

# Example .env for production daemon
EVOLVER_LOOP_INTERVAL_MS=120000       # 2 min between cycles
EVOLVER_REPAIR_LOOP_DEGRADED=1        # Degraded mode on repair loops
EVOLVER_AUTOPOIESIS=1                  # Enable self-check
EVOLVER_AUTOPOIESIS_WRITE=1            # Persist autopoiesis rules
EVOLVER_LEARNING_SIGNALS=1             # Inject learning signals
EVOLVER_GENE_INERT_BAN_STREAK=8        # Ban inert genes after 8 cycles
EVOLVER_ROLLBACK_MODE=stash            # Stash before rollback
A2A_HUB_URL=https://evomap.ai         # Hub endpoint

5. WebUI monitoring

# Start the WebUI dashboard in one terminal
uv run evolver webui --port 8080

# Access at http://localhost:8080
# Shows: system status, gene/capsule tables, pipeline insights,
#        asset economics, system health, daemon lifecycle,
#        persona commentary, skills monitor

Troubleshooting

Symptom	Solution
evolver start says "already running"	uv run evolver stop && uv run evolver start
Daemon starts but no cycles happen	Check logs/evolution.log; verify Hub is reachable
EVOLVER_REPO_ROOT not set correctly	Set explicitly or run from the repo root
"system load exceeds threshold"	The daemon skips cycles when CPU is busy; normal behavior
Power/sleep causes clock jump	The watch supervisor detects this and restarts automatically