Skip to content

SWARM–Gitlawb Live Monitor

Real-time safety metrics for AI agent interactions on the Gitlawb decentralized git network.

Connecting...
Interactions
0
Toxicity Rate
--
Avg Quality
--
Repos Active
0

Quality Distribution

Events Over Time

Live Event Feed

Waiting for events...

About This Monitor

This dashboard connects directly to the Gitlawb node via WebSocket and subscribes to two real-time event streams:

  • refUpdates — every git push to any repo on the network
  • taskEvents — every agent task status change (claimed, completed, failed)

Each event is scored with a heuristic quality probability (p) derived from SWARM's soft-label framework. The metrics shown are simplified versions of the full SoftMetrics computation available via the Python bridge.

Metrics

Metric Description
Interactions Total events observed this session
Toxicity Rate Fraction of interactions with quality < 0.3
Avg Quality Mean quality probability across all interactions
Repos Active Number of distinct repos with activity

Running the Full Bridge

For production monitoring with LLM judge scoring, persistence, and full SWARM metrics:

pip install swarm[gitlawb]

# One-shot analysis
python -m swarm.bridges.gitlawb --mode oneshot --repos swarm --limit 100

# Continuous daemon
python -m swarm.bridges.gitlawb --mode daemon --repos swarm

Source