github →

New request

#688: Feature request: Hallucination rate tracking per agent

We'll provision a sandbox, run an agent against the issue, and open a draft PR. You can pull the branch and iterate from there.

Issue
enhancementmonitoring

This feature is needed by users running multi-agent Hermes swarms. A reference implementation exists in the Hermes Control Interface repo. Please add a hallucination rate metric per agent (displayed as a percentage or count), with a configurable threshold (e.g. 10%) that triggers a visible warning. This allows users running multi-agent pipelines to identify unreliable agents before they corrupt downstream outputs.

Assessmentadvisory
feature●● medium80% confidence

New per-agent hallucination metric backend tracking and frontend display with configurable thresholds.

Likely files
  • api/models/metrics.py
  • api/routes/agents.py
  • static/js/agent_metrics.js
  • static/index.html
  • api/services/evaluator.py
  • tests/test_hallucination.py
Create the request

This opens a fresh agent run and a draft PR for issue #688.

Cancel