Turns raw agent traces into labeled evaluation data by classifying each failure against a deterministic rule-based taxonomy, attaching the evidence and sending unclear traces to human review.
-
Updated
Jul 18, 2026 - Python
#
failure-taxonomy
Here are 4 public repositories matching this topic...
Automated Quality Evaluation & Failure-Taxonomy Platform for Voice and Chat AI Agents. Hybrid deterministic rules + LLM-as-a-judge, 7-layer root-cause attribution, zero-trust PII masking, and HDBSCAN failure clustering with live interactive dashboard.
python ai-agents hdbscan conversational-ai root-cause-analysis pydantic fastapi voice-ai llm-evaluation pii-masking failure-taxonomy eval-platform
-
Updated
Oct 4, 2026 - Python
105 documented failures from an agent-run research lab, classified by phase, cause, and catching layer. Sealed dataset v1.1 plus the audited taxonomy and submission format.
-
Updated
Oct 19, 2026
A structured Chinese failure atlas and casebook for fictional / roleplay dialogue models.
-
Updated
Aug 24, 2026 - JavaScript
Add this topic to your repo
To associate your repository with the failure-taxonomy topic, visit your repo's landing page and select "manage topics."