prateekdevisingh/kakveda
Open-source failure intelligence platform for LLM & agent systems. Adds failure memory, pre-flight warnings, pattern detection, and system health scoring.
# Technical Summary Treats failures as queryable first-class entities in a **Failure Knowledge Base (GFKB)** rather than logs, enabling pattern detection and pre-flight warnings before repeat failures occur. Built on event-driven architecture with fingerprinting and classifier stages that feed into a unified dashboard integrating LLM traces, infrastructure metrics, container/Kubernetes telemetry, and APM signals. Ships with `kakveda-netra`, a native host agent collecting CPU/memory/network/process/Docker/K8s payloads, designed for self-hosted deployment with direct dashboard-driven runtime instrumentation controls.
Stars
9
Forks
5
Language
Python
License
Apache-2.0
Category
Last pushed
Mar 01, 2026
Commits (30d)
0
Get this data via API
curl "https://pt-edge.onrender.com/api/v1/quality/agents/prateekdevisingh/kakveda"
Open to everyone — 100 requests/day, no key needed. Get a free key for 1,000/day.
Featured in
Higher-rated alternatives
truera/trulens
Evaluation and Tracking for LLM Experiments and AI Agents
traceroot-ai/traceroot
Find the Root Cause in Your Code's Trace
future-agi/traceAI
Open Source AI Tracing Framework built on Opentelemetry for AI Applications and Frameworks
VishApp/multiagent-debugger
Multi-Agent Debugger: An AI-powered debugging system using CrewAI to orchestrate specialized...
evilmartians/agent-prism
React components for visualizing traces from AI agents