RagaAI-Catalyst

Why we included this project

Debugging a multi-agent system usually means staring at logs and guessing which step produced the bad output. RagaAI Catalyst is a Python SDK that traces the agent's steps, including the LLM calls and tool interactions behind them, then shows the results in a self-hosted dashboard with timeline and execution-graph views, so you can see where a run went wrong instead of inferring it. It also bundles evaluation tooling: define datasets, attach metrics like faithfulness and hallucination with configurable thresholds, and run experiments to compare versions. Prompt management, synthetic data generation, guardrails, and red-teaming round out the package, so the same toolchain covers both debugging a single run and ongoing quality checks. Teams building RAG pipelines or multi-agent systems who want tracing and evaluation without stitching several services together will find most of what they need here.