Reliable buddy

• 

  Project workflow

Reliable Buddy is an intelligent agentic orchestration system that automatically detects, diagnoses, and mitigates incidents in microservices—while enforcing guardrails and human approvals where needed. Instead of just paging humans, it closes the entire loop by coordinating specialized agents across the end-to-end response workflow:

Scout Agent gathers context (metrics/logs/traces/runbooks). Triage Agent classifies the incident. Hypothesis Agent proposes root causes with validation criteria. Experiment Agent runs diagnostics to confirm the root cause. Executor Agent applies mitigations, but only after passing safety checks (reversibility, impact limits, production policies). High-risk actions automatically route through a Retool workflow for human approval; non-reversible plans are blocked outright by the guardrail engine. Postcheck Agent verifies recovery and generates structured incident reports/audit logs. Key highlights:

✅ Sponsor-tool integrations: Tonic data simulator, TinyFish/Yutori runbook retrieval, Retool dashboards/workflows, Freepik visuals. ✅ Modular architecture with metrics tracking, audit logging, and real-time dashboards. ✅ Guardrail-first execution: reversible-only auto-actions, approval workflow for risky changes, and transparent block/review paths. Reliable Buddy showcases true agentic orchestration: autonomous agents coordinate multi-step operational tasks, yet remain safely constrained by guardrails and human-in-the-loop approvals.

Built With

• fastapi
• freepik
• retool
• tinyfish
• tonic