All projects · AI Agents & RAG
OpsAgent AI — AI Operations Agent & Supervised Execution Engine
An enterprise-grade AI Operations Agent platform that retrieves company procedures from Supabase using pgvector, determines compliant operational…
An enterprise-grade AI Operations Agent platform that retrieves company procedures from Supabase using pgvector, determines compliant operational plans, asks for missing request details, and executes approved tasks through tools (Stripe, Gmail, Slack, Google Sheets, Supabase). Adhering strictly to the engineering clue that an operations agent must first check company rules and know when human approval is required, it gates sensitive high-risk actions behind manager sign-off while allowing low-risk tasks to run autonomously.
- 1:1 pixel-accurate layout matching reference design with request details, procedure retrieved, execution plan, approval queue, and recent activity
- pgvector Semantic Retrieval matching operational requests against company SOPs with cosine similarity scores and condition verification checklists
- Missing details detector and interactive clarification dialog that prompts for required fields before generating plans
- Human-in-the-loop approval gating keeping sensitive financial operations (Stripe refunds) locked behind role-based manager sign-off
- Supervised tool execution across Supabase, Stripe, Gmail, Slack, and Google Sheets with strict input validation
- Task controls for pause, resume, cancel, retry, and automated failure detection with dynamic recovery planning
- Complete Supabase schema with pgvector extension, match_procedures RPC function, RLS policies, and seed data
Built with: Vanilla JavaScript, HTML5 / CSS3, Supabase & pgvector, Semantic Procedure RAG, Tool Execution Engine, Role-Based Access Control (RBAC), Prompt Injection Defense, Google Fonts (Sora, Outfit, Chivo Mono)