Ayvu dashboard

Ayvu is a visual dashboard for designing, running, and exposing multi-agent LLM workflows without hand-wiring orchestration code. Most teams that move past a single prompt end up stitching together scripts, ad-hoc glue code, and brittle integrations just to get one model call to feed the next. Ayvu replaces that with a visual builder: every project is a directed graph (DAG) of agents, where each agent has its own system prompt, LLM provider, model, temperature, and optional JSON Schema contract for its inputs and outputs. You draw the dependencies between agents, and Ayvu automatically routes each agent's output into the next agent's context. Runs are executed by a Supabase Edge Function that topologically sorts the graph, runs independent agents in parallel waves, and propagates failures downstream — if one agent fails, everything depending on it is safely skipped instead of running on bad input. Every run keeps a full audit trail of prompts, outputs, and token usage per agent, and the DAG view updates live via Supabase Realtime as a cascade executes. Ayvu is multi-provider by default — OpenAI, Anthropic, and Groq are unified behind one adapter, and workspaces can bring their own encrypted API keys. Every cascade also ships as a REST API out of the box: endpoints to discover an agent's schema, trigger a run, poll for results, and pull auto-generated usage instructions that are just as useful to another LLM orchestrator as they are to a human developer. Built with React, TypeScript, and Supabase (Postgres, Auth, Realtime, Edge Functions), Ayvu is designed for teams who want to go from prompt to production multi-agent workflow — visually, reliably, and without reinventing orchestration infrastructure every time.