Sim

Overview

Sim is an open-source platform designed to let teams and developers build, run, and deploy AI agent workflows quickly using a visual canvas. It focuses on composing agents, tools, and blocks into flows that can be executed immediately, making it easy to prototype multi-step agent pipelines and productionize them.

Key Features

• Visual Workflow Builder: design agent pipelines on a drag-and-drop canvas, connect nodes (agents, tools, blocks) and run flows instantly.
• Copilot Integration: an assistant that generates nodes, suggests fixes, and helps iterate on flows via natural language prompts.
• Vector Database / RAG Support: upload documents to a vector store and let agents answer grounded queries using embeddings (requires pgvector for self-hosted setups).
• Self-hosting & Local Models: multiple deployment options including an npm runner (npx simstudio), Docker Compose, and integrations to run local models with Ollama or vLLM.
• Production-Friendly: provides environment variable configuration, Docker images, and guides for running with PostgreSQL (pgvector), vLLM, or Ollama for inference.
• Realtime & Extensions: includes a realtime socket server for full functionality and supports adding tools, custom nodes, and integrations.

Deployment & Self-hosting

• Quick start (cloud): sim.ai hosted instance.
• Local/dev: npx simstudio to run locally (defaults to http://localhost:3000). Requires Docker.
• Docker Compose: official docker-compose manifests for production and Ollama-enabled setups.
• Local model support: includes instructions for using Ollama (host or container), and vLLM via OpenAI-compatible API.
• Database: PostgreSQL with the pgvector extension is required for embedding-backed features (knowledge bases, semantic search).

Developer Experience & Tech Stack

Sim is built with modern web and AI stack components:

• Frontend: Next.js (App Router), Tailwind CSS, shadcn UI
• Runtime: Bun
• DB/ORM: PostgreSQL + pgvector, Drizzle ORM
• Realtime: Socket.io
• Background jobs: Trigger.dev
• Flow editor: React Flow
• Remote code execution: E2B

The project provides dev containers, Docker images, and detailed environment variable guidance (DATABASE_URL, COPILOT_API_KEY, OLLAMA_URL, VLLM_BASE_URL, etc.) to simplify local and production deployments.

Use Cases

• Prototyping multi-agent orchestration (e.g., task decomposition, tool use, chained LLM calls)
• Building retrieval-augmented agents using your documents for grounded answers
• Running agents with local models for privacy or offline inference (Ollama, vLLM)
• Integrating into existing systems for production agent deployments

Governance & License

Sim is released under the Apache License 2.0. The README and docs provide contribution guidelines and troubleshooting instructions for common host/model/database issues.

Summary

Sim aims to reduce the friction of building complex AI agent workflows by combining a visual flow editor, Copilot-assisted development, vector-backed knowledge, and flexible self-hosting options. It's targeted at developers and teams who need to design, test, and deploy multi-step agent systems with both cloud and local model support.