ishandutta2007/Velocity

What it solves

Open Velocity is an open-source, agent-first IDE designed to prevent vendor lock-in by providing a universal gateway to multiple LLMs. It allows developers to delegate complex coding tasks to autonomous AI agents that can plan, write, and verify code without being tied to a single AI provider.

How it works

The platform uses a modular, container-native architecture where a central Gateway and Orchestrator manage the interaction between the Web UI, a Workspace Manager (which handles isolated Docker containers), and an AI Model Gateway that connects to various LLMs like Gemini, Claude, GPT-4, and Llama.

Its backend core (OpenCentravity) implements several enterprise-grade coordination features:

• Hierarchical Swarms: Tracks parent-child relationships between agents for complex branching workflows.
• Whiteboard Message Bus: A shared communication system for sub-agents to coordinate efforts.
• Distributed Lock Manager: Prevents multiple agents from editing the same file simultaneously using file-level mutexes.
• Memory Snapshots: Periodically saves agent states to allow for self-healing from failures.
• Cost Tracking: Monitors token counts and API costs in real-time to enforce budget ceilings.

Who it’s for

Developers who want a model-agnostic, self-hosted AI coding environment that supports multi-agent collaboration and maintains strict privacy and data control.

Highlights

• Universal LLM Gateway: Connects to a wide variety of models including GPT-4, Claude 3.5, Gemini 1.5 Pro, and Llama 3.
• Multi-Agent Collaboration: Ability to spawn parallel agents to work on different project parts.
• ** uma Verifiable Artifacts**: Generates task lists, screenshots, and test results to verify agent work.
• Enterprise-Grade Backend: Includes a robust SQLite database layer with sequential migrations and connection safety mechanisms.