Summary

EpicStaff addresses the "ownership problem" of AI agent workflows, often locked in code inaccessible to operations teams. This source-available, self-hosted platform features a visual editor over a Django backend, enabling operations teams to build, own, and audit AI agent flows. Engineers integrate systems via Python, while operations teams manage workflows through an intuitive interface, bridging development and operational oversight.

How It Works

The platform uses a node-based visual editor for designing AI agent workflows, allowing drag-and-drop construction. Custom Python code can be injected into any node for specific logic or integrations. A robust Django backend powers orchestration, ensuring production-ready performance. Key features include persistent agent memory (Redis/PostgreSQL) and state management for user/organization variables, facilitating context retention and reliable execution.

Quick Start & Requirements

Installation requires Git and Docker Desktop. Users clone the repository and execute a platform-specific docker-compose up --build command via their terminal to download, configure, and launch the system, accessible at http://localhost. Alternative setup options include "Partly Local Setup" and "Podman Support." Further details are available in the Wiki.

Highlighted Details

• Visual Workflow Builder: Drag-and-drop interface for designing, debugging, and validating multi-step AI agent workflows.
• Python Core & Custom Tools: Inject/execute Python code within nodes; define custom Python functions callable by agents.
• RAG & Vector Store: Supports document uploads, knowledge collections, and index customization with Naive Vector Search or GraphRAG.
• Persistent Agent Memory: Dual-layer memory (short-term context, long-term state) via PostgreSQL (pgvector) and Redis.
• LLM/Embedder Flexibility: Native support for OpenAI, Anthropic, and custom local models.
• Multi-Agent Collaboration: Enables "digital crews" where multiple AI agents collaborate.
• Webhooks & Human Input Control: Trigger workflows programmatically and pause execution for manual review.

Maintenance & Community

The project acknowledges contributions from the broader open-source ecosystem, specifically highlighting the f-flow library by Foblex as core to the visual editor. No specific community channels or details on active maintainers, sponsorships, or a public roadmap are provided in the README.

Licensing & Compatibility

The platform is described as "Source-Available," meaning its source code is accessible but usage rights are defined by its specific license (consult LICENSE file). Compatibility for commercial use or closed-source linking depends on these license terms.

Limitations & Caveats

The primary caveat is its "Source-Available" licensing, which may impose restrictions beyond typical open-source licenses. The default setup relies heavily on Docker, potentially posing a barrier for users unfamiliar with containerization. Alternative setup methods suggest the default Docker configuration might require advanced user intervention.