<2-3 sentences summarising what the project addresses and solves, the target audience, and the benefit.> Flowboard addresses repetitive e-commerce video creation by providing a local, single-user infinite canvas for AI media workflows. It allows users to compose characters, products, and scenes as a directed graph, automating prompt generation and streamlining video production for creators and power users.

How It Works

This project utilizes a local-only, infinite canvas workspace where users construct AI media workflows as directed graphs. Nodes represent reusable assets like characters and products, connected by data dependencies. Generation is proxied via a mandatory Chrome extension to Google Flow (Veo 3.1 / GEM_PIX_2), with auto-prompts synthesized by a local LLM CLI, creating a unified campaign canvas.

Quick Start & Requirements

• Install: Clone repo, run make install (or pip/npm manually), then make agent and make frontend.
• Prerequisites: Python 3.11, Node 20+, Chrome/Chromium (for MV3 extension), one LLM CLI (Claude Code default, Gemini, Codex beta) on PATH, and a Google Flow Pro or Ultra plan (free/trial tiers are non-functional). WSL2 recommended for Windows.
• Links: Repo, Google Flow.

Highlighted Details

• Node-Based Graph: Reusable "Ref" nodes (Character, Visual Asset) anchor identity and assets. Composition nodes (Image, Video) build scenes and generate videos via Veo 3.1 i2v.
• Auto-Prompt Synthesis: LLM CLI generates context-aware prompts from upstream aiBrief descriptions, adapting motion vocabulary to scene context.
• Multi-LLM Support: Swappable providers for Auto-Prompt/Vision/Planner (Claude Code, Gemini CLI, OpenAI Codex).
• Workflow Ergonomics: Features include a drop-add popover, easy edge editing, clone variant functionality, and an activity feed for monitoring operations.

Maintenance & Community

A shared community for FlowKit and Flowboard exists via a Facebook group: facebook.com/groups/flowkit.flowboard.community. Sponsorships are noted for Vietnam MoMo, Binance Pay, and International card payments.

Licensing & Compatibility

The project is proposed under the MIT license, with the license file pending. It is designed as a personal, local-only, single-user tool.

Limitations & Caveats

• Google Flow Plan: Pro or Ultra is strictly required; free/trial accounts will fail.
• Chrome Extension: Mandatory for all Google Flow interactions; the agent does not connect directly.
• Local-Only: Operates as a single-user loopback service.
• LLM CLI: Auto-prompting requires a configured LLM CLI; manual prompting is unaffected. OpenAI Codex provider is in beta.
• Content Filters: Veo/Imagen content filters are surfaced directly.