Inspiration
In the modern creator and marketing landscape, building a brand requires an active presence across multiple channels—LinkedIn, Twitter/X, Newsletters, WhatsApp, Facebook, and Instagram. However, each platform has completely different native cultures, formatting conventions, character limits, and audience expectations.
Cross-posting the exact same copy fails, while manually rewriting content for 6+ distinct channels takes hours of tedious work. We asked ourselves: What if a single creative idea could trigger an autonomous multi-agent swarm that researches, strategizes, adapts, and drafts publication-ready assets across every channel simultaneously?
That inspiration led to OmniCast AI—an autonomous multi-platform studio powered by Google Antigravity and Google Gemini.
What it does
OmniCast AI transforms a single user topic or raw prompt into a comprehensive, multi-platform campaign through a synchronized 9-node visual agentic workflow:
- 🔍 Deep Research Dossier (Level 1): Crawls live web intelligence to gather verified facts, market data, audience sentiment, and viral narrative angles.
- 🧠 Strategic Campaign Architecture (Level 2): Formulates a core strategic thesis, demographic targeting profile, and channel-by-channel adaptation blueprints.
- 📐 Master Adaptation Matrix (Level 3): Calibrates native whitespace rules, character limits, hook structures, and call-to-action dynamics.
- 💼 LinkedIn Architect (Level 4): Crafts high-authority thought-leadership posts with hook lines, clean whitespace, and engagement questions.
- 🐦 X / Twitter Master (Level 4): Authors high-retention 7-tweet viral narrative threads (
Tweet 1/7to7/7). - 💬 WhatsApp Specialist (Level 4): Formats high-impact broadcasts using native
*bold*syntax and emoji bullet points. - 📧 Newsletter Editor (Level 4): Writes complete 600+ word editorial deep dives featuring 3 A/B test subject lines, preview text, H2 sections, and an actionable playbook.
- 👥 Facebook Engine (Level 4): Composes authentic, story-driven community posts that spark comment discussions.
- 📸 Instagram Specialist (Level 4): Crafts scroll-stopping captions with frictionless CTAs and 15–20 targeted hashtags.
🌟 Key Studio Features:
- Interactive Visual Canvas: Live animated SVG wires showing agent-to-agent data flow in real time.
- Full Granular Control: In-studio markdown modal inspector with 1-click text editing, direct copying, and
.txtdownloads. - 1-Click Bulk Export: Download the entire multi-channel campaign as a structured
.ziparchive. - Auto-Save History Drawer: Locally persisted campaign library with instant restore and review capabilities.
- Virality Scorecard: Real-time multi-dimensional scoring evaluating hook strength, clarity, and 60-second retention estimates.
How we built it
OmniCast AI was architected from the ground up using Google Antigravity and the modern Google Cloud ecosystem:
- Autonomous Agent Orchestration: Built with Google Antigravity, organizing specialized agent roles (
ResearchAgent,PlannerAgent,DistributionSwarm) communicating via asynchronous Server-Sent Events (SSE). - AI & LLM Engine: Powered by Google Gemini (
gemini-3.5-flash-liteandgemini-3.6-flash) via the Google Generative Language API and Vertex AI for ultra-fast, high-reasoning output. - Backend: Asynchronous FastAPI Python application with custom streaming pipelines, multi-agent state machines, and dynamic rate-limit management.
- Frontend UI/UX: High-performance responsive studio canvas built with Tailwind CSS, FontAwesome, and vanilla JavaScript with zero framework overhead.
- Cloud Infrastructure: Containerized using Docker and deployed globally serverless on Google Cloud Run in
us-central1with automated scaling and health monitoring.
Challenges we ran into
- Cross-Platform Tone Bleed: Ensuring that platform agents didn't cross-contaminate formats (e.g., preventing newsletter content from reading like tweets). We resolved this by engineering strictly isolated agent prompt contexts and individual platform adaptation rules.
- Real-time SSE Streaming in Multi-Agent Swarms: Managing asynchronous event loops across sequential and parallel agents while streaming live state updates to the frontend SVG wire graph without UI stuttering.
- Pacing & Natural Progression: Striking the perfect balance between instant responsiveness and deliberate multi-step reasoning so users can see each specialized agent contributing to the campaign live.
Accomplishments that we're proud of
- Building a full-stack, enterprise-grade multi-agent studio with a stunning, production-quality visual workflow graph.
- Achieving end-to-end multi-platform generation with deep research grounding in ~45–50 seconds.
- Delivering a zero-friction user experience with in-place editing, 1-click bundle downloads, and persistent campaign history.
- Deploying 100% live on Google Cloud Run with robust containerized reliability.
What we learned
- How to leverage Google Antigravity to rapidly design, scaffold, test, and orchestrate complex multi-agent architectures.
- Best practices for prompt isolation and few-shot calibration when generating highly distinct content formats from a shared research grounding.
- How to deploy and manage production-ready serverless AI workloads on Google Cloud Run.
Links & Live Demo
- 🌐 Live Web Application: https://omnicast-ai-296127548041.us-central1.run.app
- 💻 GitHub Repository: https://github.com/ibrahimdot20/OmniCast-AI## Inspiration
What it does
How we built it
Challenges we ran into
Accomplishments that we're proud of
What we learned
What's next for OmniCast AI
Built With
- fastapi
- google-antigravity
- google-cloud-run
- google-gemini
- javascript
- multi-agent-systems
- python
- tailwind-css
- vertex-ai
