🦅 VantAIge: The Strategic Brand Engine

Status: Alpha / Active Development
Stack: Next.js, Cloud Firestore, Gemini Multimodal Live via Vertex AI (WebSockets)

VantAIge is an AI-powered "Marketing Director" that combines real-time situational awareness (vision/audio) with a deep, persistent memory of brand identity (Vibe Profiles).

---

📋 For Judges

Requirement	Details
🔗 Public Code Repository	https://github.com/dexkcd/vantaige
🖥️ Spin-Up Instructions	See Quick Start below
☁️ Proof of Google Cloud Deployment	See GCP Proof below
🏗️ Architecture Diagram	See Architecture Diagram below (Pro tip: Export or screenshot this for the submission image carousel)

---

🚀 Quick Start (Spin-Up Instructions)

Reproducible local setup for judges:

Prerequisites

• Node.js 18+
• Python 3.10+ (3.12+ recommended)
• gcloud CLI (install) — authenticated with gcloud auth application-default login
• Google Cloud project with billing enabled; APIs: aiplatform.googleapis.com, Firestore

1. Clone and install

git clone https://github.com/dexkcd/vantaige.git
cd vantaige
npm install

2. Environment

Create .env.local in the project root:

GOOGLE_CLOUD_PROJECT=your_gcp_project_id
GOOGLE_CLOUD_LOCATION=us-central1
NEXT_PUBLIC_WS_URL=ws://localhost:8000/ws

3. Start the WebSocket backend (Terminal 1)

cd backend
python3 -m venv .venv
source .venv/bin/activate   # Windows: .venv\Scripts\activate
pip install -r requirements.txt
uvicorn main:app --reload --host 0.0.0.0 --port 8000

4. Start the Next.js app (Terminal 2)

npm run dev

5. Open

Visit http://localhost:3000. Choose New Session or Continue Session (with passcode), then Start conversation.

---

☁️ Proof of Google Cloud Deployment

VantAIge runs on Google Cloud Run and uses multiple GCP services. Code references:

GCP Service	Code Location	Purpose
Vertex AI (Gemini Live)	backend/main.py	WebSocket bridge to Gemini Multimodal Live API
Vertex AI (Imagen, Veo)	src/app/actions/memory.ts	Image generation (Imagen 4.0), video generation (Veo 3.1)
Cloud Firestore	src/lib/firestore.ts	Vibe profiles, session logs, brand assets, marketing plans
Firebase Storage	src/lib/storage.ts	Brand asset images, Veo video output
Cloud Run	scripts/deploy-cloud-run.sh	Deployment script for both services

Optional proof: A short screen recording of the app running on GCP (e.g. Cloud Run console, logs) can be included in your submission.

---

🏗️ Architecture Diagram

flowchart TB
    subgraph Client["🖥️ Frontend (Next.js)"]
        UI[Studio UI]
        Canvas[Webcam + Screen Capture]
        Mic[Microphone]
    end

    subgraph GCP["☁️ Google Cloud"]
        subgraph CloudRun["Cloud Run"]
            WS[Python WebSocket Bridge<br/>backend/main.py]
            Next[Next.js App]
        end

        subgraph VertexAI["Vertex AI"]
            Live[Gemini Multimodal Live API]
            Imagen[Imagen 4.0]
            Veo[Veo 3.1]
        end

        Firestore[(Cloud Firestore)]
        Storage[Firebase Storage]
    end

    Canvas --> WS
    Mic --> WS
    WS <-->|WebSocket| Live
    UI --> Next
    Next -->|REST| Imagen
    Next -->|REST| Veo
    Next --> Firestore
    Next --> Storage
    Live -.->|tool calls| Next

Loading

Pro tip: Screenshot this diagram or export it as an image for the submission file upload / image carousel.

---

🎯 Core Mission

Transition AI from a "chat box" to a proactive partner that understands physical products, digital designs, and brand DNA simultaneously.

🛠 Feature Roadmap & Status

1. 👁️ The "Live Vision" Engine

• Dual-Stream Input (Webcam + Screen-Share)
• Camera lens toggle (front ↔ back) with live restart
• Low-Latency Compositor (1FPS Gemini Stream)
• Proactive Visual Auditing (Interruption logic for brand mismatches)

2. 🎙️ High-Fidelity Interaction (The "Director")

• True Barge-in Support (VAD-driven buffer clearing)
• Affective Intelligence (Tone detection & adaptation)
• Zero-Latency Hand-off (Syncing UI assets with voice)

3. 🧠 Strategic Brain (Memory Layer)

• Vibe Profile: Persistent brand DNA storage (Firestore)
• Session Continuity: Recalling past decisions & palettes
• Real-time Updates: upsert_vibe_profile mid-call
• Cross-session trend analysis (Gemini analysis over session logs + roadmap)

4. 🎨 Real-Time Execution

• Nano Banana Asset Gen: Image generation during live calls (Imagen 4.0 via Vertex AI)
• Short-Form Video: TikTok/YouTube Shorts (9:16) via Veo 3.1, with a daily limit of 5 videos or 30 seconds total per brand (UTC day) to control generation costs
• Launch Pack Sidebar: Pinned assets & copy for review
• Strategic Grounding: Google Search integration for trends
• Blog Post Generation: Long-form blog articles (Markdown or HTML) generated from live sessions and the Vibe Profile, auto-pinned in the Launch Pack Review tab so they can be copy-pasted into Medium, Ghost, Substack, or custom blogs.

5. 🔄 The "Refine" Loop (Agentic Workflow)

• Kanban Bridge: Turning ideas into "Draft Plans"
• Roadmap Task Detail View with modal, image/video posts, status workflow
• Session Management: New session creates passcode; Continue session restores by passcode
• Mobile-first Studio shell: Responsive header/actions, flexible panel heights
• GTM Strategy: Define a Go-To-Market strategy with phases (e.g. Awareness, Launch). AI or user can create/update the strategy; tasks and assets can be assigned to phases and viewed in a phase-grouped Strategy Flow.

---

📚 Further Documentation

• PROJECT.md — Full technical architecture, deployment, IAM, and extension guide
• backend/README.md — Python WebSocket bridge setup
• AGENTS.md — Coding standards and workflow details

---

Last Updated: 2026-03-07