Skip to content

TECHNICAL_GUIDE.md

Latest commit

 

History

History
189 lines (143 loc) · 6.2 KB

TECHNICAL_GUIDE.md

File metadata and controls

189 lines (143 loc) · 6.2 KB

Intelligent Content Generation Accelerator

A multimodal content generation solution for retail marketing campaigns using Microsoft Agent Framework with HandoffBuilder orchestration. The system interprets creative briefs and generates compliant marketing content (text and images) grounded in enterprise product data, brand guidelines, and product images.

Overview

This accelerator provides an internal chatbot that can:

  • Interpret Creative Briefs: Parse free-text creative briefs into structured fields (overview, objectives, target audience, key message, tone/style, deliverable, timelines, visual guidelines, CTA)
  • Generate Multimodal Content: Create marketing copy and images using GPT models (text and image generation models)
  • Ensure Brand Compliance: Validate all content against brand guidelines with severity-categorized warnings
  • Ground in Enterprise Data: Leverage product information, product images, and brand guidelines stored in Azure services

Architecture

Backend

  • Runtime: Python 3.11 + Quart + Hypercorn
  • Deployment: Azure Container Instance (ACI) in private VNet
  • Authentication: System-assigned Managed Identity

Frontend

  • Framework: React + Vite + TypeScript + Fluent UI
  • Deployment: Azure App Service with Node.js proxy
  • Features: Server-Sent Events (SSE) for streaming responses

Specialized Agents (Microsoft Agent Framework)

The solution uses HandoffBuilder orchestration with 6 specialized agents:

Agent Role
TriageAgent Coordinator that routes user requests to appropriate specialists
PlanningAgent Parses creative briefs, develops content strategy, returns for user confirmation
ResearchAgent Retrieves products from CosmosDB, fetches brand guidelines, assembles grounding data
TextContentAgent Generates marketing copy (headlines, body, CTAs) using GPT
ImageContentAgent Creates marketing images via GPT image models with product context
ComplianceAgent Validates content against brand guidelines, categorizes violations

Compliance Severity Levels

Level Description Action
Error Legal/regulatory violations Blocks acceptance until modified
Warning Brand guideline deviations Review recommended
Info Style suggestions Optional improvements

Azure Services

Service Purpose
Azure OpenAI (GPT) Text generation and content creation
Azure OpenAI (GPT Image) Image generation (can be separate resource)
Azure Cosmos DB Products catalog, chat conversations
Azure Blob Storage Product images, generated images
Azure Container Instance Backend API hosting
Azure App Service Frontend hosting
Azure Container Registry Container image storage

Creative Brief Fields

The system extracts the following fields from free-text creative briefs:

  1. Overview - Campaign summary
  2. Objectives - Goals and KPIs
  3. Target Audience - Demographics and psychographics
  4. Key Message - Core messaging
  5. Tone and Style - Voice and manner
  6. Deliverable - Expected outputs
  7. Timelines - Due dates and milestones
  8. Visual Guidelines - Image requirements
  9. CTA - Call to action

Product Schema

{
  "product_name": "string",
  "category": "string",
  "sub_category": "string",
  "marketing_description": "string",
  "detailed_spec_description": "string",
  "sku": "string",
  "model": "string",
  "image_description": "string (auto-generated via GPT-5 vision)",
  "image_url": "string"
}

Getting Started

Prerequisites

  • Azure subscription with access to:
    • Azure OpenAI (GPT model - GPT-4 or higher recommended)
    • Azure OpenAI (GPT Image model - can be same or different resource)
    • Azure Cosmos DB
    • Azure Blob Storage
    • Azure Container Instance
    • Azure App Service
    • Azure Container Registry
  • Azure CLI >= 2.50.0
  • Docker (optional - ACR can build containers)
  • Python 3.11+
  • Node.js 18+

Quick Deployment (Recommended)

Using Azure Developer CLI (azd):

# Clone the repository
git clone <repository-url>
cd content-generation-solution-accelerator

# Deploy everything with one command
azd up

See AZD_DEPLOYMENT.md for detailed azd deployment instructions.

Manual Deployment

For more control over individual resources:

# Clone the repository
git clone <repository-url>
cd content-generation-solution-accelerator

# Run deployment script
./scripts/deploy.sh

See DEPLOYMENT.md for detailed manual deployment instructions.

Local Development

# Backend
cd src/backend
pip install -r requirements.txt
python app.py

# Frontend
cd src/app/frontend
npm install
npm run dev

Configuration

Environment Variables

See src/backend/settings.py for all configuration options. Key settings:

Variable Description
AZURE_OPENAI_ENDPOINT Azure OpenAI endpoint for GPT model
AZURE_ENV_GPT_MODEL_NAME GPT model deployment name
AZURE_OPENAI_GPT_IMAGE_ENDPOINT Azure OpenAI endpoint for GPT image model (if separate)
AZURE_ENV_IMAGE_MODEL_NAME GPT image model deployment name (gpt-image-1-mini)
AZURE_COSMOS_ENDPOINT Azure Cosmos DB endpoint
AZURE_COSMOS_DATABASE_NAME Cosmos DB database name
AZURE_BLOB_ACCOUNT_NAME Storage account name
BRAND_* Brand guideline parameters

Brand Guidelines

Brand guidelines are configured via environment variables with the BRAND_ prefix:

BRAND_TONE=Professional yet approachable
BRAND_VOICE=Innovative, trustworthy, customer-focused
BRAND_PROHIBITED_WORDS=guarantee,best,only,exclusive
BRAND_REQUIRED_DISCLOSURES=Terms apply,See details
BRAND_PRIMARY_COLOR=#0078D4
BRAND_SECONDARY_COLOR=#107C10

Documentation

License

MIT License - See LICENSE for details.