MisyBot AI

AI-powered assistant for small businesses that creates promotional content, automates campaigns, and engages customers in real time—your creative strategist, support agent, and growth partner in one.

Comment

🧠 About the Project: Misybot

🚀 Inspiration

Misybot was born from a simple but powerful question:
Can businesses access the creative and strategic power of AI without needing a full marketing team?
In Colombia and across Latin America, entrepreneurs run cafés, boutiques, and local services with limited time and resources. We wanted to build an assistant that doesn't just answer questions—it executes, creates, and orchestrates real outcomes.

💡 What It Does

Misybot is an autonomous AI assistant that helps small businesses:

  • 🎥 Generate promotional videos with AI-written scripts
  • 🖼️ Create branded images and synthetic voice content
  • 📅 Schedule posts with calendar optimization
  • 📊 Scan market trends and predict engagement
  • 🧠 Respond to customer FAQs with contextual memory
  • 📈 Deliver real-time and monthly performance reports

It connects a network of modular agents through a central orchestrator, allowing users to launch campaigns, analyze metrics, and engage customers—all from a single chat interface.

🛠️ How We Built It

  • Backend: NestJS, TypeORM, PostgreSQL, Azure Blob Storage, Azure Service Bus
  • Frontend: React with floating chat assistant and dark theme
  • Agents: Modular services with DTO validation, persistent context, and clean logging
  • Meta-agent: Orchestrates multi-agent workflows in real time
  • Security: JWT + OAuth authentication for session-based execution
  • AI Integration: GPT-OSS model accessed via Foundry, embedded in orchestration and content-generation agents

We’re showcasing 5 core agents, but we’ve architected and are actively developing over 40 specialized agents with advanced capabilities like:

  • Predictive trend modeling
  • Engagement forecasting
  • Monthly campaign planning
  • Strategic content optimization

🧠 GPT-OSS Integration

During the hackathon, we integrated the OpenAI gpt-oss model via Foundry into our orchestration layer. This allowed us to shift from static workflows to dynamic, context-aware execution.
The model powers agents like AgentVideoScriptor, AgentTrendScanner, and AgentFaqResponder, enabling them to generate business-specific content, analyze metrics, and respond with strategic tone and relevance.

Due to infrastructure constraints and credential sensitivity, the model is not publicly exposed, but its usage is documented and demonstrated in our agent flows.

⚔️ Challenges We Ran Into

  • Ensuring real-time orchestration without blocking the UI
  • Managing service dependencies (.env for PostgreSQL, OpenAI, Blob, JWT, OAuth) with robust logging
  • Refactoring legacy codebases while preparing for high-stakes demos
  • Navigating legal inconsistencies in insolvency processes and documenting every step
  • Designing a frontend that feels conversational, proactive, and non-intrusive

🏆 Accomplishments That We're Proud Of

  • Built a scalable multi-agent backend with persistent context
  • Designed a proactive chat assistant that offers smart tips and executes tasks
  • Developed a scheduling algorithm that optimizes post timing based on engagement predictions
  • Created a modular architecture that supports over 40 agents in parallel development
  • Delivered a real-time dashboard for campaign metrics and monthly reports
  • Successfully embedded gpt-oss into agent workflows via Foundry

📚 What We Learned

  • How to enforce Clean Code and DTO validation across multi-agent systems
  • The importance of context persistence in AI workflows
  • How to balance UX simplicity with backend complexity
  • Strategic communication and documentation in legal and financial negotiations
  • How to build AI assistants that don’t just talk—they act

🔮 What's Next for Misybot

  • Launching the full suite of 40+ agents with predictive intelligence
  • Integrating monthly engagement calendars and automated campaign planning
  • Expanding to new verticals: retail, education, health, and logistics
  • Building a self-optimizing agent layer that learns from user behavior
  • Offering Misybot as a plug-and-play SaaS for small businesses worldwide

Misybot isn’t just a chatbot—it’s a creative strategist, campaign manager, and growth partner for businesses that deserve more than templates. We built it to empower, not just inform.

Built With

  • axios
  • azure-app-services
  • azure-blob-storage
  • azure-service-bus
  • bcrypt
  • class-validator
  • docker
  • github-actions
  • jwt
  • microservices
  • modular
  • mongodb
  • multi-agent
  • nestjs
  • node.js
  • oauth-2.0
  • openai-gpt-models-(for-content-generation-and-orchestration)
  • postgresql
  • react
  • tailwindcss
  • typeorm
  • typescript
Share this project:

Updates