egg&geese

Autonomous marketing agents ("Geese") that identify high-value social conversations ("Nests") and deposit authentic, persona-driven content ("Eggs") that hatch into viral engagement.

Comment

Inspiration I'm more convinced to buy a product when it's recommended in Reddit instead of seeing it being the promoted product on Amazon. Visibility determines a lot of things. The inspiration for egg&geese came from witnessing the disconnect between authentic community engagement and traditional marketing. We saw countless businesses struggling to connect genuinely with their target audiences on social platforms, often resorting to spammy, inauthentic tactics that damage their brand. We wanted to create a system that combines AI intelligence with human authenticity - allowing businesses to participate in conversations naturally, providing real value while subtly introducing their products where genuinely relevant.

What it does

egg&geese is an autonomous marketing ecosystem that deploys AI-powered "geese" personas to engage authentically in Reddit communities. Each goose is a customizable persona trained on your product, with its own tone, voice, and personality. The geese autonomously find relevant conversations, craft authentic comments that add real value, and naturally mention your product when appropriate. Each comment is an "egg" - a potential customer touchpoint. The system uses Claude Sonnet 4.5 for natural language generation, Parallel AI for intelligent URL content extraction, and Redis for maintaining conversational memory, ensuring each goose stays consistent and authentic across interactions.

How I built it

The stack combines cutting-edge AI with robust infrastructure:

  • Backend: FastAPI (Python) with Reddit OAuth via PRAW for authentic posting
  • Frontend: Next.js 14 with TypeScript and Tailwind CSS for a beautiful, responsive UI
  • AI Layer: Anthropic Claude Sonnet 4.5 and Google Gemini Pro for conversational intelligence
  • Content Intelligence: Parallel AI SDK for extracting product information from URLs
  • Memory Layer: Redis for maintaining goose conversation history and context
  • CMS: Sanity for structured content management
  • Authentication: Reddit OAuth 2.0 flow for secure posting capabilities

The architecture allows each goose to operate autonomously while maintaining brand consistency and authentic communication patterns.

Challenges I ran into

  1. API Integration Complexity: Integrating multiple AI APIs (Claude, Gemini, Parallel) each with different authentication schemes and response formats required careful error handling and fallback strategies.
  2. React Hydration Issues: Browser extensions like Grammarly were injecting attributes into the DOM, causing hydration mismatches that required strategic use of suppressHydrationWarning.
  3. Product Persistence: Initially products weren't persisting because the backend lacked a storage mechanism - we implemented in-memory storage as a temporary solution before full Sanity integration.
  4. Parallel API Documentation: The Parallel AI API required specific headers (parallel-beta) that weren't clearly documented, requiring trial-and-error debugging.
  5. Type Safety Across Boundaries: Ensuring type safety between the Python backend and TypeScript frontend required careful interface definitions and optional field handling.

What's next for egg&geese

  1. Full Sanity CMS Integration: Complete the integration for persistent storage of products, geese, and eggs
  2. Advanced Conversation Matching: Implement semantic search to find the most relevant Reddit threads for each goose
  3. Multi-Platform Support: Expand beyond Reddit to Twitter, Discord, and LinkedIn
  4. Analytics Dashboard: Build comprehensive analytics showing egg performance, engagement rates, and conversion tracking
  5. A/B Testing for Geese: Allow users to deploy multiple goose variants and automatically optimize based on engagement
  6. Community Guidelines Compliance: Add AI-powered checks to ensure all comments comply with subreddit rules and Reddit's terms of service
  7. Collaboration Features: Enable teams to manage multiple geese with role-based access control
  8. Mobile App: Build native iOS/Android apps for managing geese on the go

Built with

  • Frontend: Next.js 14, React, TypeScript, Tailwind CSS
  • Backend: Python, FastAPI, Pydantic
  • AI/ML: Anthropic Claude Sonnet 4.5, Google Gemini Pro, Parallel AI SDK
  • Database/Storage: Redis, Sanity CMS
  • APIs: Reddit API (PRAW), Reddit OAuth 2.0
  • Deployment: Uvicorn (ASGI server)
  • Development Tools: Python dotenv, httpx (async HTTP client)

Built With

  • ai
  • anthropic
  • claude
  • fal
  • lightpanda
  • mcptotal.ai
  • parallel
  • postman
  • redisvl
  • sanity
  • skyflow
Share this project:

Updates