An experimental AI agent that automatically discovers, categorizes, and visualizes technology ecosystems based on keyword searches.

Ecosystem Mapper takes a keyword (e.g., "agentic AI", "vector databases", "MLOps") and automatically:

1. Discovers recent projects and resources across GitHub and the web
2. Analyzes the landscape to create a taxonomy/categorization
3. Identifies representative examples within each category
4. Generates visual ecosystem maps showing the landscape

Input: User-provided keyword (e.g., "agentic AI")

Data Sources:

• GitHub API: Search for repositories created in the last 3 months matching the keyword
• Tavily Search: Web search to find articles, tools, and resources

Analysis Engine:

• Model: Google Gemini 3 Flash Preview (via OpenRouter)
• Task: Cross-analyze projects and resources to:
  • Create a taxonomy/categorization scheme
  • Identify main categories within the ecosystem
  • Find representative examples for each category
  • Determine relationships and groupings

Output: Structured taxonomy with categorized examples

Input: Taxonomy and categorized examples from Stage 1

Process:

• Use example ecosystem maps as style references
• Generate visual representations of the ecosystem
• Create market map-style visualizations showing:
  • Categories and subcategories
  • Representative projects/companies
  • Relationships and groupings

Output: Ecosystem map images

# Analyze the agentic AI ecosystem
python agent.py --keyword "agentic AI"

# Explore vector database landscape
python agent.py --keyword "vector databases"

# Map out the RAG (Retrieval Augmented Generation) ecosystem
python agent.py --keyword "RAG frameworks"

Set up your .env file (see .env.example):

# OpenRouter Configuration
OR_RESEARCH_MODEL_NAME=google/gemini-3-flash-preview
OR_IMAGE_GEN_MODEL_NAME_1=google/gemini-3-pro-image-preview
OPENROUTER_API_KEY=your_openrouter_api_key_here

# Tavily Search API
TAVILY_API_KEY=your_tavily_api_key_here

# GitHub Personal Access Token
GITHUB_PAT=your_github_personal_access_token_here

Ecosystem-Mapper/
├── agent.py                 # Main orchestration script
├── modules/
│   ├── github_collector.py  # GitHub API integration
│   ├── tavily_search.py     # Tavily search integration
│   ├── taxonomy_analyzer.py # Gemini-powered taxonomy creation
│   └── visualizer.py        # Map generation (Stage 2)
├── example-maps/            # Reference ecosystem maps
├── outputs/                 # Generated taxonomies and maps
└── README.md

• Python 3.10+
• OpenRouter API access (for Gemini)
• GitHub API token
• Tavily API key
• Required packages: requests, python-dotenv, openai (for OpenRouter)

• ✅ Stage 1a: Data collection modules (GitHub + Tavily)
• ✅ Stage 1b: Taxonomy analysis with Gemini
• 🚧 Stage 2: Visualization generation (planned)

The example-maps/ folder contains ecosystem maps from public sources that serve as visual style references for the maps we'll generate.

Private repository - Daniel Rosehill

This is an experimental project exploring automated ecosystem mapping and visualization. The agent uses AI to understand domain landscapes and create structured categorizations automatically.