Gaia Research

PRD: Gaia Research x Gaia Skill Tree Consolidation

1. Executive Summary

Objective: Unify the energetic, open-science brand of “Gaia Research” with the prestigious, evidence-backed product of “Gaia Skill Tree” into a single, cohesive narrative and web presence before official launch. The Dilemma: The current lab mascot (Milim) is chaotic and fun, while the Skill Tree (The Atlas) is a solemn, ceremonial ledger. The Solution: Treat this dichotomy as a storytelling feature rather than a bug. Convert the Skill Tree repository into a lightweight, headless developer toolkit (“Gaia-Lite”) and rebuild the UI in a modern React framework within the Gaia Research repository.

2. Brand & Narrative Architecture

We will maintain a dual-brand strategy linked by narrative.

• Gaia Research (The Lab): The high-energy, chaotic-but-brilliant collective. Hosted by Milim (Chief Capability Scout). This acts as the DevRel face, appealing to builders who want to break things and push boundaries.
• Gaia Skill Tree (The Atlas): The permanent, ceremonial artifact produced by the lab. It is the solemn ledger where the lab’s discoveries are permanently recorded.
• The Bridge: Milim and the explorers “scout” the wild frontier of AI capabilities, and when a capability is verified, it is formally written into The Atlas.

3. Technical Architecture (The Decoupling)

Currently, gaia-skill-tree is a monolith containing developer tools and static HTML generation scripts. We will decouple these into two distinct surfaces:

A. The Web Surface (gaia-research repo)

• Stack: Modern React framework (Next.js/Remix).
• Purpose: The single deployed website (e.g., gaiaskilltree.com or future domain).
• Structure:
  • / (Home): The Milim-hosted Research Lab. High energy, blogs, open-science manifestos.
  • /atlas (The Registry): The Skill Tree UI. The design strictly shifts to the dark, serious, Honor Red/Apex Gold ledger.
• Data Fetching: Ingests registry.json directly from the core package to render the DAG and leaderboards dynamically, eliminating the need for Python HTML generators.

B. The Core Developer Package (gaia-skill-tree repo)

• Stack: Python / Node.
• Purpose: A lightweight, headless toolkit (“Gaia-Lite”).
• Contents:
  • registry.json (The ground-truth database).
  • gaia_cli (For interacting with the registry).
  • mcp (The Model Context Protocol server).
  • Python/Node SDKs.
• To be Deprecated: All build_docs.py, generateProfilePages.py, and custom static HTML scripts will be deleted.

4. Phased Execution Plan (Post-Current Sprints)

Phase 1: Freeze & Prepare

• Continue current sprints without disrupting the existing Python-based gaia-skill-tree deployment.
• Finalize the React component library in gaia-research for the Milim brand.

Phase 2: The “Gaia-Lite” Extraction

• Create a branch in gaia-skill-tree to safely delete the HTML generation scripts.
• Ensure the CLI, MCP, and SDKs run flawlessly in isolation, relying purely on registry.json.

Phase 3: React Atlas Rebuild

• Scaffold the React application in gaia-research.
• Build the registry.json parser in TypeScript.
• Recreate the Skill Tree visual elements (badges, trees, leaderboards) as React components.

Phase 4: Launch & Route Consolidation

• Deploy the unified React app.
• Point gaiaskilltree.com (or the new dev domain) to the unified gaia-research deployment.
• Publish the stripped-down gaia-skill-tree core to package managers (PyPI/npm).