GADE - Gradient-Aware Development Environment

Allocate AI compute dynamically based on code difficulty.

GADE measures difficulty across your codebase and focuses 80% of tokens on the 20% hardest regions.

Quick Start

pip install gade

# Analyze a repository (works on all systems)
python -m gade analyze ./my-project --top 20

# View heatmap
python -m gade heatmap ./my-project

Windows Users: If gade command doesn't work, use python -m gade instead. This is the recommended universal approach.

Cloud API

Use the hosted API — no setup required:

curl -X POST "https://web-production-8b5ca.up.railway.app/analyze" \
  -H "Content-Type: application/json" \
  -d '{"repo_path": "./my-project", "top_k": 10}'

📖 API Docs: web-production-8b5ca.up.railway.app/docs

Features

• 5 Difficulty Signals: Edit churn, complexity, errors, uncertainty, gradient
• 80/20 Allocation: Smart token distribution by difficulty
• Multi-LLM Support: OpenAI, Anthropic, Google, Ollama, Azure, Bedrock
• Agentic AI Ready: MCP server, OpenAI tools, LangChain integration
• REST API: Hosted cloud API or self-host

Installation

pip install gade

Python SDK

from gade import analyze

result = analyze("./my-project")
for node in result.get_top_k(10):
    print(f"{node.node_name}: {node.difficulty_score:.2f}")

CLI Commands

Difficulty Tiers

License

MIT