Universal Dependency Resolver

Your Python backend needs PyPI. Your React frontend needs npm. Your Rust CLI needs crates.io. Your Go service needs Go modules. One udr lock gives you one lock file for all of them.

# Resolve across ecosystems in one command
udr resolve flask>=2.0 torch@pypi react@^18

# Lock your project across all manifests
udr lock

# Check system + CVE + deprecated packages — 18 ecosystems at once
udr check --cve --deprecated

Install

pip install ud-resolver

# For full capacity — Rust-backed PubGrub + Z3 + richer system data:
pip install "ud-resolver[z3,pubgrub,system]"

The base install resolves dependencies, detects GPU/OS/CPU, and handles GPU variant selection — no extras needed. The extras add speed (Rust PubGrub on large graphs), conflict detection (Z3 for CUDA XOR rules), and richer telemetry (GPU temperature, per-process memory). All solvers fall back gracefully when an extra is missing.

Optional extras

Features

Why UDR?

• Cross-ecosystem resolution: A Python package that transitively depends on an npm package gets solved in one pass, not two.
• SAT-solver engine: Real Z3/PubGrub CDCL solver, not greedy backtracking. Finds valid solutions dependency graph heuristics miss.
• System-aware: GPU type + CUDA version are resolution constraints — torch 2.1.2+cu121 selected automatically when NVIDIA GPU detected; for PyPI torch the pytorch wheel index is consulted to cap and rewrite to the matching +cu<ver> build (e.g. --cuda 12.1 → torch 2.5.1+cu121).
• Supply chain built-in: CVE scanning, license compliance, deprecation checks, lock-file signing (Ed25519), SBOM export (SPDX/CycloneDX), policy engine.
• 3 solver backends: AutoSolver profiles your graph and selects Z3, PubGrub, or Hybrid — with fallback chain if the first choice fails.

Quick Start

# Resolve cross-ecosystem packages
udr resolve numpy pandas scikit-learn
udr resolve react vue -e npm
udr resolve serde tokio -e crates

# Lock a project's dependencies
udr lock
udr lock --manifest requirements.txt --dry-run

# Validate lock file
udr verify

# Show dependency tree
udr graph flask django

# Scan a GitHub repo without cloning
udr scan --github https://github.com/user/repo

# List all supported ecosystems
udr list-ecosystems

🐍 Use as a Python Library

import asyncio
from backend.core.data_aggregator import DataAggregator
from backend.core.system_scanner import SystemScanner
from backend.orchestrator.resolve import create_solver

async def main():
    scanner = SystemScanner()
    system_info = await scanner.scan_all()

    aggregator = DataAggregator()
    info = await aggregator.get_package_info(
        "torch", ecosystem="pypi",
        include_dependencies=True, include_versions=True,
    )

    resolver = create_solver()
    result = resolver.resolve_dependencies(
        packages=[{"name": "flask", "version": ">=2.0"}],
        system_info=system_info,
    )

asyncio.run(main())

🔄 How It Works

Your request ──► Fetch metadata from registry APIs
                      │
                      ▼
              Scan target system (OS, GPU, CUDA, runtimes)
                      │
                      ▼
               Resolve conflicts with AutoSolver (Z3 / PubGrub / Hybrid)
                      │
                      ▼
               Export to 15 formats or write lock file

Links

• GitHub — source, issues, releases
• Documentation — hosted docs site (User Guide, CLI, API, Architecture)
• CLI Reference
• Architecture
• API Docs
• Changelog
• License: MIT