rmpca - Route Optimization Engine

High-performance route optimization for road networks. Solve Chinese Postman Problem (CPP) and Vehicle Routing Problem (VRP).

Features

• CPP Solvers: Internal Rust solver + external `rust-optimizer` (5-8x faster, 20-60% better efficiency)
• VRP Solvers: Greedy, Clarke-Wright, Sweep, Two-Opt, OR-Tools, Neural
• Data Sources: Overture Maps, OSM PBF, PMTiles, PostGIS
• Binary Format: `.rmp` - compact binary road network storage (~90% compression)
• MCP Server: 25+ tools for AI agent integration
• ML Features: Solver selection, quality prediction, graph embeddings

Installation

cargo install v2rmp

For better CPP results, install the external optimizer:

cargo install rust-optimizer

Quick Start

# Extract road network
rmpca extract --source overture --bbox "-73.6,45.5,-73.5,45.6" --output roads.geojson

# Compile to binary
rmpca compile --input roads.geojson --output map.rmp

# Optimize route (use external engine for best results)
rmpca optimize --input map.rmp --output route.gpx --cpp-engine external-rust-optimizer

# Or use the full pipeline
rmpca pipeline --bbox "..." --output-prefix my_route

CLI Commands

MCP Server

Connect AI agents to v2rmp with 25+ tools:

{
  "mcpServers": {
    "rmpca": {
      "command": "rmpca",
      "args": ["serve"]
    }
  }
}

Tools: `extract_overture`, `compile`, `optimize`, `v2rmp_rust_optimizer`, `vrp_solve`, `partition`, `list_r2_bucket`, etc.

🤖 Offline AI Agent

`v2rmp` features a specialized, 100% offline LLM agent designed to orchestrate the route optimization pipeline. Built on a fine-tuned 1.5B parameter model, it translates natural language routing requests directly into tool commands—without needing internet access or cloud APIs.

# Pull and run the local agent via Ollama
ollama run spacialglaciercom/v2rmp-agent

For more info, visit the Ollama Hub page.

Performance

License

MIT OR Apache-2.0