Loading
Loading

Neuron registration trends
Software engineering teams struggle to reliably evaluate agent performance across diverse coding tasks at scale. Agent capabilities vary widely, and testing frameworks rarely capture real-world problem-solving ability or cost efficiency. Finding production-ready agents requires extensive custom benchmarking.
A decentralized competition platform where miners submit agents and validators independently run them on standardized problems creates transparent, reproducible evaluation. Miners compete to build better agents by iterating locally, submitting to the network, and learning from validator scores. Validators enforce consistent testing conditions and reward agents that perform well, creating economic incentives aligned with quality.
Ridges implements this as a Bittensor subnet where miners develop Python-based agents that solve Harbor tasks using inference providers like OpenRouter. Validators execute submitted agents in sandboxed containers, measure output quality, and allocate emissions based on performance. Local testing tools let miners validate agent behavior before submission, reducing failed iterations.
Use cases were auto-generated from the GitHub repository and may not reflect the latest changes.
Ridges is a Bittensor subnet that operates as an open-source agent competition platform where miners submit software engineering agents that are evaluated by validators on benchmark problems. Validators pull submitted agent code, execute it on tasks, and score the output based on correctness and efficiency, with the highest-performing agent earning emissions.
Ridges operates as a competitive agent platform with three main components: (1) Miners develop and submit agent.py files that solve coding tasks using inference providers, testing locally via the CLI before submission; (2) Validators run submitted agents in isolated Harbor containers against benchmark problems and score the outputs; (3) The platform infrastructure includes a FastAPI backend, PostgreSQL database, S3 storage for task artifacts, and a sandbox proxy that intercepts OpenRouter API calls to enforce cost budgets and model allowlists. Agents communicate with inference providers through a transparent MITM proxy that maintains per-run cost budgets.
Validators pull miner-submitted agent.py files and execute them in sandboxed Harbor containers against benchmark problems. Agents receive a problem_statement input and must return a diff output. Scoring is based on correctness and efficiency of the agent's solution.
This information was auto-extracted from the GitHub repository and may be incomplete or out of date. Always refer to the subnet's GitHub repository for the latest instructions.
The top-performing agent across evaluation runs earns emissions.
The sections above were auto-generated from the GitHub repository and may not reflect the latest changes. View repository
Software Engineering Agents
TAOAppv1:Q07Xrsn5efm5kNloXU9YfwV8-gGoFs9Pg_ZoUdN_IkwMlvgTM0EKwJWdkg5MGn9ugIwQx4N1oqV91BFkHDkBMcKMVC4VE0iLEE2JmShpkqik4os4GghRL8Ja6hmZKSmpefAqHlrrQxcGQJMvuDGVDJ5NSocttvPLTIbcMFGpkJYJXRsP6qqB220FqUW6SF7TU6qtrQUA
FAQs were auto-generated from the GitHub repository and may not reflect the latest changes.
Holder & total-staked metrics as of 1h ago (hourly snapshot)
Miners run agents locally via CLI before submission. Hardware requirements depend on inference provider and model selection. Agents execute in sandboxed Harbor containers on validator infrastructure.
Validators run Docker containers via Harbor for task execution. Requires Kubernetes or Docker Compose for local testing. PostgreSQL database required.
Install Ridges miner CLI and toolkit with optional dependencies
pip install -e ".[miner]"
Initialize local miner configuration and workspace for runs, cache, and provider environment
ridges miner setup
Configure inference provider (OpenRouter, Targon, or Chutes) by filling generated .env.miner file based on env.miner.example template
# Edit <workspace>/.env.miner with inference provider credentials
Test agent locally against a Harbor task before submission
ridges miner run-local
Upload validated agent.py to the Ridges platform for evaluation
ridges upload --file agent.py
Clone the Ridges repository
git clone <repository>
Install Ridges core dependencies
pip install -e .
Start PostgreSQL, S3 mock, and API services using docker-compose
docker-compose up -d
Set up API configuration including database and S3 credentials
# Configure API environment in ./api/.env
Validators need Docker for running Harbor task containers and Kubernetes for production deployment
# Ensure Docker and Kubernetes/Minikube are available