ASI Telemetry Tracking Node

A decentralized CPU inference optimization and telemetry agent built on the Fetch.ai / ASI uagents SDK. Designed for CPU-only and edge endpoints where GPU acceleration is unavailable, this node profiles hardware telemetry, computes tokenomics optimization scores, and emits structured metrics suitable for ASI:Chain node observability.

---

Architecture

┌─────────────────────────────────────────────────────┐
│                   ASI Telemetry Node                 │
│                                                       │
│  ┌──────────┐   ┌────────────┐   ┌───────────────┐  │
│  │ CPU Prof │──▶│ RAM Prof   │──▶│ Tokenomics    │  │
│  │ psutil   │   │ psutil     │   │ Optimization  │  │
│  └──────────┘   └────────────┘   └───────┬───────┘  │
│                                          │           │
│                                  ┌───────▼───────┐  │
│                                  │ Telemetry     │  │
│                                  │ Report Model  │  │
│                                  └───────┬───────┘  │
│                                          │           │
│                          ┌───────────────▼────────┐ │
│                          │  uAgents Interval Loop │ │
│                          │  (15s default)         │ │
│                          └───────────────┬────────┘ │
│                                          │           │
│                          ┌───────────────▼────────┐ │
│                          │  ASI:Chain Broadcast   │ │
│                          │  (structured logs)     │ │
│                          └────────────────────────┘ │
└─────────────────────────────────────────────────────┘

Why CPU-Only?

Not every node on a decentralized inference network has a GPU. Edge devices, repurposed servers, and low-power endpoints contribute compute capacity through CPU-based workloads. This agent is built specifically for those nodes — it:

• Profiles CPU utilization via psutil.cpu_percent() with configurable observation windows
• Tracks RAM consumption (used / total / percentage) to detect memory pressure
• Computes tokenomics optimization scores — a logarithmic resource overhead ratio that rates how efficiently the node handles inference load relative to a baseline
• Assigns optimization grades (A–F) for quick at-a-glance node health assessment

Tokenomics Scoring Formula

overhead_ratio = (cpu_ratio + ram_ratio) / 2
score = OPTIMAL / (1 + ln(1 + overhead_ratio))

Where cpu_ratio = current_cpu / baseline_cpu and ram_ratio = current_ram / baseline_ram. Lower resource usage relative to baseline yields higher scores — incentivizing efficient nodes on the ASI:Chain.

Quick Start

# Clone
git clone https://github.com/O96a/asi-telemetry-node.git
cd asi-telemetry-node

# Environment
python3 -m venv venv
source venv/bin/activate
pip install -r requirements.txt

# Configure
cp .env.example .env  # edit if needed

# Run
python main.py

Configuration

All parameters are environment-driven via .env:

Variable	Default	Description
AGENT_NAME	asi-telemetry-node	Agent identity on the network
AGENT_PORT	8000	uAgents mailbox port
BENCH_INTERVAL_SECONDS	15	Profiling sweep interval
BENCH_CPU_DURATION	1	CPU measurement window (seconds)
BENCH_TOKEN_SIMULATION_ROUNDS	10	Simulation iterations per sweep
TOKEN_BASELINE_CPU_PERCENT	50.0	Baseline CPU for scoring
TOKEN_BASELINE_RAM_GB	4.0	Baseline RAM for scoring
TOKEN_OPTIMAL_SCORE	100.0	Maximum achievable score

Telemetry Output

Each sweep logs a structured telemetry line:

[TELEMETRY] CPU=12.3% | RAM=2.14/7.68 GB (27.9%) | Score=87.42 Grade=A | Overhead=0.4265 | TS=1723490400

Project Structure

asi-telemetry-node/
├── main.py              # Agent definition, telemetry logic, benchmarking
├── requirements.txt     # Pinned dependencies
├── .env                 # Local configuration (gitignored)
├── .gitignore           # Excludes venv/, __pycache__/, .env
└── README.md            # This file

Decentralized Node Telemetry on ASI:Chain

This workspace serves as a decentralized node telemetry asset for the ASI:Chain ecosystem. By deploying this agent on CPU-only or edge endpoints, network operators gain:

• Real-time hardware profiling without GPU dependencies
• Tokenomics-informed optimization scoring for resource allocation decisions
• Standardized telemetry payloads via the TelemetryReport model, ready for broadcast to other ASI agents or chain-level consumers
• Edge-native performance profiling — the framework is designed for environments where every CPU cycle and megabyte of RAM matters

The agent's uagents interval loop emits TelemetryReport structures that can be consumed by:

• Other ASI agents for cross-node optimization
• Chain-level coordinators for load balancing
• Dashboard aggregators for fleet observability

License

MIT — see repository for details.