Skip to content

Quick Start

Get SolanaLM running in under 5 minutes.

One-Command Setup

The fastest way to start is using the quick start script:

python scripts/quick_start.py

This automatically:

  • Starts the gateway server
  • Launches a demo inference node
  • Runs basic connectivity tests

Manual Setup

For more control, start components individually:

Step 1: Start the Gateway

The gateway is the central API server that routes requests to nodes:

python scripts/run_gateway.py

You should see:

INFO: Gateway starting on http://localhost:8001
INFO: Node registry initialized
INFO: Training coordinator initialized
INFO: Gateway ready to accept connections

Step 2: Start an Inference Node

In a new terminal, start an inference node:

python scripts/run_node.py \
    --node-type inference \
    --node-id my-node-1 \
    --wallet YOUR_WALLET_ADDRESS

The node will:

  1. Initialize the local model
  2. Register with the gateway
  3. Start accepting inference requests

Step 3: Test the System

Run the end-to-end test suite:

python scripts/test_end_to_end.py

Or test manually with the Python SDK:

import asyncio
from client.python.solanalm_client import SolanaLMClient

async def test():
    async with SolanaLMClient("http://localhost:8001") as client:
        # Check network status
        status = await client.get_network_status()
        print(f"Network: {status}")

        # List available models
        models = await client.list_available_models()
        print(f"Models: {models}")

        # Run inference
        if models:
            response = await client.inference(
                model=models[0],
                prompt="Hello, how are you?",
                wallet_address="test-wallet"
            )
            print(f"Response: {response.response}")

asyncio.run(test())

Access Points

Once running, access these endpoints:

Service URL Description
Gateway API http://localhost:8001 Main API endpoint
Health Check http://localhost:8001/health System health status
Metrics http://localhost:8001/metrics Prometheus metrics
Web Dashboard http://localhost:5173 Vue.js monitoring UI

Monitoring Options

Web Dashboard

Start the Vue.js dashboard for visual monitoring:

cd frontend
npm install && npm run dev

Access at http://localhost:5173

Terminal TUI

Use the terminal-based dashboard for command-line monitoring:

# Monitor a running node
python -m core.tui --node-url http://localhost:8100

# Use light theme
python -m core.tui --theme light

Keyboard shortcuts: 1-4 switch tabs, r refresh, q quit

Example: Complete Inference Flow

import asyncio
from client.python.solanalm_client import SolanaLMClient

async def main():
    async with SolanaLMClient("http://localhost:8001") as client:

        # 1. Check network status
        status = await client.get_network_status()
        print(f"Service: {status.get('service')}")
        print(f"Active nodes: {status.get('network_stats', {}).get('active_nodes', 0)}")

        # 2. List available models
        models = await client.list_available_models()
        print(f"\nAvailable models: {models}")

        # 3. Run inference
        response = await client.inference(
            model="microsoft/DialoGPT-small",
            prompt="What is the meaning of life?",
            wallet_address="my-wallet-address",
            max_tokens=100,
            temperature=0.7
        )

        print(f"\nResponse: {response.response}")
        print(f"Tokens: {response.tokens_generated}")
        print(f"Cost: {response.cost_sol} SOL")
        print(f"Latency: {response.processing_time}s")

asyncio.run(main())

Running Multiple Nodes

Scale your local network with multiple nodes:

python scripts/run_gateway.py

python scripts/run_node.py \
    --node-type inference \
    --node-id inference-1 \
    --wallet wallet-1 \
    --port 8100

python scripts/run_node.py \
    --node-type inference \
    --node-id inference-2 \
    --wallet wallet-2 \
    --port 8101

python scripts/run_node.py \
    --node-type training \
    --node-id training-1 \
    --wallet wallet-3 \
    --port 8200

Using the OpenAI-Compatible API

SolanaLM provides an OpenAI-compatible endpoint:

from openai import OpenAI

client = OpenAI(
    base_url="http://localhost:8001/v1",
    api_key="your-solana-wallet-address"
)

response = client.chat.completions.create(
    model="microsoft/DialoGPT-small",
    messages=[
        {"role": "user", "content": "Hello!"}
    ]
)

print(response.choices[0].message.content)

What's Next?

Now that you have SolanaLM running: