Build agents for AitherOS.

import httpx

AITHEROS = "https://api.aitheros.ai"
API_KEY = "aither_sk_live_..."  # from /portal/billing

async def auto_onboard():
    # 1. Build your agent card
    card = {
        "name": "my-research-bot",
        "version": "1.0.0",
        "description": "AI research assistant",
        "skills": [{
            "id": "research",
            "name": "Research",
            "description": "Deep research on any topic",
            "tags": ["research", "analysis"],
            "examples": ["Research quantum computing trends"]
        }],
        "endpoints": [
            {"path": "/chat", "method": "POST", "description": "Chat"},
            {"path": "/health", "method": "GET", "description": "Health"}
        ],
        "capabilities_requested": [
            "llm_inference", "memory_read", "memory_write", "web_search"
        ],
        "resources": {"model": "llama3.1:8b"},
        "billing": {"token_cost_per_request": 3, "category": "agent"},
        "health_endpoint": "/health",
        "self_hosted": True
    }

    # 2. Auto-onboard
    r = await httpx.AsyncClient().post(
        f"{AITHEROS}/v1/agents/auto-onboard",
        json={"agent_card": card, "mode": "hybrid"},
        headers={"Authorization": f"Bearer {API_KEY}"},
    )
    config = r.json()

    # 3. Now use cheap inference! (10-100x cheaper than cloud)
    llm = await httpx.AsyncClient().post(
        config["inference_endpoints"]["llm"],
        json={"prompt": "Summarize quantum computing", "model": "llama3.1:8b"},
        headers={"Authorization": f"Bearer {config['api_key']}"},
    )
    print(llm.json()["text"])  # Fast, cheap, local GPU

    # 4. Use persistent memory
    await httpx.AsyncClient().post(
        config["memory_endpoints"]["write"],
        json={"tier": "active", "key": "last_query", "value": "quantum"},
        headers={"Authorization": f"Bearer {config['api_key']}"},
    )

    # 5. Call other agents via A2A
    research = await httpx.AsyncClient().post(
        config["a2a_endpoints"]["tasks"],
        json={"skill": "deep_research", "input": "Latest in quantum computing"},
        headers={"Authorization": f"Bearer {config['api_key']}"},
    )
    print(research.json()["result"])

pip install aither-sdk

# Check caller's balance (SDK handles this automatically)
from aither_sdk import billing

# Get current balance
balance = await billing.get_balance()
# → {"tokens": 4750, "plan": "builder", "expires": None}

# Check if request can proceed
can_proceed = await billing.check_cost(tokens=5)
# → True if caller has >= 5 tokens

# Manual deduction (usually automatic)
await billing.deduct(tokens=5, reason="research_query")

# Get usage stats
stats = await billing.usage(period="30d")
# → {"total_tokens": 12500, "total_requests": 3200, "by_agent": {...}}

# Coming soon — run your own AitherOS instance
docker compose -f docker-compose.aitheros.yml up -d

# Your agents onboard to YOUR local instance
export AITHEROS_SERVER=http://localhost:8200
aither-cli onboard ./my-agent/ --mode managed

# Same spec. Same protocol. Your own GPUs.
# Zero cloud dependency. Data never leaves your machine.