Infrintia Python SDK

The Infrintia SDK provides ComputeClient and AsyncComputeClient for interacting with the Infrintia GPU compute marketplace from Python.

---

Installation

pip install infrintia

---

ComputeClient

Synchronous client for the Infrintia API.

Initialization

from infrintia import ComputeClient

client = ComputeClient(
    base_url="https://api.infrintia.crossgl.net",
    username="alice",
    password="secure-password-123"
)

Parameter	Type	Required	Description
base_url	str	Yes	Infrintia API base URL
username	str	No	Username for authentication
password	str	No	Password for authentication
api_key	str	No	API key (alternative to username/password)
timeout	int	No	Request timeout in seconds (default: 60)

Methods

create_user(username, email, password)

Create a new user account.

user = client.create_user(
    username="alice",
    email="alice@example.com",
    password="secure-password-123"
)

Returns: dict with user_id, username, email, credits, created_at.

get_me()

Get the authenticated user's profile.

profile = client.get_me()
print(f"Credits: {profile['credits']}")

list_models()

List all available models on the marketplace.

models = client.list_models()
for m in models:
    print(f"{m['name']} — {m['min_price_per_token']} credits/token")

Returns: list[dict] with model metadata.

list_hosts()

List all active GPU hosts.

hosts = client.list_hosts()
for h in hosts:
    print(f"{h['host_id']}: {h['gpu']} ({h['status']})")

run_model(model, input_text, max_tokens=128, temperature=0.7)

Submit an inference job.

job = client.run_model(
    model="meta-llama/Llama-3-8B-Instruct",
    input_text="What is gravity?",
    max_tokens=256,
    temperature=0.7
)
print(f"Job {job['job_id']} — status: {job['status']}")

Returns: dict with job_id, status, model, credits_reserved.

get_job(job_id)

Get the status and details of a specific job.

status = client.get_job("job_def456")

list_jobs(status=None, limit=20, offset=0)

List jobs for the authenticated user.

jobs = client.list_jobs(status="completed", limit=10)

stream_job(job_id)

Stream results from a running job. Yields tokens as they are generated.

for token in client.stream_job("job_def456"):
    print(token, end="", flush=True)

Yields: str — individual tokens.

get_metrics()

Get platform-wide metrics.

metrics = client.get_metrics()
print(f"Active hosts: {metrics['active_hosts']}")

---

AsyncComputeClient

Async version of ComputeClient for use with asyncio.

Initialization

from infrintia import AsyncComputeClient

client = AsyncComputeClient(
    base_url="https://api.infrintia.crossgl.net",
    username="alice",
    password="secure-password-123"
)

Async Methods

All methods mirror ComputeClient but are async:

import asyncio
from infrintia import AsyncComputeClient

async def main():
    client = AsyncComputeClient(
        base_url="https://api.infrintia.crossgl.net",
        username="alice",
        password="secure-password-123"
    )

    job = await client.run_model(
        model="meta-llama/Llama-3-8B-Instruct",
        input_text="Write a poem about silicon.",
        max_tokens=128
    )

    async for token in client.stream_job(job["job_id"]):
        print(token, end="", flush=True)
    print()

asyncio.run(main())

Available Async Methods

Method	Description
await create_user(...)	Create user account
await get_me()	Get authenticated user profile
await list_models()	List available models
await list_hosts()	List GPU hosts
await run_model(...)	Submit inference job
await get_job(job_id)	Get job status
await list_jobs(...)	List user's jobs
async for token in stream_job(job_id)	Stream job results
await get_metrics()	Get platform metrics