Install the SDK · Docs

1. Install

Shell

pip install resontech

Python 3.11+. Your machine needs only httpx and boto3; PyTorch runs on the workers.

2. Get your keys

Key	Where	Used for
Platform API key	Profile → Developer → Create API key. Starts with `rsk_`, shown once	Every SDK call: submit, list, stop
Bucket keys	Profile → Storage → Provision Bucket. Secret shown once; Rotate Key mints a new pair	Uploading datasets and scripts, reading outputs
Predict key	Returned by `rt_submit_inference()`, one per runtime	Calls to that endpoint, in the `X-API-Key` header

3. Connect and check

setup.py

from resontech import ResonTech, ResonTechConfig

sdk = ResonTech(ResonTechConfig(
    base_url="https://api.beta.reson.tech",
    s3_endpoint="https://s3.beta.reson.tech",
    platform_api_key="rsk_live_...",   # or set RESON_API_KEY
    s3_access_key_id="...",
    s3_secret_access_key="...",
))

print(sdk.worker_stats())               # lists workers: you're connected

There is no login(): each request carries the platform key as a bearer token. A bad, revoked or expired key raises ResonAuthError. Every config field: SDK client config.

If something’s off

Situation	Do this
worker_stats() returns no workers	Your setup works; no public GPUs are free right now. Retry later
Lost the platform key	Revoke it in Profile → Developer and create a new one. Revocation takes effect within seconds
Lost the bucket secret	Profile → Storage → Rotate Key. The old pair stops working at once
Keys in a repo or CI	Omit `platform_api_key` and set `RESON_API_KEY`
Self-hosted (Private Fabric)	Override `base_url` and `s3_endpoint`

Submit training via SDK
Submit inference via SDK
A runnable starting point: job_training_sdk (a small CNN on MNIST)

PreviousWelcome Next Overview