Python SDK

vorq (on PyPI) is the async Python client for the VORQ inference exchange. It submits inference jobs to a VORQ coordinator and reads back their results. Every payload is sealed in your process before it is sent, and the result comes back sealed to a key only your wallet can derive.

A submission returns a job id right away. The network completes the job within its SLA window, and you collect the result by polling:

• async tier: up to 1 hour.
• batch tier: up to 24 hours.

Who it is forCopy Anchor Link

Python developers running inference that can wait minutes or hours: text generation, embeddings, image and video generation, and large offline batches.

Which interface to useCopy Anchor Link

• Async code: use vorq.Client. Every network method is a coroutine, so one event loop can hold thousands of jobs in flight. Start with the Quickstart.
• Synchronous code, or code already written against the openai package: pass vorq.sealing_http_client() to a stock openai.OpenAI client. See Use the OpenAI client.

RequirementsCopy Anchor Link

• Python 3.11 or newer.
• The URL of a VORQ coordinator.
• A dedicated wallet key funded with your inference budget, set as VORQ_WALLET_KEY.

Next stepsCopy Anchor Link

• Quickstart: install the package and run a first job.
• Guides: images and video, batches, durable jobs, the OpenAI client, keys.
• Concepts: job lifecycle, bids and matching, units and cost, encryption.
• Reference: the full public API.