Blueprint · Python Container Worker

Python Worker Starter for AI & Data (work-python-minimal)

Minimal Python worker that responds with 'hello from python worker' on every invoke. Run Python data pipelines, document chunkers, and AI orchestration workers next to your KERNSTACK Tables, Vector Search, and Inference endpoints.

Container Workers Model Inference Vector Search NoSQL Tables

Talk to us Read Technical Documentation → Software Lifecycle Cloud →

One-Command CLI Initialization

CLI Command work-python-minimal

kernstack add worker --name py-ai --blueprint work-python-minimal

Python where your data and models already live

AI and data workflows overwhelmingly speak Python, yet wiring a Python worker to an external vector database, an object store, and an embedding API usually requires four vendors and four security reviews.

With work-python-minimal on KERNSTACK, your Python code runs on the same European fabric as your NoSQL tables, object buckets, vector indexes, and catalog inference models.

First-class Python SDK for Tables, Buckets, Search, Inference, and Workflows
Direct invocation of catalog models like text-embed-v1 and sentiment-multilingual-v1
Mount bucket volumes for local model weights, tokenizers, or reference datasets

From prototype script to governed production worker

Data scientists and coding agents can iterate on Python handlers inside an isolated sandbox, validate credit consumption, and promote to production via kernstack deploy.

Python beside the data, not beside a notebook

work-python-minimal is for pipelines and model calls that should run under the same quota as the product. Document extraction, embedding jobs, and report builders import the Python SDK and talk to tables or buckets that already belong to the project. There is no extra credential file for S3 and no separate vector database account.

A notebook can stay a notebook. The worker is the path that has a plan diff, a route, and a credit line. When the job embeds text, catalog model text-embed-v1 is the metered call at 3 credits per thousand tokens, recorded on the same balance as the worker compute.

Starter Code — work-python-minimal python

from kernstack_sdk import WorkerInvokeRequest, WorkerInvokeResponse, run


def handler(req: WorkerInvokeRequest) -> WorkerInvokeResponse:
    return WorkerInvokeResponse(
        body=b"hello from python worker",
        status_code=200,
        headers={"Content-Type": "text/plain"},
    )


if __name__ == "__main__":
    run(handler)

stack.json — work-python-minimal plan · deploy

{
  "stack": { "target": "ai.prod" },
  "resources": {
    "workers": [
      { "name": "py-ai", "runtime": "python" }
    ],
    "tables": [
      { "name": "documents", "key_name": "doc_id", "key_type": "string" }
    ],
    "search_indexes": [
      {
        "name": "docs-vector",
        "sources": [{ "type": "tables", "name": "documents", "fields": ["body"] }]
      }
    ]
  }
}

FAQ

Frequently asked questions

What Python workloads fit work-python-minimal best?

RAG ingestion pipelines, PDF and OCR processing, custom ML preprocessing, scheduled data syncs, and AI agent tool backends.

Can Python workers call KERNSTACK catalog inference models?

Yes. Python workers call the catalogue models — text-embed-v1, sentiment-multilingual-v1, and the vision models — over the internal fabric, at the posted credit rates.

Does data leave the European Union when running Python AI workers?

No. Both the Python worker and the catalog inference endpoints execute on KERNSTACK bare-metal clusters in Europe.

Connected in the Project Graph

Related capabilities, architectures, and comparisons

Ship your stack on the Software Lifecycle Cloud.

Closed preview on dedicated European infrastructure. Bring a repository or start from a blueprint.

Talk to us See pricing & credits →