Job Types (Environments)¶

Job types are pre-configured execution environments with specific packages and resource limits.

Why Job Types?¶

Installing packages at runtime is slow. Job types pre-install packages so execution is fast:

WITHOUT Job Types (slower)          WITH Job Types (fast)
──────────────────────────          ─────────────────────
Start container                     Start container (pre-built)
uv pip install numpy pandas         Execute code
Execute code                        Return results
Return results

Defining Job Types¶

Define job types in your tako_vm.yaml:

job_types:
  - name: data-processing
    requirements:
      - pandas
      - numpy
    memory_limit: "1g"
    cpu_limit: 2.0
    timeout: 60

  - name: ml-inference
    requirements:
      - numpy
      - scikit-learn
      - joblib
    memory_limit: "2g"
    cpu_limit: 2.0
    timeout: 120

  - name: api-client
    requirements:
      - requests
      - httpx
    memory_limit: "256m"
    timeout: 30
    network_enabled: true

Using a Job Type¶

Specify the job_type parameter in your request:

import requests

response = requests.post("http://localhost:8000/execute", json={
    "code": '''
import json
import numpy as np

with open("/input/data.json") as f:
    data = json.load(f)

values = np.array(data["values"])
result = {"mean": float(np.mean(values)), "std": float(np.std(values))}

with open("/output/result.json", "w") as f:
    json.dump(result, f)
''',
    "input_data": {"values": [1, 2, 3, 4, 5]},
    "job_type": "data-processing"
})

print(response.json()["output"])
# {'mean': 3.0, 'std': 1.414...}

Job Type Options¶

Field	Description	Default
`name`	Unique identifier	required
`requirements`	pip packages to install	`[]`
`python_version`	Python version	`"3.11"`
`base_image`	Custom Docker base image	`python:{version}-slim`
`memory_limit`	Container memory limit	`"512m"`
`cpu_limit`	CPU cores	`1.0`
`timeout`	Default timeout (seconds)	`30`
`network_enabled`	Allow outbound network	`false`
`environment`	Environment variables	`{}`
`shared_code`	Python files to include	`[]`

Network Access¶

By default, containers have no network access for security. To enable network for specific job types:

job_types:
  - name: api-caller
    requirements:
      - requests
    network_enabled: true           # Enable network

When network_enabled: true, containers can access any external host. For strict egress control in production, use external firewalls or Kubernetes NetworkPolicy.

List Job Types¶

response = requests.get("http://localhost:8000/job-types")

for jt in response.json():
    print(f"{jt['name']}:")
    print(f"  Packages: {jt['requirements']}")
    print(f"  Memory: {jt['memory_limit']}")
    print(f"  Network: {'yes' if jt.get('network_enabled') else 'no'}")

Building Images¶

In development mode, images are auto-built on first use. To pre-build:

# Build images via the REST API (requires a running server)
curl -X POST http://localhost:8000/job-types/data-processing/build

CLI support for building images is planned — see GitHub #30.

Images are named tako-vm-{name}:latest.

Programmatic Registration¶

You can also register job types in code:

from tako_vm.job_types import JobType, JobTypeRegistry

registry = JobTypeRegistry()

registry.register(JobType(
    name="custom-env",
    requirements=["numpy", "pandas>=2.0"],
    memory_limit="2g",
    cpu_limit=2.0,
    timeout=60,
    environment={
        "NUMBA_CACHE_DIR": "/tmp",
        "OMP_NUM_THREADS": "2"
    },
))

Production Mode¶

In production mode (production_mode: true):

Auto-build is disabled
All job types must be pre-built
Requests for missing job types fail with an error

production_mode: true

This ensures consistent, tested images in production.