Instill Core Python SDK

Overview

Welcome to Instill Python SDK, where the world of AI-first application comes alive in the form of Python.

Before you jump into creating your first application with this SDK tool, we recommend you to get familiar with the core concepts of Instill Product first. You can check out our documentation here:

• Instill Core
• Instill SDK

Setup

Requirements

• Python 3.9 - 3.12
• CUDA 12.2 - 12.8 (if you want to build a model with GPU support)

Installation

GRPC_PYTHON_LDFLAGS=" -framework CoreFoundation" pip install grpcio --no-binary :all:

Install it directly into an activated virtual environment:

pip install instill-sdk

or add it to your Poetry project:

poetry add instill-sdk

Check import

After installation, you can check if it has been installed correctly:

python
>>> import instill
>>> instill.__version__

Config Instill Core or Managed Instill Core instance

Before we can start using this SDK, you will need to properly config your target instance. We support two ways to setup the configs, which are

Config file

create a config file under this path ${HOME}/.config/instill/sdk/python/config.yml, and within that path you will need to fill in some basic parameters for your desired host.¹

Within the config file, you can define multiple instances with the alias of your liking, later in the SDK you can refer to this alias to switch between instances.²

hosts:
  alias1:
    url:    str
    secure: bool
    token:  str
  alias2:
    url:    str
    secure: bool
    token:  str
  ...
  ...

Example:

hosts:
  default:
    url: localhost:8080
    secure: false
    token: instill_sk***
  cloud:
    url: api.instill-ai.com
    secure: true
    token: instill_sk***

At runtime

If you do not like the idea of having to create a config file, you can also setup your target instance by doing the following at the very beginning of your script.

from instill.config import global_config

global_config.set_default(
    url="api.instill-ai.com",
    token="instill_sk***",
    secure=True,
)

Usage

Before we get into this, please make sure a local instance of Instill VDP and Instill Model is running, and the config file had been populated with the correct url and api_token

Let's get started!

Import packages

To Form a pipeline, it required a start operator and a end operator, we have helper functions to create both

from instill.clients import InstillClient

Get the client

Get the unified client that connect to all the available services offered by Instill VDP and Instill Model, including

• mgmt_service
• pipeline_service
• model_service
• artifact_service

client = InstillClient()

user = client.mgmt_service.get_user()
# name: "users/admin"
# uid: "4767b74d-640a-4cdf-9c6d-7bb0e36098a0"
# id: "admin"
# ...
# ...

Please find more usages for this SDK in our documentation

You can also find some notebook examples in our notebooks directory

Create a model

Now create a model text-generation in Instill Model for later use

import instill.protogen.common.task.v1alpha.task_pb2 as task_interface
model_id = "model_text-generation"
client.model_service.create_model(
    model_id,
    task_interface.Task.TASK_TEXT_GENERATION,
    "REGION_GCP_EUROPE_WEST4",
    "CPU",
    "model-definitions/container",
    {},
)

Build and deploy the model

Instill Model is an advanced MLOps/LLMOps platform that was specifically crafted to facilitate the efficient management and orchestration of model deployments for unstructured data ETL. With Instill Model, you can easily create, manage, and deploy your own custom models with ease in Instill Core or on the cloud with Instill Cloud.

Follow the model creation guide to build and deploy your model.

Create pipeline

In the section we will be creating a pipeline using this python-sdk to harness the power of Instill VDP!

The pipeline receipt below is a sample for demo. It simply returns the input string value.

pipeline_id = "pipeline_demo"
client.pipeline_service.create_pipeline(
    pipeline_id,
    "this is a pipeline for demo",
    {
        "output": {"result": {"title": "result", "value": "${variable.input}"}},
        "variable": {"input": {"instillFormat": "string", "title": "input"}},
    },
)

Validate the pipeline

Before we trigger the pipeline, it is recommended to first validate the pipeline recipe first

# validate the pipeline recipe
client.pipeline_service.validate_pipeline(pipeline_id)

Trigger the pipeline

Finally the pipeline is done, now let us test it by triggering it!

# we can trigger the pipeline now
client.pipeline_service.trigger_pipeline(pipeline_id, [], [{"input": "hello world"}])

And the output should be exactly the same as your input.

Contributing

Please refer to the Contributing Guidelines for more details.

Community support

Please refer to the community repository.

License

See the LICENSE file for licensing information.

Footnotes

1. You can obtain an api_token by simply going to Settings > API Tokens page from the console, no matter it is Instill Core or Managed Instill Core. ↩

2. SDK is default to look for instance named default first, and will fall back to the first instance entry in the config file if default not found ↩