Onboarding assets

aiXplain gives you full control and flexibility to bring your own models, tools, datasets, and utilities into the platform. This avoids vendor lock-in and allows you to monetize your own AI components.

All onboarded assets are deployed as private assets by default. You can request to make them public if you wish to monetize them. Public assets allow you to set your own price. See the Monetization and asset privacy section below for details.

Onboard a hosted AI model

We support onboarding models up to 8GB in size and requiring up to 16GB RAM (equivalent to ~13B parameter LLMs). These models are hosted directly on aiXplain’s infrastructure and are scalable via Kubernetes.

• Guide: Onboard a custom model
• Guide: Onboard LLMs from Hugging Face
• Contact us at care@aixplain.com

Onboard a passthrough AI model

If you have a model hosted on a PAYG (pay-as-you-go) cloud provider, we can onboard it as a passthrough asset. We support both public and private configurations.

For subscription-based models, reach out to care@aixplain.com to discuss integration options.

Onboard a dataset

Datasets can be used for:

• Indexing for RAG-based agents
• Batch input for agent processing
• Fine-tuning or benchmarking

Datasets support a single input and a single output/reference column, along with unlimited metadata.

• Programmatic onboarding guide
• No-code onboarding guide

Onboard a corpus

Corpora allow multiple input and reference columns, and can be used to derive datasets for the purposes listed above.

• Guide: Onboard a corpus

Onboard a file

Files of any extension can be onboarded and deployed as private assets. These assets will be accessible from your team’s dashboard.

Onboard a utility (custom Python function)

You can deploy a Python function as a private utility asset, making it reusable across multiple agents.

• Utilities must be a single encapsulated function (including all imports)
• Size limits apply; utilities run in a sandboxed environment but can access the internet

1. Define a Python function that encapsulates the tool’s logic
2. Convert it into a utility asset using ModelFactory.create_utility_model
3. Deploy it — the utility appears in your dashboard with a unique asset ID and pricing
4. Add it to any agent via AgentFactory.create_model_tool

Example:

from aixplain.factories import ModelFactory, AgentFactory

def get_today_date(format="%Y-%m-%d"):
    from datetime import datetime
    return datetime.now().strftime(format)

# Deploy as utility
utility = ModelFactory.create_utility_model(
    name="Get date",
    description="Returns today's date",
    code=get_today_date
)

# Convert to tool and attach to agent
tool = AgentFactory.create_model_tool(model=utility.id)
agent.tools.append(tool)

Monetization and asset privacy

• All onboarded assets are private by default
• Private assets are counted against your team’s dedicated storage quota
• Costs are based on compute and storage usage
• You may request to make an asset public to allow other users to access it
• Public assets allow you to set your own usage pricing
• aiXplain takes a 20% revenue share on public asset usage
• For public assets, ensure your team has sufficient prepaid credits to cover usage by others

For help with onboarding or monetizing any asset, contact us at care@aixplain.com.