Nebius Tools

Legacy NebiusTools example for the discontinued Token Factory text-to-image service.

Nebius disabled its text-to-image APIs on April 13, 2026. None of the three generation examples below can run against that retired service; changing the Nebius image model ID does not restore it. See Nebius’s deprecation notice. This code is retained as a historical reference. Do not run it as written.

nebius_tools.py
"""Run `uv pip install openai agno` to install dependencies.

This example demonstrates how to use NebiusTools for text-to-image generation with Nebius Token Factory.
"""

import base64
import os
from pathlib import Path
from uuid import uuid4

from agno.agent import Agent
from agno.tools.models.nebius import NebiusTools
from agno.utils.media import save_base64_data

# ---------------------------------------------------------------------------
# Create Agent
# ---------------------------------------------------------------------------


# Create an Agent with the Nebius text-to-image tool
agent = Agent(
    tools=[
        NebiusTools(
            # You can provide your API key here or set the NEBIUS_API_KEY environment variable
            api_key=os.getenv("NEBIUS_API_KEY"),
            image_model="black-forest-labs/flux-schnell",  # Fastest model
            image_size="1024x1024",
            image_quality="standard",
        )
    ],
    name="Nebius Image Generator",
    markdown=True,
)

# Example 1: Generate a basic image

# ---------------------------------------------------------------------------
# Run Agent
# ---------------------------------------------------------------------------
if __name__ == "__main__":
    response = agent.run(
        "Generate an image of a futuristic city with flying cars and tall skyscrapers",
    )

    if response.images:
        image_path = Path("tmp") / f"nebius_futuristic_city_{uuid4()}.png"
        Path("tmp").mkdir(exist_ok=True)
        image_base64 = base64.b64encode(response.images[0].content).decode("utf-8")
        save_base64_data(
            base64_data=image_base64,
            output_path=str(image_path),
        )
        print(f"Image saved to {image_path}")

    # Example 2: Generate an image with the higher quality model
    high_quality_agent = Agent(
        tools=[
            NebiusTools(
                api_key=os.getenv("NEBIUS_API_KEY"),
                image_model="black-forest-labs/flux-dev",  # Better quality model
                image_size="1024x1024",
                image_quality="hd",  # Higher quality setting
            )
        ],
        name="Nebius High-Quality Image Generator",
        markdown=True,
    )

    response = high_quality_agent.run(
        "Create a detailed portrait of a cyberpunk character with neon lights",
    )

    # Save the generated image
    if response.images:
        image_path = Path("tmp") / f"nebius_cyberpunk_character_{uuid4()}.png"
        Path("tmp").mkdir(exist_ok=True)
        image_base64 = base64.b64encode(response.images[0].content).decode("utf-8")
        save_base64_data(
            base64_data=image_base64,
            output_path=str(image_path),
        )
        print(f"High-quality image saved to {image_path}")

    # Example 3: Generate an image with the SDXL (Stability Diffusion XL model) model
    sdxl_agent = Agent(
        tools=[
            NebiusTools(
                api_key=os.getenv("NEBIUS_API_KEY"),
                image_model="stability-ai/sdxl",  # Stability Diffusion XL model
                image_size="1024x1024",
            )
        ],
        name="Nebius SDXL Image Generator",
        markdown=True,
    )

    response = sdxl_agent.run(
        "Create a fantasy landscape with a castle on a floating island",
    )

    # Save the generated image
    if response.images:
        image_path = Path("tmp") / f"nebius_fantasy_landscape_{uuid4()}.png"
        Path("tmp").mkdir(exist_ok=True)
        image_base64 = base64.b64encode(response.images[0].content).decode("utf-8")
        save_base64_data(
            base64_data=image_base64,
            output_path=str(image_path),
        )
        print(f"SDXL image saved to {image_path}")

Current alternative

Use the OpenAI image-generation example or the Gemini image-generation agent.

Full source: cookbook/91_tools/models/nebius_tools.py