Our bots 🤖

TLDR: How do our bots work ❓

• We run the bots inside a free-tier Space, which acts as a server.
• We use Gradio apps as APIs to use them in our bots

Building blocks of a Discord Bot 🤖

1. Create an application
2. Create a Hugging Face Space
3. Add commands

After that, we'll have a working discord bot. So how do we spice it up with machine learning?

Using ML demos in your bot 🧠

• Almost any Gradio app can be used as an API! This means we can query most Spaces on the Hugging Face Hub and use them in our discord bots.

  

Here's an extremely simplified example 💻:

from gradio_client import Client

musicgen = Client("huggingface-projects/transformers-musicgen", hf_token=os.getenv("HF_TOKEN"))

# call this function when we use a command + prompt
async def music_create(ctx, prompt): 
    # run_in_executor for the blocking function
    loop = asyncio.get_running_loop()
    job = await loop.run_in_executor(None, music_create_job, prompt)

    # extract what we want from the outputs
    video = job.outputs()[0][0]

    # send what we want to discord
    await thread.send(video_file)

# submit as a Gradio job; this makes retrieving outputs simpler
def music_create_job(prompt):
    # pass prompt and other parameters if necessary
    job = musicgen.submit(prompt, api_name="/predict")
    return job

In summary, we:

1. Use a command and specify a prompt ("piano music", for example)
2. Query a specific Gradio Space as an API, and send it our prompt
3. Retrieve the results once done and post them to discord

🎉 And voila! 🎉

For further explorations (depending on your needs), we can recommend checking these out 🧐:

• Events in discord bots (to automate some behavior)
• Handling concurrency (important if you're making many concurrent requests at once)
• UI (discord buttons, interactive fields) (can add a lot of functionality)