Add Modal (serverless GPU provider) example with advanced features (config.toml)

Update the app.py to show how to setup the config.toml with Modal

I already have it working. Let me know if you are interested in PR.

---

Please reply with a 👍 if you want this feature.

For anyone interested in contributing, I suggest checking out the documentation on persisting volumes in Modal: https://modal.com/docs/guide/volumes#persisting-volumes for more details. Alternatively, you can consider copying the config.toml during the image building process in Modal.

The app.py file itself serves as a minimal example to run Tabby on Modal, so it may not be ideal to update it directly. But an example under experimental/ is certainly good.

Hey can I have this?

I haven't managed to use the persisting-volumes (it gives an error that it can not write on a lazy file system or so), but if you index the repos at the image build stage, you can do it like this:

# Afterwords, run: tabby scheduler --now
def render_config():
    return """# Index three repositories' source code as additional context for code completion.

[[repositories]]
name = "tabby"
git_url = "https://github.com/TabbyML/tabby"
"""

def scan_repos():
    import subprocess

    subprocess.run(
        [
            "/opt/tabby/bin/tabby",
            "scheduler",
            "--now"
        ]
    )

def write_config():
    with open("/data/config.toml", "w") as f:
        f.write(render_config())
        f.close()

image = (
    Image.from_registry(
        IMAGE_NAME,
        add_python="3.11",
    )
    .dockerfile_commands("ENTRYPOINT []")
    .run_function(download_model)
    .run_function(write_config)
    .run_function(scan_repos)
    .pip_install("asgi-proxy-lib")
)