Catching up: Sparkle's Roadmap and Life's Lemons

Hey there, fam! It’s been a rough few months, but I’m back and ready to get moving on Sparkle again. This update has been a long time coming.

First, the disappearance. The last few months threw a lot at us. My wife’s health took a hard turn, and for the previous 3 months she’s been battling one illness after another, ending with COVID for the second time. Between a day job, freelance work, our five dogs and four cats, and a 10-year-old home for the summer, family had to come first.

Things have finally settled down now that she’s on the mend, and I’m honestly more excited than ever to get back into Sparkle.

Sparkle Update

So, where do things stand? I’ve been more or less rewriting the codebase from the ground up. The first version of Sparkle was really a proof-of-concept prototype, no proper error handling, no tests. Now I’m hardening it up and getting it ready for GitHub.

Before that, there are a couple of features I really want to nail down first. One is Sparkle Server, which lets you attach your agents to an OpenAI-compatible chat endpoint and start talking to them with just a few lines of code. I want that flow to be smooth. I’m also reworking function calling and tool usage, aiming for a more intuitive API than what I have now.

SparkleServer::register(
    'nova-sonnet',
    fn () => Agent::create(
        llm: Anthropic::create(
            AnthropicModel::Sonnet35,
        ),
        prompt: view('prompts.characters.nova-v2'),
    )
);

Once those pieces are in place, along with solid documentation, I’ll push Sparkle to GitHub for an alpha/beta release. You’ll finally get to try it out and break it! From there it’s on to 1.0, which I’m super stoked about.

For 1.0, I’m focused on chat through Sparkle Server, plus image support, function calling, and tool use. I’m also building a tight integration with the OpenWebUI project so Sparkle can lean on everything they’ve already built. And tracing, which is honestly one of the most useful things you can have when you’re working with LLMs.

There’s more. I want Sparkle to work well with the big three: OpenAI, Anthropic, and Ollama. I’ll support other providers and models eventually, but these are my daily drivers, so I want every feature working solidly across all three first.

Post-1.0, I’ll be heads-down building a bunch of drop-in tools for you all, with plenty of examples and real documentation. Audio support (hello, Whisper!), embeddings, document splitting, storage for chat-the-docs workflows. I’ve got a long list.

So that’s where things are: the roadmap, and a bit of why I went quiet. I’m aiming to have something up on GitHub in the next few weeks to a month, life permitting. After that, it’s go time!

I can’t thank you all enough for the excitement and support. Knowing you’re waiting on Sparkle has honestly kept me going these past few months, and I can’t WAIT to get it in your hands. More soon!