I wanted to share an open source Telegram-based personal AI assistant I built. It’s a model-agnostic agent with memory, skills, tools (like web search, browser user, etc.) operating in a persistent workspace. It also has support for scheduled tasks, and can build powerful HTML-based apps that live in the workspace.

Here are some of my favorite use cases:

* Send Lilo photos of food, and it tracks your calories.

* Leave a voice note on your run to pause your supplements, and Lilo adds a TODO.

* Have Lilo remind you when the Knicks game starts and even send you score updates every 5 minutes.

* Have Lilo read an article out loud. Or give you a summary of the top stories on Hacker News.

* Forward a Uber receipt, and pull it up later to file for a reimbursement at work.

* Schedule a meeting with Jess next week, ask for suggestions on location, and next week, remind you to leave for the meeting on time.

While Telegram is my most frequently used channel, Lilo can also be accessed by email, WhatsApp, a website and a mobile app. Email is particularly useful: I often forward receipts, invites, etc for Lilo to handle.

How is this different from OpenClaw and Hermes Agent? Here are some reasons:

- Runs on a remote machine/in the cloud rather than your local machine - your local data is safe, and the assistant is available 24/7.

- More visual/ more GUI - Lilo comes with a default set of apps like a TODO list that you can interact with not just by text, but also with a GUI on the mobile and web app.

- The Telegram integration is very comprehensive (handles replies, voice notes, reactions, etc.).

I use Lilo a ton to manage my life. Would love to hear your feedback!

Github: https://github.com/abi/lilo

Comments URL: https://news.ycombinator.com/item?id=48237730

Points: 2

# Comments: 1

Video demo: https://youtu.be/RteRVM7BSps

Github URL: https://github.com/abi/1-800-CODER

How is pointing implemented? GPT-Realtime-2 only supports image inputs (unlike Gemini Live which also supports video inputs). So, the app sends a screenshot including the cursor when the model emits a speech_stopped event. That way, the agent always has a fresh visual before it replies.

Limitations:

- GPT-Realtime-2 is okay at front-end changes, probably at a GPT-4o level. Small modifications like copy changes, adding/removing elements, formatting updates work really well at low latency. In fact, for these types of changes, this might be my ideal interface. But if you wanted this app to be more useful for larger changes or generating UI from scratch, you’d want to hook up a subagent system that runs a smarter model like GPT 5.5 or Claude Opus.

- GPT-Realtime-2 is expensive. The good news though is that bandwidth is really high here so you might save time with this interface.

Comments URL: https://news.ycombinator.com/item?id=48136879

Points: 1

# Comments: 0

I’ve found it super useful in my personal life and is pretty much my #1 app.

Comments URL: https://news.ycombinator.com/item?id=48014156

Points: 3

# Comments: 0

For a visual intro, here’s a YouTube video demonstrating the features and use cases: https://youtu.be/Jz0l_izoA1w

I started this project because I wanted a few small AI-powered apps for myself — a bookmarks tool, a calorie tracker, a TODO list — but deploying N separate apps with N deployments, URLs, and auth configs is too much effort for a single-user use case. So I built one container that holds all the apps, runs them at the same URL, and lets the agent inside modify them. If I want to change my bookmarks app, I don't open Claude Code and push to a repo — I tell the agent, and it edits the HTML directly. Not great for a large SaaS with lots of users but works great for a single-user app.

Each app is just an HTML file but with access to a filesystem API, full network access and full agentic capabilities.

Since then, Lilo has grown to also support a filesystem/workspace that can hold more than just apps. You can upload PDFs or screenshots and have the AI analyze and organize them for you. The AI also remembers key details about you in a “LLM wiki” style tree of markdown files. It’s a full-on personal assistant.

Inspired by OpenClaw, I added support for additional channels like WhatsApp, email, and Telegram. Now I take a photo of my lunch, text it to Lilo, and the calorie tracker updates. If I didn't eat the pizza crust, I text "didn’t eat the crust" and it adjusts the entry. Cal AI couldn't do that. And unlike say a calorie tracker WhatsApp bot, I also have a nice visual interface to look at my meals.

This combo of personal assistant + personal apps is very powerful. And very flexible. The UI is nice for glancing at data. The chat is nice for operations the UI doesn't cover. I don't have to build a search into every app, I can just ask the agent.

Lilo is open source and alpha software. Bring your own keys. The setup is not the easiest (a lot of API keys and you need to self host). All security advisories for LLM apps with network access apply here. But at the start, since there is no personal data, no data exfiltration is possible but credential exfiltration certainly is. Your entire workspace can be backed up and versioned using a git repo so the data is durable.

I’d love to hear feedback, and hope people find this as useful as I have.

Comments URL: https://news.ycombinator.com/item?id=47894947

Points: 7

# Comments: 4

The HTTP layer is fine. Websockets work great. This is how the Codex app server works, I believe: https://openai.com/index/unlocking-the-codex-harness/ Same pattern I've used in my agentic OS/personal assistant project: https://github.com/abi/lilo Works great!

Chat history shouldn’t be hard to add with local storage and Indexed DB.

Just wrappers all the way down