flybird

I built GoodSMS as a side project, an on-device LLM SMS assistant that drafts fast, context-aware replies to your text messages.

Most “AI messaging” tools today operate as cloud services or full chat apps. We wanted something closer to a digital-twin input method—a thin layer that sits on top of your existing SMS app and quietly helps you respond faster.

What GoodSMS does

• Runs an LLM locally (no cloud round-trip for drafting) • Reads the incoming SMS text, thread context, and your previous writing style • Generates 3–5 possible replies instantly • Lets you accept/edit/paste into your messaging app • Supports short messages, long-form replies, quick actions, confirmations, and scheduling • Privacy-first architecture: no message content leaves your device unless you explicitly opt into cloud inference

Why we built it

Most people lose time triaging simple messages: “ok,” “sure,” “on my way,” “what’s the address again,” etc. Others tend to miss messages or delay replies because the friction is too high.

GoodSMS tries to behave like a personal executive assistant for SMS, especially useful for: • Busy professionals • Parents coordinating logistics • Service providers handling many similar conversations • Anyone who wants faster, cleaner messaging flow

Technical notes

• Android-first implementation using a custom input-method wrapper • On-device LLM inference (quantized 3B–8B models) • Optional cloud-compute escalation for long messages • Conversation-thread reconstruction • Lightweight ranking layer for human-like prioritization • Zero dependency on carrier APIs

Open Questions / Looking for Feedback

I would really appreciate feedback from this community on:

How to improve the on-device inference/latency tradeoff

Whether there is value in adding a plug-in layer (e.g., automate routine replies, reminders, follow-ups)

Ideas for a secure way to integrate with RCS and third-party messaging

Whether a more advanced “agentic” mode would be useful or too risky

I just launched the first public version today. Happy to answer all technical questions, share architectural details, or discuss edge cases.

cloudflare and akami are still very expensive especially for video. Any chance for a p2p version of cloudflare?

YC company peer5 developed P2P CDN for many years but only ends up being acquired by MS. Is there a chance to leverage p2p replacing cloudflare or fastly?

Does OWUI stand for Open Web UI?

My understanding is that Open Web UI and Perplexica each have their own focus.

However, through this tutorial, they can jointly use the same local SearxNG and Ollama service instance.

Thank you for your valuable feedback.

During the installation process, Terminus requires over 110+ images. Considering Docker Hub's rate limiting, the script initially prompts users to download all images locally.

The script executes the actual installation command at the end. The installation process generally follows these steps: 1. Install K8S/K3S 2. Import images sequentially 3. Install system applications using Helm 4. Wait for system startup, then user activation

We understand your concerns. As a new system without an established reputation, we recognize it's challenging for users to try it out.

We've added installation methods for Windows WSL and Raspberry Pi, allowing users to test in virtual machines or temporary environments without concerns about their exiting setups.

Over the next two months, we plan to develop a graphical installation tool to further simplify the process.

Thanks again for your input.

You are absolutely right.

Thanks for correcting and elaborating on this. We used a simplified version (maybe a bit oversimplfiied) to make it easy for the average audience to understand how AI search works. We will be more careful with accuracy in the future.

You're right! This is actually one of people's concerns about Perplexity's future.

Additionally, the way I see it, running a local Perplexity alternative is not necessarily for better results, but for better privacy protection. Plus, it's cool.