LiveStore#
https://github.com/livestorejs/livestore
LiveStore is a next-generation state management framework that combines a reactive SQLite database with a built-in sync engine to provide a data layer for applications. It serves as a replacement for traditional state management libraries like Redux or MobX by offering a client-centric approach where data is stored locally in an embedded SQLite database that can be queried reactively using either a built-in query builder or raw SQL. The SQLite back-end is what I find most interesting as it makes the data easily accessible for extending and debugging.
How does LiveStore compare to Redux or MobX?#
LiveStore and Redux are alike because, in the LiveStore docs' words, "both are based on event-sourcing methodologies." For MobX, the same page lists it with six other libraries that "solve in-memory state management (or server-cache management) but not persistence, offline support, or cross-client syncing." LiveStore keeps its state in SQLite, supports client persistence and has a built-in sync engine. This roundup calls it "a replacement for traditional state management libraries like Redux or MobX." The mapping below comes from the Technology comparison page, read November 16, 2026.
| Redux | LiveStore | What the docs say |
|---|---|---|
| Actions | Events | Both are used to describe "things that have happened" |
| Views | State, such as SQLite tables | Both are derived from the history of events/actions |
| Reducers | Materializers | Redux reducers are "similar to LiveStore's materializers" |
The same page names three ways LiveStore differs: SQLite for "flexible queries and aggregations", "client-persistence out of the box" and a "built-in sync engine". It names one downside: "As a downside compared to Redux, LiveStore has a slightly larger bundle size."
The page names seven libraries, including MobX: Zustand, Redux Toolkit, MobX, Jotai, XState, Recoil and TanStack Query.
How does LiveStore sync data between clients?#
This roundup says LiveStore "combines a reactive SQLite database with a built-in sync engine." LiveStore syncs events through a central sync backend. Each client then builds its own SQLite state from those events. The Syncing page, read November 16, 2026, describes a "push/pull" model "similar to how Git works".
A client pulls upstream events before it pushes its own. The page says "The sync backend acts as the global authority", so LiveStore "can't be used in a fully decentralized/P2P manner." For adapters with several client sessions, such as web, it also supports "local syncing across client sessions (e.g. across browser tabs or worker threads)".
The FAQ asks "Is there a hosted sync backend provided by LiveStore?" Its answer is "No, LiveStore is designed to be self-hosted or be used with a 3rd party sync backend." The docs describe these sync providers.
| Provider | Package | What its docs say |
|---|---|---|
| Cloudflare Workers | @livestore/ | It "uses Durable Objects for connectivity and, by default, persists events in the Durable Object's own SQLite." D1 is optional. |
| ElectricSQL | @livestore/ | "Protocol: HTTP push/pull with long-polling support" |
| S2 | @livestore/ | "Protocol: HTTP push/pull, live pull via SSE". The docs call S2 "the official S2 backend (s2.dev)", and SSE means Server-Sent Events. |
| Your own | SyncBackend | Implement the SyncBackend interface on the client. The docs show its pull and push methods as a "Slightly simplified API". A server side must also "Implement the actual sync backend protocol", which validates and persists events. |
Which platforms and frameworks does LiveStore support?#
The README lists adapters for "web, mobile, server/edge, desktop". The docs' page index also lists framework integrations for React, Vue, Solid, Svelte and custom elements. The web adapter persists with OPFS, and the Expo adapter uses native SQLite.
| Platform | Package | What its docs page says |
|---|---|---|
| Web | @livestore/ | LiveStore "currently only support OPFS to locally persist its data". In Safari and Firefox private browsing it "automatically falls back to in-memory storage". |
| iOS and Android with Expo | @livestore/ | Uses native SQLite through the expo-sqlite library. |
| Server | @livestore/ | "Works with Node.js, Bun and Deno." |
| Cloudflare Workers | @livestore/ | Runs LiveStore apps "on Cloudflare Workers with stateful Durable Objects". |
| Electron | Web adapter | "LiveStore can already be used in Electron through the web adapter". A native adapter is tracked in a GitHub issue. |
| Tauri | Web adapter | "LiveStore doesn't yet support a native Tauri adapter", but "you can already use the web adapter with Tauri." |
How do you query LiveStore's SQLite data?#
You query LiveStore's SQLite data with its query builder or with raw SQL. The SQL queries page says LiveStore "also provides a small query builder". Raw SQL needs "the result schema". The page also says "Query results should be treated as immutable/read-only" and recommends paginating queries that could return many rows.
The SQLite page says "LiveStore operates two SQLite databases by default: a state database (your materialized tables) and an event log database".
| Database | What the docs say |
|---|---|
| State database | "Query state via your materialized tables and LiveStore APIs" |
| Event log database | It is "not intended for direct reads or writes". |
Is LiveStore ready for production, and when does it not fit?#
The State of the project page says "LiveStore is currently in beta" and "LiveStore is not yet ready for all production scenarios." Among the reasons the When to use LiveStore page gives for not using LiveStore are an existing source-of-truth database, data "highly connected across users" and a bundle you want "as small as possible". The README says "Livestore is licensed under the Apache License 2.0."
The same State of the project page says "There is currently no specific timeline for a 1.0 release". It also says "minor releases can contain three kinds of breaking changes (patch releases don't contain breaking changes)": API changes, client storage format changes and sync backend storage format changes.
| Topic | The docs say it fits | The docs say it does not fit |
|---|---|---|
| Existing data | A new app, because LiveStore "doesn't yet provide a way to re-use an existing database" | You have "an existing database which is the source of truth of your data" |
| Data shape | Apps that need "solid offline support" and "audit logs" | Data "highly connected across users" |
| Size | None listed | You want the app bundle "as small as possible", and LiveStore "adds a few hundred kB to your app bundle size" |
The table comes from the When to use LiveStore page, read November 16, 2026. Under Considerations, it also sets a database limit: "All the client app data should fit into a in-memory SQLite database". Depending on the device, "having databases up to 1GB in size should be okay". For more data, it says "you can consider segmenting your database into multiple SQLite databases". The FAQ asks "Can I use my existing database with LiveStore?" and answers "Not currently."
For an existing server database, the Technology comparison page names ElectricSQL, Zero and PowerSync: "Choose them when you need to keep an existing server database authoritative."
LivePortrait#
https://github.com/KwaiVGI/LivePortrait
I’m a sucker for the various AI Image > Video projects because they serve as a good bellwether for gen-AI capabilities. What was impressive months ago can look like an 8-bit game today. I also appreciate being able to run projects locally so I can justify the purchase of a GPU for “testing”. LivePortrait is easy to setup and pretty fun to experiment with. I’m sure there are practical uses as well.
What does LivePortrait need to run locally?#
The README, read November 16, 2026, covers Linux, Windows and Apple Silicon Macs, and says to have git, conda and FFmpeg installed. It then has you create a conda environment with Python 3.10 and download the pretrained weights from Hugging Face.
After that, setup splits by machine.
| Machine | Humans mode | Animals mode |
|---|---|---|
| Linux or Windows | The "Check your CUDA versions" step is "optional" if you only want to run Humans mode. | "Animals mode is ONLY tested on Linux and Windows with NVIDIA GPU." X-Pose is "required by Animals mode". |
| macOS, Apple Silicon | "Humans mode works as usual", with a separate requirements file. The README's macOS command sets PYTORCH_ | "Animals mode is not supported." |
Running python inference.py writes an mp4 file that "includes the following results: driving video, input image or video, and generated result." Running python app.py starts a Gradio interface, which the README offers "for a better experience". It also links a Windows one-click installer and an online demo on Hugging Face.
For your own driving video, the README recommends a 1:1 crop and says "the first frame of driving video is a frontal face with neutral expression."
Can you use LivePortrait commercially?#
Not as shipped, going by two license files read November 16, 2026. LivePortrait's LICENSE file says: "If you want to use the LivePortrait project for commercial purposes, you should remove and replace InsightFace’s detection models to fully comply with the MIT license." Animals mode also needs X-Pose, and the X-Pose license allows use "solely for your non-commercial research purposes."
| Part | What the LICENSE file says |
|---|---|
| MIT License text | It starts with "MIT License" and "Copyright (c) 2024 Kuaishou Visual Generation and Interaction Center". |
| InsightFace | "The code of InsightFace is released under the MIT License. The models of InsightFace are for non-commercial research purposes only." |
The README's Ethics section says "Portrait animation technologies come with social risks, particularly the potential for misuse in creating deepfakes." It adds "we do not assume any legal responsibility for the use of the results generated by this project."
BillionMail#
https://github.com/aaPanel/BillionMail
BillionMail is an open-source, self-hosted email marketing platform and mail server designed to provide businesses and individuals with complete control over their email campaigns without the high costs and limitations of commercial services. It combines mail server functionality with comprehensive email marketing capabilities, including newsletter management, advanced analytics for tracking delivery and engagement metrics, customizable templates, and unlimited sending capacity. The platform integrates RoundCube webmail access, supports both transactional and promotional email sending, and emphasizes privacy by keeping all data on your own servers rather than relying on third-party services. Licensed under AGPLv3, BillionMail offers a Docker-based installation process and aims to be a fully-featured alternative to expensive closed-source email marketing platforms while maintaining developer-friendly deployment and complete freedom from monthly subscription fees.
What does it take to self-host BillionMail?#
The Quickstart page, read November 16, 2026, asks for a Linux server on amd64 or arm64 with "CPU: 1Core", "RAM: 1GB" and "Hard disk: 20GB" at minimum. It also asks for open mail ports.
| Service | TCP port |
|---|---|
| Postfix SMTP | 25 |
| Postfix SMTPS | 465 |
| Postfix Submission | 587 |
| Dovecot IMAP | 143 |
| Dovecot IMAPS | 993 |
| Dovecot POP3 | 110 |
| Dovecot POP3S | 995 |
The page says "Currently, mail servers can only communicate via port 25". It tells you to submit a ticket to the server provider to open port 25, or to use an SMTP relay server. The SMTP relay page says "Most cloud servers (e.g., AWS EC2, DigitalOcean) block port 25 by default". It names "AWS SES, SendGrid, Mailgun" as relays and says "only one SMTP relay can be used for a domain."
The Quickstart page marks the one-click installation script as "Recommended" and lists Docker Compose as the other method. The Docker page asks for Docker and Docker Compose, with BILLIONMAIL_HOSTNAME in the .env file set to your domain, such as mail.domain.com.
The README's License section says "BillionMail is licensed under the AGPLv3 License." It then lists three things you can do: "Use the software for free.", "Modify and distribute the code." and "Use it privately without restrictions." Section 13 of the GNU AGPLv3 says a modified version "must prominently offer all users interacting with it remotely through a computer network (if your version supports such interaction) an opportunity to receive the Corresponding Source of your version".
How do you install and manage BillionMail?#
The README lists three ways to install: an install script, a one-click install on aaPanel and a Docker install. Afterwards you manage the server with bm commands such as bm start, bm stop and bm status, listed on the commands page.
| Command | What the commands page says |
|---|---|
| bm default | "View login default information" |
| bm status | "Check the running status" |
| bm update | "Update BillionMail" |
| bm change-password | "Modify the BillionMail administrator password" |
The page says "bm command equals bash bm.sh", and points to bm help for more commands.
Which DNS records does BillionMail need?#
BillionMail needs six DNS records: the Domain page lists A, MX, SPF, DKIM, DMARC and PTR records. It says the free Let's Encrypt certificate needs "A Record DNS resolution" configured first. The PTR record is "Generally configured at the server provider or network operator".
The README's management commands include bm show-record, described as "Show domain DNS record". The Docker page says that to apply for Let's Encrypt free certificates, "port 80 must be used."


