Trending Open-Source Github Projects: php-src, Impeccable, claude-mem, SPACE, DeepInfer & Filee #303 Transcript, AI Summary & Key Points

AI Engineering from Scratch is a free, open-source curriculum developed by rohitg00 for learning to build and ship AI systems end to end. Its 20-phase sequence covers development tooling, mathematical and machine-learning foundations, deep learning, NLP, computer vision, generative AI, LLM applications, agent engineering, Model Context Protocol (MCP), agent skills, reinforcement learning, and related topics, with implementations in Python, TypeScript, Rust, and Julia. The repository describes 511 lessons and approximately 329 hours of material; each lesson produces a reusable artifact such as a prompt, skill, agent, or MCP server. Learners are instructed to read the lesson documentation, type and run the code from the repository root, record command evidence and outputs, and make a small change before continuing. It provides routes for complete foundations, mathematics and machine learning, production LLM applications, agent engineering, MCP, agent skills, and Claude certification preparation. The repository is distributed under the MIT license and is accompanied by a website containing the same lesson content.

ascii-magic-art is an open-source library of 634 animated ASCII art pieces for web pages, READMEs, wallpapers, and terminals, organized into 24 categories. Its npx command-line browser lets users navigate pieces and categories, change rendering styles such as ASCII, dots, binary, dithering, and custom character sets, and copy a prompt for adding a selected piece to a coding project. Pieces can also be played directly in the terminal, embedded as the <ascii-art> web component, or installed as ES modules that expose metadata and a time-based draw function; runtime modules render frames to HTML or canvas for PNG, GIF, and video export. The repository includes a validator for frame dimensions, character sets, animation, and drawing time. It is distributed under the MIT license, with two photo-based scenes retaining the licenses of their Unsplash source photos.

Claude-Mem is an open-source persistent-memory plugin and service for coding agents. It captures agent activity through lifecycle hooks, stores sessions, observations, and summaries in SQLite, and uses hybrid full-text and Chroma vector search to retrieve relevant context across sessions. Its MCP search workflow uses progressive disclosure: the agent first searches a compact index, then reviews a timeline, and finally fetches full observations for selected result IDs. A local worker service, managed by Bun, provides the HTTP API, search endpoints, and web viewer; the project also supports integrations with Claude Code and other listed agent environments, configurable context injection, private-content exclusion tags, and optional cloud synchronization. The repository states that it is distributed under the Apache License 2.0 and requires Node.js 20 or later, with Bun, uv, and SQLite used by the runtime. It can be installed through its npx installer or Claude Code's plugin marketplace.

DeepInfer is DeepSeek's open-source, high-performance inference framework for offline and large-scale inference of DeepSeek models. It provides optimized kernels and model and sampler runtimes, attention data and context parallelism, mixture-of-experts expert parallelism with EPLB, speculative decoding through DSpark and confidence-scheduled verification, bounded replay for decoder prefill, asynchronous CPU batch scheduling, batch-invariant inference, deterministic sampling, and structured outputs constrained by regular expressions, JSON Schema, or EBNF grammars. The repository currently documents support for DeepSeek-V4.1-Flash on NVIDIA SM100 GPUs and Huawei Ascend 950 NPUs. It runs on Linux with Python 3.12 and requires CUDA 13 for NVIDIA or CANN 9.2 for Ascend. The release does not include a KV-cache system, prefill/decode disaggregation, or load balancing across data-parallel replicas needed for a complete inference service. DeepInfer is released under the Apache License 2.0.

Dzine is an AI creative workspace for generating, editing, and developing images and videos for individual creators and creative teams. It combines image and video generation, enhancement, lip syncing, text-to-speech, motion capture, image-to-3D conversion, generative fill, background and object editing, style transfer, asset libraries, script analysis, reusable character assets, and real-time collaboration in one workspace. Users can upload images as references, generate content from text or images, adjust composition through layering, positioning, sketching, and predefined styles, and organize prompts, model choices, edits, approvals, contributors, and project assets. The service includes image, video, and audio generation nodes, team controls for members, seats, quotas, workspaces, and credits, and copyright tools that flag potential copyright, likeness, trademark, and similarity risks. It uses credits for AI generations and other creative operations and offers free and paid plans.

editonny is a free, open-source local video-editing skill for Claude Code and Codex, created and maintained by Matheus Gomes. Users describe an editing result in plain language; the skill inspects footage, selects complete takes, measures cuts against the waveform, assembles a project, adds captions or motion scenes when requested, reviews a representative sample, and exports a checked MP4. Transcription, rendering, OCR caption checks, and media processing run locally, while optional external generators require authorization before costs or media transmission. It preserves original footage and requires confirmation of editing settings before processing. The repository includes a command-line installer, a local guide, bundled motion scenes and clips, project files, verification manifests, and diagnostic and test commands. First-party code and documentation are licensed under the MIT License; third-party assets retain their own licenses.

Filee is a free, open-source desktop file-conversion tool developed and published by Myeonggu Kim (KnifeLemon). It opens a format menu at the cursor when a key is held and files are dragged, showing conversion targets appropriate to the selected files; conversions run locally and save results beside the originals. It supports image, document, spreadsheet, presentation, PDF, e-book, video, audio, subtitle, archive, CAD, font, data, and 3D-model formats, with batch and parallel conversion, automatically discovered multi-step routes, watch folders, a right-click menu, and a command-line interface. Optional engines such as Ghostscript, LibreOffice, Pandoc, calibre, and FFmpeg are downloaded only when needed. Filee runs on Windows, macOS, and Linux, and is licensed under the MIT License.

Free My TV is an open-source, ad-free replacement home screen and screensaver for Google TV and Android TV, LG webOS TVs, and Samsung Tizen TVs. It puts selected apps on a configurable shelf and provides non-sponsored trending picks from installed services, weather, sunset-based light and dark themes, market prices, and a photo screensaver. The project can be installed through ADB on Android TV, Developer Mode on LG, or TizenBrew and Samsung Developer Mode on Tizen; it does not require root or uninstall the factory home screen, and its Original option restores the previous setup. On Android, the plain Java app builds without Gradle or libraries into an APK and uses the TV's local ADB interface to set the home activity, screensaver, and selected package states. A separate watchdog restores the stock home screen after repeated crashes. The LG and Samsung versions use the same interface in HTML, CSS, and JavaScript, packaged as webOS and Tizen applications. Content comes from TMDB, Open-Meteo, and optionally Binance and Watchmode, depending on the user's configuration. The repository is MIT-licensed and provides build and installation scripts, configuration for city, language, services, apps, weather, and prices, and optional transfer of personal photos to supported Android TVs. It states that the app has no accounts, analytics, or tracking.

fx0 is an open-source web application and lightweight design agent from Vercel Labs, built with libfx. A user describes a page and the agent produces it as a single HTML file, with a live sandboxed preview, design editing and publishing tools, web search and page-fetching tools, and optional subagents. The app stores chats and design versions, allowing previous versions to be reopened or downloaded. libfx supplies the agent loop, streaming, tool calls, retries and durable sessions; on Vercel, Vercel World and Vercel Queues let a turn continue after the function handling it crashes or reaches its time limit. Model calls are billed through Vercel AI Gateway using the signed-in user's team, with an optional sponsored trial. The repository includes local setup and deployment instructions using Node.js, Neon Postgres and Upstash Redis. It is licensed under Apache-2.0; its bundled design styles come from VoltAgent's awesome-design-md under the MIT license.

gitai is an open-source, MIT-licensed version-control system designed for AI coding agents. It moves repository operations onto a server: clients fetch individual objects and sparse trees, while the server handles history queries, blame, path history, ancestry, code search, and derived-data caching. Changes land by rebasing onto the current branch tip and using compare-and-swap rather than a merge queue; concurrent lands can be batched, and accepted reference updates are recorded in an ordered changeset log. Conflicts can be resolved through a versioned, vendor-neutral protocol using an LLM provider, a deterministic tool, or no resolver. The server uses normal Git repositories underneath, with SQLite or PostgreSQL for changesets and references and file or cloud object storage for content-addressed packs. It includes a graft CLI and Go SDK, optional cache daemons, agent tokens and branch policies, semantic review rules, post-landing verification and reversion, and agents that monitor changesets. The repository describes it as an independent implementation inspired by Anthropic's internally developed Scion design, not affiliated with or endorsed by Anthropic.

Harp is an open-source macOS terminal and unofficial fork of the open-source Warp terminal, designed for running multiple Claude Code and Codex sessions side by side. It replaces Warp's tab bar with a sidebar whose rows show whether an agent is working, waiting, or finished, with unread indicators, named groups, and optional tmux synchronization that preserves session layout across restarts and clients. The sidebar footer can display provider-specific limit rings for five-hour and weekly windows. Harp reads Codex rate-limit data from local session logs and captures Claude Code plan limits through its status-line integration; it does not fetch limits or read credentials. Its tmux integration can adopt sessions started elsewhere, including over SSH, and a Claude Code hook can report background tasks. Harp adds no servers of its own, while Warp's existing AI and cloud features continue to communicate with Warp's servers. The application has its own bundle identifier, URL scheme, settings directory, and updater, so it can run alongside Warp. It can be installed from a release DMG or built from source with the Rust toolchain. The repository states that the code is licensed under GNU AGPL v3, except for the warpui and warpui_core crates, which use MIT licenses. The video describes it as Apple silicon-only.

Impeccable is an open-source design skill for AI coding agents, developed by pbakaus. It provides a shared design vocabulary with 23 commands, including init, shape, critique, audit, polish, harden, animate, typeset, layout, and live. The init command gathers product and design context into PRODUCT.md and DESIGN.md files, while other commands review, plan, modify, or refine a frontend project. Its CLI and browser extension run 61 deterministic detector rules for common AI-generated frontend design problems without an LLM or API key; it also supports LLM-based critique checks. The live command provides browser-based visual iteration, and the package can be installed from a project root with npx impeccable install.

Knowledge Work Plugins is an open-source repository of role-specific plugins for Claude Cowork and Claude Code. Each plugin bundles domain skills, MCP connector configuration, slash commands, and, where applicable, sub-agents for workflows such as productivity, sales, customer support, product management, marketing, legal, finance, data, enterprise search, and biomedical research. Plugins use file-based Markdown and JSON rather than application code, infrastructure, or build steps; users can install them through Cowork or the Claude Code plugin marketplace and customize their connectors, company context, and workflow instructions.

ManuAGI & AutoGPT Tutorials is an email newsletter covering AI, AGI, and AI-agent project updates. Its subscription page says it provides new project updates every week and is published through beehiiv.

php2shell-checker is an OpenAI version-based PHP security checker. It checks a website's advertised PHP version from the X-Powered-By and Server headers, or checks a local/server installation, then compares the version with listed upstream fixed releases and reports OK, POTENTIALLY VULNERABLE, or UNKNOWN. Its URL mode sends a HEAD request and falls back to GET only for 405 or 501 responses; it does not follow redirects, inspect response bodies, or send exploit probes. Because HTTP headers may not reveal backported fixes, an OK result describes only the advertised version and does not verify the installation. It provides phpcheck-url and phpcheck-server skills, supports JSON output, and is licensed under the Apache License 2.0.

Rakazo is an open-source platform for running persistent AI teammates. Each bot has conversations, memory, routines, and history, and can access a browser, terminal, files, and a graphical desktop. Users can access bots through the web app, Electron desktop app, or Expo mobile app, and can take control of the graphical desktop when needed. The platform supports shared team computers and isolated private computers, bot delegation to peer bots or short-lived subagents, voice interaction, and user-provided model credentials through Pi. It can use Docker, E2B, Daytona, Box, or trusted local computers as computer providers, and supports integrations through Composio, Pipedream Connect, MCP, OpenAPI, and Treg sources. Rakazo can be self-hosted with Docker and is also available as a managed application. The repository describes it as being in beta and identifies a TypeScript stack using React, Vite, Tailwind CSS, Electron, Expo, Hono, oRPC, PostgreSQL, Prisma, Better Auth, and Graphile Worker.

Scrawl is a Rust screen-recording and annotation tool for macOS that lets users capture a screen or iPhone, narration, clicks, and red ink from a shortcut. Each capture is packaged as a bundle containing annotated keyframes, a contact sheet, full-resolution frames, a time-aligned transcript and interaction timeline, and a structured manifest; captures can be pasted into an active application or sent to coding-agent sessions such as Claude Code, OpenCode, and Codex. Its command-line tool can inspect bundles, extract frames around a timestamp or marked event, route and send captures, import or reprocess recordings, and export annotated MP4 video. Screen captures use logged window, cursor, click, scroll, and drawing activity to auto-crop keyframes around the demonstrated activity, while on-device transcription uses Apple Speech or optional local engines and uses word timings to divide narration and select keyframes. The repository also includes ScrawlKit, a Swift package for recording an iOS app's video, narration, touches, strokes, and application events for delivery to Scrawl. The project is distributed under the MIT license. The macOS app supports Apple silicon Macs running macOS 15 or later; the repository also provides a Windows preview build. Downloads are available from the project's website, through a Homebrew cask, or from GitHub releases.

Sill is a free, open-source remote-desktop app for showing one window or the whole Mac desktop on an iPhone or iPad. It captures the selected content with ScreenCaptureKit, encodes it as HEVC with VideoToolbox, and sends it over an encrypted direct connection; taps, clicks, scrolling, keyboard input, trackpad gestures, and Apple Pencil input are sent back to the Mac. Devices can connect over a local network, USB, peer-to-peer Wi-Fi, or a user-owned VPN such as Tailscale, with one-time pairing by scanned or entered code and no account or intermediary server. Sill for Mac requires Apple silicon and macOS 14 or later, while the iPhone and iPad app requires iOS or iPadOS 17 or later. It is distributed as a notarized Mac download, an App Store app, Homebrew cask, or source code, and is licensed under Apache License 2.0.

SPACE (Sandboxed Platform for Agentic Code Execution) is an open-source sandbox platform from Perplexity for running agent, user, and code-execution workloads in isolated Firecracker microVMs. Applications use its HTTP API to create Linux environments from templates, run multiple commands in a sandbox with a writable filesystem, collect output over a separate server-sent-events data plane, and pause, suspend, resume, or delete sandbox instances. Lifecycle changes are asynchronous and return operation IDs for polling; command sessions use attachment tokens. The platform supports Kubernetes deployments, a local Kind setup, an operator CLI, warm pools, base snapshots, observability features, and an E2B compatibility wrapper. SPACE is implemented in Rust and distributed under the Apache License 2.0.

TwinDesk is a free, open-source cross-device workspace app for Mac, Windows, Linux, and Android. It shares a mouse, keyboard, screens, applications, files, folders, clipboard, sound, and notifications between paired computers and phones; Android devices also support calls, texts, photos, phone files, and apps in desktop windows. Devices communicate directly over the same Wi-Fi network, Bluetooth LE, USB, Wi-Fi Direct, a phone hotspot, or a direct network link, selecting or switching connections as needed. Its TwinDesk Mesh beta connects multiple devices at once with a shared clipboard, notification pool, and screen layout. The repository describes one TLS session per device link and an Electron application paired with native helpers for input capture and injection, virtual displays, window capture, audio, and drag-and-drop operations. TwinDesk requires no account or cloud relay and states that paired devices communicate directly, with no telemetry apart from an optional GitHub release update check. The repository includes desktop and Android applications, build instructions for macOS, Windows, Linux, and Android, and is licensed under the GNU General Public License v3.0.

Zi Wei Dou Shu is an open-source Chinese astrology charting engine following Ni Haixia's Tian Ji teaching system. It generates charts from a Gregorian date and an hour-branch index, applying true-solar-time correction, China's 1986–1991 daylight-saving rules, configurable leap-month conventions, late Zi-hour handling, selectable star-brightness tables, natal/decade/annual Four Transformations, flying-star self-transformations, twelve spirits, and more than 1,100 pattern rules. The TypeScript engine is in lib/ziwei/, with generateChart() as its main entry point, and includes classical-text data and downloadable chart-interpretation datasets. The repository's code is MIT-licensed; its sample and interpretation datasets use CC BY 4.0, while included quotations from Ni Haixia's Tian Ji use CC BY-NC-SA 4.0. The open-source repository does not include the interpretation database, AI prompts, backend routes, accounts, payments, or deployment configuration.