SmartSoftHub

💻AIPex 🛠️ AIPex is an open-source browser extension that transforms your browser into an intelligent automation platform using natural language commands. 🔰It allows you to automate almost any task in the browser, including tab management, form filling, and data extraction from web pages. 🔰AIPex integrates with your current browser without the need to switch to another one, ensuring complete privacy protection and requiring no subscription. 🔗 Links: https://github.com/AIPexStudio/AIPex 🏷️ Tags: #TypeScript

☁️Telegram-Drive 🛠️This is a cross-platform desktop application with open source code, which turns your Telegram account into an unlimited cloud storage. 🔰The program allows you to upload, download, and organize files. 🔰It is developed using Tauri, Rust, and React, ensuring high performance and responsiveness. 🔗Links: https://github.com/caamer20/Telegram-Drive 🏷️Tags: #TypeScript

☝🏻Milla Jovovich launched GitHub and released a memory system for AI The actress announced the launch of the open-source project MemPalace in collaboration with developer Ben Sigman. Within a day, the repository gathered about 5,400 stars. 🔻MemPalace is an open-source local memory system for AI for LLMs and agents. It uses the ancient "memory palace" method to organize all your conversations, decisions, projects, and files into virtual wings, rooms, and corridors, so that the AI never forgets details between sessions. In the LongMemEval test (the leading memory performance test), it showed a 100% result. A separate feature is the built-in "shorthand language" for AI agents called AAAK. This is a way to record the same information thirty times shorter without losing meaning: months of correspondence fit into about 120 units of text, which the model swallows in fractions of a second. And all this works locally, on a laptop, without the cloud and paid services. 🔗 Links: Github 🏷️ Tags: #ai #llm

💻 Openclaude 🛠️ OpenClaude is a tool that allows you to use the Claude Code interface with various language models, such as OpenAI, Gemini, DeepSeek, Ollama, and others, through an OpenAI API-compatible interface. 🔰 It provides simple setup and support for more than 200 models, expanding the capabilities of Claude Code beyond Anthropic models. 🔗 Links: https://github.com/Gitlawb/openclaude 🏷️ Tags: #typescript #openai #claude #ollama #gemini #deepseek

💥 Luma Labs has released UNI-1 — a new model for image generation 🔰 The San Francisco-based startup, with a staff of just ~150 people, has introduced a model that outperforms both Nano Banana 2 from Google and GPT Image 1.5 from OpenAI in key benchmarks. 🔰 In human preference tests (Elo rating), UNI-1 took the first place in three categories at once: overall quality, style and editing, and reference-based generation. In pure text-to-image, it's currently second to Nano Banana. 🔰 The main architectural difference is that UNI-1 has abandoned the diffusion approach used by Stable Diffusion, Midjourney, and Flux. ☝🏻Instead, the model is built as an autoregressive transformer: text and images are processed in a single sequence of tokens. 🧠 Simply put, the model doesn't "pull an image out of the noise," but generates it token by token — just like language models generate text. 🔗 Links: https://lumalabs.ai/uni-1 🏷️ Tags: #ai #image

🗣️ WAAS 🛠️ This is a tool that provides a graphical interface and API for using OpenAI Whisper, designed for transcribing audio and video files. 🔰 Key features include the ability to upload files, automatic transcription with subsequent email delivery of results, and a local editor for adjusting transcriptions in the browser. 🔰 Additionally, WAAS supports various output formats such as SRT and text files, and provides an API for integration with other services. 🔗 Links: https://github.com/schibsted/WAAS 🏷️ Tags: #JavaScript #whisper #openai

👁️‍🗨️ Deep-eye 🛠️ This is an advanced tool for scanning vulnerabilities and conducting penetration tests, based on artificial intelligence. 🔰 It integrates support for several AI providers, such as OpenAI, Grok, OLLAMA, and Claude, for automated vulnerability detection, intelligent payload generation, and professional reporting. 🔰 The program supports more than 45 attack methods, including SQL injections, XSS, SSRF, as well as API and GraphQL security testing. 🔰 Additionally, deep-eye offers advanced reconnaissance capabilities, including passive OSINT, DNS enumeration, and subdomain detection. 🔗 Links: https://github.com/zakirkun/deep-eye 🏷️ Tags: #Python #ai #security #automation

🖥️Swarms 🛠️ This is an industrial framework for orchestrating multi-agent systems, designed to create fully autonomous organizations using the collaborative work of several AI agents. 🔰The key features of Swarms include multi-agent architectures (hierarchical, sequential, and parallel), advanced communication protocols between agents, an ultra-optimized runtime, memory systems for multi-agents, simulation environments, and enterprise-level security and compliance features. 🔗Links: https://github.com/kyegomez/swarms 🏷️Tags: #python #ai #llm

💻Fara 🛠️Fara-7B is a compact language model with 7 billion parameters, developed by Microsoft for automating interactions with web interfaces. 🔰It is capable of monitoring the browser and performing actions on behalf of the user, such as clicking, entering text, scrolling, and selecting elements, to perform various web tasks. 🔰The model is trained on data generated using a multi-agent system, which creates and verifies a variety of web tasks, ensuring the accuracy and safety of the actions performed. 🔰Fara-7B supports automating online shopping, travel booking, restaurant reservations, information search, and account management. 🔗Links: https://github.com/microsoft/fara 🏷️Tags: #python #llm

🌏 Geoai 🛠️ GeoAI is a powerful Python package designed for integrating artificial intelligence with the analysis and visualization of geospatial data. 🔰 It provides a unified framework for processing satellite images, aerial photographs, and vector data using advanced deep learning models. 🔰 Key features include interactive search and download of remote sensing data, automated preparation of datasets, model training for classification, detection, and segmentation, as well as visualization of results. 🔰 GeoAI integrates with popular AI frameworks such as PyTorch and Transformers, and supports working with various data formats, including GeoTIFF, GeoJSON, and Shapefile. 🔗 Links: https://github.com/opengeos/geoai 🏷️ Tags: #python #ai #geo

💥 ByteDance presents Seedance 2.0 — a new generation of AI models for video generation 🎬 Early access to the model is already being distributed to select creators, and a full public launch is expected on the Dreamina (international version) and Jimeng (China) platforms on February 24th. The model has transitioned from generating individual clips to full-fledged multi-frame storytelling. Now, a single prompt can yield an entire scene consisting of several cohesive shots, where the characters, lighting, color correction, and frame space remain consistent from start to finish.🎞 🔗Links: https://seedance2.ai 🏷️Tags: #video #ai

🦀Openclaw 🛠️OpenClaw is an open-source personal AI assistant that operates locally on user devices and integrates with various messengers. 🔰It supports integration with platforms such as Telegram, Discord, Slack, and iMessage, allowing users to interact with the assistant through familiar applications. 🔰OpenClaw has persistent memory, remembering the user's preferences and context, which makes the interaction more personalized. 🔰The program can perform real tasks, such as managing email, scheduling meetings, automating workflows, and controlling smart homes. 🔗Links: https://github.com/openclaw/openclaw 🏷️Tags: #TypeScript #aiagent

💻Pentestagent 🛠️PentestAgent is a comprehensive tool for conducting offensive operations in the field of cybersecurity, equipped with an AI agent and MCP architecture, designed to automate pentesting tasks, vulnerability hunting, and threat hunting. 🔰It integrates tools such as Nmap, Metasploit, FFUF, and SQLMap, allowing users to perform penetration testing, bug hunting, and threat analysis through natural language interaction. 🔰Among the key features are autonomous pentesting using intelligent task trees (PTT), execution of predefined workflows, and generation of Markdown-formatted reports with detailed findings and recommendations. 🔰Additionally, it supports AI response expansion using a local knowledge base, which provides more contextual and informative results. 🔗Links: https://github.com/GH05TCREW/pentestagent 🏷️Tags: #pentest #ai #python

👍 Generating 3D models — Krea has released the top-notch tool Realtime Edit 🔰 The service allows you to superimpose any generated content over photos, videos, and even 3D models, and all edits are done in real time 🔰The generation delay is just 50 milliseconds, so the changes are visible almost instantly 🔗Links: https://www.krea.ai/ 🏷️Tags: #ai #3d

☝🏻Alibaba has released CosyVoice 3 — a speech synthesis model with voice cloning in just 3 seconds of audio The model, with only 0.5B parameters in size, outperforms models three times larger in quality and is distributed with open-source code. The AI is trained on a million hours of audio data and supports 9 languages. The main feature is zero-shot voice cloning: simply upload 3-10 seconds of any person's recording, and the model will reproduce their timbre without additional training. The new speech tokenizer was trained on several tasks at once — speech recognition, emotion detection, language identification, and sound event analysis. ☝🏻This allows the model to capture not only "what is said", but also "how it is said" — with pauses, intonations, and emotional coloring. 🔬 The model supports streaming with a delay of just 150 ms — practically in real time. You can also control emotions, tempo, and volume through instructions. 🔗Links: https://github.com/FunAudioLLM/CosyVoice 🏷️Tags: #ai #voice

👍 400 neural networks in one place 🛠️Poe is an online platform for communicating with various neural networks that will solve any task for you ◦ Hundreds of AI services: for studying, writing, coding, email newsletters, tutors, analysts, designers, services for generating videos, pictures and audio, and so on ◦ All AI agents are available with a single click — click and immediately use the top model ◦ The authors conveniently divided the neural networks into categories ◦ You can use them in the browser, as well as download applications for PC, iOS and Android ◦ Without registration, and without restrictions 🔗Links: https://poe.com/ 🏷️Tags: #ai #agents

💥 We've found a collection of the best repositories for you 🛠️ Awesome-Repos — a selection of the best repositories for almost any task: backend, frontend, DevOps, API, learning, startup ideas, and a bunch of other usefulness. 🔗Links: https://github.com/pawelborkar/awesome-repos 🏷️Tags: #awesome #repos

💥Resemble AI released Chatterbox Turbo 🛠️The first open-source speech synthesis model that outperforms market leaders. 🔰In blind tests, Chatterbox Turbo beats closed-source ElevenLabs Turbo 2.5 and Cartesia Sonic 3. ☝🏻The main feature is voice cloning in just 5 seconds of audio without any training. 🔰At the same time, the model is the first among open-source solutions to support control of emotional expression: you can smoothly adjust the intensity from monotonous speech to dramatic delivery with a single parameter. 🔰Another unique feature is paralinguistic tags. 🔰You can insert tags like [laugh], [sigh], [cough], [chuckle] into the text, and the model will reproduce these reactions naturally, using the same cloned voice. 💬 🔗Links: https://github.com/resemble-ai/chatterbox 🏷️Tags: #ai #voiceclone

📲Open-AutoGLM 🛠️Open-AutoGLM — a ready-made framework to turn an Android phone into a "smart agent" that understands the screen and performs actions based on a user's text request. 🔰Open-AutoGLM is built around the AutoGLM-Phone-9B model and uses it as a brain that sees a screenshot of the screen, understands the interface, and plans step by step what to press and where to enter text. 🔗Links: https://github.com/zai-org/Open-AutoGLM 🏷️Tags : #llm #phone #ai

🏞️ IC-Custom 🛠️IC-Custom is an open-source tool for various image customization based on in-context learning. 🔰It supports the accurate placement of an object from a reference image using a mask or segmentation, and the generation of a new scene with the subject from the reference based on a text description. 🔰IC-Custom can be applied to tasks such as virtual fitting, product placement, furniture arrangement, and character creation. 🔗Links: https://github.com/TencentARC/IC-Custom/ 🏷️Tags: #ComfyUI #imageeditor #ai