Hasty Briefsbeta

Bilingual

GitHub - seerr-team/seerr: Open-source media request and discovery manager for Jellyfin, Plex, and Emby.
2 months ago
- Seerr is a free, open-source media request management tool.
- Integrates with Jellyfin, Plex, Emby, Sonarr, and Radarr.
- Supports PostgreSQL and SQLite databases.
- Manages requests for Movies, Shows, and Mixed Libraries.
- Features customizable request system and user-friendly UI.
- Includes granular permissions and notification agents.
- Mobile-friendly design for on-the-go management.
- Supports watchlisting and blocklisting media.
- Documentation and community support available via Discord and GitHub.
- API documentation accessible locally.
GitHub - maxritter/claude-pilot: Claude Code is powerful. Pilot makes it reliable. Start a task, grab a coffee, come back to production-grade code. Tests enforced. Context preserved. Quality automated
2 months ago
- Claude Pilot is a tool designed to enforce production-grade code quality by integrating linting, formatting, and type checking as mandatory hooks.
- It ensures Test-Driven Development (TDD) is followed strictly with a RED → GREEN → REFACTOR cycle for every feature.
- Endless Mode maintains context across sessions, preventing degradation and enabling seamless continuation of work.
- The tool includes verifier sub-agents that perform independent code reviews before marking tasks as complete.
- Pilot is optimized for output quality with minimal system complexity, avoiding bloated prompts and unnecessary scaffolding.
- It supports quick mode for direct execution and /spec mode for structured planning, implementation, and verification.
- Pilot works with existing projects without restructuring them, adapting to project conventions via the /sync command.
- The tool is available for macOS, Linux, and Windows (WSL2) and can be installed via a curl command.
- Pilot uses different Claude models (Opus and Sonnet) based on the phase of development to balance reasoning quality and speed.
- It includes persistent memory, team collaboration features via a private Git repository, and extensive customization options.
- All development tools run locally, ensuring no code or project data leaves the machine, making it enterprise-compliant for data privacy.
- Pilot is source-available under a commercial license, with different tiers for solo developers and teams.
GitHub - brave/brave-browser: Brave browser for Android, iOS, Linux, macOS, Windows.
2 months ago
- Repository contains build tools for Brave desktop browser on macOS, Windows, and Linux.
- Fetches and syncs code from Chromium and brave-core projects.
- Supports building for Android and iOS with specific commands.
- Offers different build types: Component, Release, Static, and Debug.
- Provides instructions for syncing code, applying patches, and updating dependencies.
- Includes guidelines for Google Safe Browsing API key setup.
- Lists security rules and IPC review guidelines from Chromium.
- Encourages community involvement through Q&A and translations.
- Provides troubleshooting solutions for common problems.
GitHub - moonshine-ai/moonshine: Fast and accurate automatic speech recognition (ASR) for edge devices
2 months ago
- Moonshine Voice is an open-source AI toolkit for real-time voice applications, running on-device for privacy and speed.
- Optimized for low latency, Moonshine processes audio while the user is speaking, improving responsiveness.
- Offers higher accuracy than Whisper Large V3 with models as small as 26MB, suitable for constrained devices.
- Cross-platform support includes Python, iOS, Android, MacOS, Linux, Windows, Raspberry Pi, IoT devices, and wearables.
- Includes high-level APIs for transcription, speaker identification (diarization), and command recognition, simplifying voice application development.
- Supports multiple languages: English, Spanish, Mandarin, Japanese, Korean, Vietnamese, Ukrainian, and Arabic.
- Community support is available via Discord, with examples and documentation to help developers get started.
- Moonshine outperforms Whisper in live speech scenarios with flexible input windows and caching for streaming, reducing latency.
- Language-specific models provide higher accuracy by focusing on one language, unlike Whisper's multilingual approach.
- The API is designed for ease of use, abstracting complex details to allow developers to focus on application logic.
GitHub - steipete/gogcli: Google Suite CLI: Gmail, GCal, GDrive, GContacts.
2 months ago
- gogcli is a fast, script-friendly CLI for various Google services including Gmail, Calendar, Drive, Docs, Sheets, and more.
- Supports JSON-first output, multiple accounts, and least-privilege authentication.
- Features include email tracking, calendar event management, document export, and task management.
- Allows secure credential storage using OS keyring or encrypted on-disk keyring.
- Supports auto-refreshing tokens and least-privilege auth with --readonly and --drive-scope options.
- Can be installed via Homebrew, AUR, or from source.
- Includes commands for managing OAuth credentials, service accounts, and domain-wide delegation.
- Offers detailed documentation for setup, usage, and integration testing.
GitHub - nautechsystems/nautilus_trader: A high-performance algorithmic trading platform and event-driven backtester
2 months ago
- NautilusTrader is an open-source, high-performance algorithmic trading platform.
- It supports backtesting and live trading with no code changes required.
- The platform is AI-first, designed for Python-native environments.
- Core components are written in Rust for performance and safety.
- Supports multiple asset classes including FX, Crypto, Futures, and more.
- Features include high-frequency trading, advanced order types, and multi-venue capabilities.
- Available on Linux, macOS, and Windows with Docker deployment options.
- Includes modular adapters for integrating various trading venues and data providers.
- Offers high-precision and standard-precision modes for core value types.
- Community contributions are welcome under the GNU Lesser General Public License v3.0.
GitHub - ruby/ruby: The Ruby Programming Language
2 months ago
- Ruby is an interpreted object-oriented programming language often used for web development.
- It offers scripting features for text processing, file management, and system tasks.
- Key features include simple syntax, object-oriented capabilities, operator overloading, and garbage collection.
- Ruby is highly portable, working on Unix-like systems, Windows, and macOS.
- Installation options and source code are available on the Ruby official website.
- Development branches and source code can be accessed via Git repositories.
- Community support includes mailing lists, Stack Overflow, and a bug tracking system.
- Ruby was created by Yukihiro Matsumoto (Matz) in 1995.
GitHub - letta-ai/letta-code: The memory-first coding agent
2 months ago
- Letta Code is a memory-first coding harness built on the Letta API.
- It uses persistent agents that learn over time and are portable across models like Claude, GPT, Gemini, and more.
- Install via npm: `npm install -g @letta-ai/letta-code`.
- Use `/connect` to configure your own LLM API keys and `/model` to swap models.
- Agents persist across sessions, improving with use, unlike session-based CLI tools.
- Use `/init` to initialize an agent's memory system and `/remember` to guide memory updates.
- Supports skill learning via `.skills` directory and `/skill` command.
- Community packages available for Arch Linux via AUR.
GitHub - alibaba/zvec: A lightweight, lightning-fast, in-process vector database
2 months ago
- Zvec is an open-source, in-process vector database built on Proxima, designed for embedding directly into applications.
- Features include blazing-fast searches, simple setup, support for dense and sparse vectors, hybrid search, and runs anywhere.
- Requirements include Python 3.10 - 3.12, with installation via pip or npm, and supports Linux and macOS.
- Example usage includes defining a collection schema, inserting documents, and querying by vector similarity.
- Zvec is ideal for production workloads, with detailed benchmarks available in the documentation.
- Community contributions are welcome, with a Contributing Guide available for those interested.
GitHub - minio/minio: MinIO is a high-performance, S3 compatible object store, open sourced under GNU AGPLv3 license.
2 months ago
- MinIO is a high-performance, S3-compatible object storage solution under GNU AGPL v3.0.
- Designed for AI/ML, analytics, and data-intensive workloads with industry-leading performance.
- MinIO AIStor offers enterprise-grade support for commercial or proprietary usage.
- MinIO community edition is now distributed as source code only; pre-compiled binaries are no longer maintained.
- Two options to use MinIO community edition: install from source or build a Docker image.
- Compile and run MinIO server from source using Golang (minimum version go1.24).
- MinIO deployment starts with default root credentials (minioadmin:minioadmin).
- Test MinIO deployment using the MinIO Console or S3-compatible tools like MinIO Client (mc).
- Build a Docker image for MinIO using the provided Dockerfile and run it with standard Docker commands.
- Deploy MinIO on Kubernetes using the MinIO Operator or community-maintained Helm charts.
- MinIO Server includes an embedded web-based object browser accessible via http://127.0.0.1:9000.
- Use MinIO Client (mc) for managing buckets and objects with commands similar to UNIX utilities.
- MinIO contributions are guided by the MinIO Contributor's Guide.
- MinIO source is licensed under GNU AGPLv3, and documentation under CC BY 4.0.
GitHub - DebugSwift/DebugSwift: A toolkit to make debugging iOS applications easier 🚀
2 months ago
- DebugSwift is a comprehensive debugging tool for iOS development, supporting iOS 14.0+, Swift 6.0+, and Xcode 16.0+.
- Features include HTTP monitoring, WebSocket inspection, request limiting, smart content formatting, encryption support, real-time metrics, memory leak detection, and thread checking.
- Additional tools include performance widgets, crash reports, console logs, device info, APNS tokens, custom actions, and visual debugging aids like grid overlays and view hierarchy inspection.
- Supports SwiftUI render tracking (beta), file browsing, UserDefaults and Keychain inspection, database browsing, and push notification simulation.
- Easy integration via Swift Package Manager, CocoaPods, or direct XCFramework download.
- Fully supports Apple Silicon Macs with native arm64 simulator builds, eliminating the need for architecture exclusions.
- Customizable setup options allow enabling/disabling specific features, beta features, and configuring app groups for file access.
- Advanced network debugging includes URL filtering, history clearing, and automatic decryption of API responses.
- SwiftUI-specific debugging features include render tracking, overlay styles, and logging.
- Encourages community contributions and is licensed under MIT.
GitHub - THUDM/slime: slime is an LLM post-training framework for RL Scaling.
2 months ago
- slime is an LLM post-training framework for RL scaling, offering high-performance training and flexible data generation.
- It supports models like GLM-4 series, Qwen3 series, DeepSeek V3 series, and Llama 3.
- Core modules include training (Megatron), rollout (SGLang + router), and data buffer.
- slime powers projects like P1 (physics reasoning), RLVE (verifiable environments), TritonForge (GPU kernels), APRIL (rollout optimization), and qqr (agent evolution).
- Arguments are categorized into Megatron, SGLang, and slime-specific, with detailed usage documentation available.
- Contributions are welcome, with guidelines for code style consistency and debugging provided.
- Special thanks to projects like SGLang, Megatron-LM, and others, with citation instructions included.
GitHub - google-deepmind/superhuman
2 months ago
- Hosted by Google DeepMind's Superhuman Reasoning team, led by Thang Luong.
- Includes Nature paper and 2024 IMO-silver achievement details.
- Features advanced benchmarks for AI mathematical reasoning evaluation: IMO-AnswerBench, IMO-ProofBench, IMO-GradingBench.
- Introduces a math research agent powered by Gemini Deep Think for iterative solution generation and verification.
- Software licensed under Apache License 2.0; other materials under CC-BY 4.0.
- Disclaimer: Not an official Google product.
GitHub - TelegramMessenger/MTProxy
2 months ago
- Install dependencies like git, curl, build-essential, libssl-dev, and zlib1g-dev on Debian/Ubuntu or openssl-devel, zlib-devel, and Development Tools on CentOS/RHEL.
- Clone the MTProxy repository and build it using the 'make' command.
- Obtain proxy-secret and proxy-multi.conf files from Telegram's core website.
- Generate a secret for users to connect to your proxy using '/dev/urandom' and 'xxd'.
- Run mtproto-proxy with specified parameters including secrets, ports, and worker count.
- Generate a proxy link or use @MTProxybot to register your proxy.
- Enable random padding for packets by adding 'dd' prefix to the secret to avoid ISP detection.
- Create and configure a systemd service file for MTProxy to ensure it runs at startup.
- Use Telegram's official Docker image, noting it may be outdated.
GitHub - SynkraAI/aios-core: Synkra AIOS: AI-Orchestrated System for Full Stack Development - Core Framework v4.0
2 months ago
- Synkra AIOS is an AI-powered self-modifying development framework based on Agentic Agile Development.
- It offers revolutionary capabilities for AI-driven development across various domains like software development, entertainment, creative writing, business strategy, and personal well-being.
- The framework follows a clear hierarchy: CLI First → Observability Second → UI Third.
- Key principles include CLI as the source of truth, new features must work via CLI before UI, and observability is for understanding, not controlling the CLI.
- Two key innovations: Agentic Planning (collaborative agents create detailed PRD and Architecture documents) and Contextualized Development Engineering (agents transform plans into hyper-detailed development stories).
- Installation is simple with a single command: `npx aios-core init meu-projeto`.
- The framework includes an interactive installer with modern features like real-time validation, progress indicators, and package manager selection.
- Synkra AIOS comes with 11 specialized agents including analyst, pm, architect, sm, dev, and qa.
- The Autonomous Development Engine (ADE) transforms requirements into functional code through a 7-epic system.
- Expansion packs (Squads) allow extending AIOS to any domain with modular teams of AI agents.
- AIOS Pro offers advanced features for larger teams, including advanced squads, memory layer, and enterprise integrations.
- The framework implements a multi-layered validation system for code quality and consistency.
- Community contributions are welcomed through forks, branches, and pull requests.
GitHub - cinnyapp/cinny: Yet another matrix client
2 months ago
- Cinny is a Matrix client with a focus on simple, elegant, and secure interfaces.
- Web app available at app.cinny.in, with dev branch at dev.cinny.in (may be unstable).
- Desktop app can be downloaded from the cinny-desktop repository.
- Self-hosting options include downloading a tarball from GitHub or using Docker images from DockerHub/GitHub Container Registry.
- Default homeservers and explore pages are defined in config.json.
- Redirects must be set up for assets; example configurations provided for Netlify, Nginx, and Caddy.
- Hash routing can be enabled to avoid configuring redirects (URLs will include /#/).
- For subdirectory deployment, rebuild the app after updating the base path in build.config.ts.
- PGP public key provided for tarball verification.
- Recommended to use a version manager like NVM due to frequent version changes.
- Development server can be started with 'npm ci' followed by 'npm start'.
- Build the app with 'npm run build' to compile into the dist/ directory.
- Dockerfile included for building and serving the app with Nginx on port 80.
- Run the Docker container with 'docker run -p 8080:80 cinny:latest' and access at http://localhost:8080.
GitHub - unslothai/unsloth: Fine-tuning & Reinforcement Learning for LLMs. 🦥 Train OpenAI gpt-oss, DeepSeek, Qwen, Llama, Gemma, TTS 2x faster with 70% less VRAM.
2 months ago
- Notebooks are beginner-friendly, allowing users to add datasets, run, and deploy trained models.
- Performance comparison of various models (e.g., gpt-oss, Qwen3, Gemma 3) showing speed and memory improvements.
- Unsloth supports faster embedding fine-tuning (~1.8-3.3x) and new batching algorithms for longer context RL.
- New RoPE & MLP Triton Kernels & Padding Free + Packing offer 3x faster training and 30% less VRAM.
- Training a 20B model with >500K context is now possible on an 80GB GPU.
- FP8 Reinforcement Learning is now supported on consumer GPUs.
- DeepSeek-OCR improves language understanding by 89%.
- Unsloth Docker image simplifies setup and environment issues.
- Vision RL now supports training VLMs with GRPO or GSPO.
- Quantization-Aware Training recovers ~70% accuracy.
- Memory-efficient RL introduces faster RL with 50% less VRAM and 10× more context.
- Support for Mistral 3, Gemma 3n, Qwen3, and other models.
- Dynamic 2.0 quants set new benchmarks on 5-shot MMLU & Aider Polyglot.
- Unsloth supports all models (TTS, BERT, Mamba), FFT, and MultiGPU.
- Long-context Reasoning (GRPO) allows training reasoning models with just 5GB VRAM.
- Unsloth Dynamic 4-bit Quantization increases accuracy with <10% more VRAM than BnB 4-bit.
- Support for Llama 4, Phi-4, Vision models, and Llama 3.3 (70B).
- Cut Cross Entropy supports 89K context for Llama 3.3 (70B) on an 80GB GPU.
- Memory usage cut by 30%, supporting 4x longer context windows.
- Installation guides for pip, Conda, and Docker.
- Example code for fine-tuning gpt-oss-20b provided.
- RL support includes GRPO, GSPO, FP8 training, DrGRPO, DAPO, PPO, and more.
- Benchmarks show Unsloth's speed, VRAM reduction, and longer context capabilities.
- Citations and acknowledgments for contributors and libraries used.
GitHub - danielmiessler/Personal_AI_Infrastructure: Agentic AI Infrastructure for magnifying HUMAN capabilities.
2 months ago
- PAI is a Personalized AI Platform designed to magnify human capabilities by focusing on user goals, preferences, and continuous learning.
- PAI's mission is to activate people by helping them identify and pursue their purpose through AI-augmented self-discovery and making high-quality AI accessible to everyone.
- PAI differentiates itself from other AI systems by being goal-oriented, pursuing optimal output, and continuously learning from user interactions.
- The platform is built on 16 principles, including user centricity, clear thinking, deterministic infrastructure, and continuous learning.
- PAI's architecture consists of primitives like persistent memory, custom skills, and intelligent routing, making it a complete system rather than just a tool.
- PAI is modular, with capabilities distributed as Packs—self-contained, AI-installable modules that can be added incrementally.
- Installation options include a full release for a complete setup, a bundle for customization, or individual packs for specific needs.
- PAI is open-source and platform-agnostic, working with various AI systems like Claude Code, OpenCode, and custom setups.
- Future roadmap includes features like local model support, granular model routing, remote access, and outbound phone calling.
- PAI is actively developed, with frequent updates and a community-driven approach to contributions and improvements.
GitHub - patchy631/ai-engineering-hub: In-depth tutorials on LLMs, RAGs and real-world AI agent applications.
2 months ago
- Comprehensive resource for learning and building with AI
- 93+ Production-Ready Projects across all skill levels
- In-depth tutorials on LLMs, RAG, Agents, and more
- Real-world AI agent applications
- Examples to implement, adapt, and scale in projects
- Resources for beginners, practitioners, and researchers
- FREE Data Science eBook with 150+ essential lessons
- Projects categorized into Beginner, Intermediate, and Advanced levels
- Multi-component systems and agentic workflows for experienced practitioners
- Complex systems, fine-tuning, and production deployments for advanced users
- Open for contributions with guidelines on how to get involved
- Licensed under the MIT License
GitHub - Jeffallan/claude-skills: 65 Specialized Skills for Full-Stack Developers. Transform Claude Code into your expert pair programmer.
2 months ago
- 66 specialized skills across 12 categories including languages, frameworks, infrastructure, APIs, testing, DevOps, security, data/ML, and platform specialists.
- Skills activate automatically based on requests, e.g., NestJS Expert for backend development or React Expert for frontend tasks.
- Complex tasks combine multiple skills like Feature Forge, Architecture Designer, Fullstack Guardian, Test Master, and DevOps Engineer.
- Includes context engineering with /common-ground to validate Claude's hidden assumptions.
- Requires Atlassian MCP server for workflow commands.
- Comprehensive documentation available: Quick Start Guide, Skills Guide, Common Ground Guide, Workflow Commands Guide, and more.
- Open-source project under MIT License, with contribution guidelines and community support via GitHub Issues and Discussions.

first prev4next

About|Login

Bilingual

GitHub - seerr-team/seerr: Open-source media request and discovery manager for Jellyfin, Plex, and Emby.
2 months ago
- Seerr is a free, open-source media request management tool.
- Integrates with Jellyfin, Plex, Emby, Sonarr, and Radarr.
- Supports PostgreSQL and SQLite databases.
- Manages requests for Movies, Shows, and Mixed Libraries.
- Features customizable request system and user-friendly UI.
- Includes granular permissions and notification agents.
- Mobile-friendly design for on-the-go management.
- Supports watchlisting and blocklisting media.
- Documentation and community support available via Discord and GitHub.
- API documentation accessible locally.
GitHub - maxritter/claude-pilot: Claude Code is powerful. Pilot makes it reliable. Start a task, grab a coffee, come back to production-grade code. Tests enforced. Context preserved. Quality automated
2 months ago
- Claude Pilot is a tool designed to enforce production-grade code quality by integrating linting, formatting, and type checking as mandatory hooks.
- It ensures Test-Driven Development (TDD) is followed strictly with a RED → GREEN → REFACTOR cycle for every feature.
- Endless Mode maintains context across sessions, preventing degradation and enabling seamless continuation of work.
- The tool includes verifier sub-agents that perform independent code reviews before marking tasks as complete.
- Pilot is optimized for output quality with minimal system complexity, avoiding bloated prompts and unnecessary scaffolding.
- It supports quick mode for direct execution and /spec mode for structured planning, implementation, and verification.
- Pilot works with existing projects without restructuring them, adapting to project conventions via the /sync command.
- The tool is available for macOS, Linux, and Windows (WSL2) and can be installed via a curl command.
- Pilot uses different Claude models (Opus and Sonnet) based on the phase of development to balance reasoning quality and speed.
- It includes persistent memory, team collaboration features via a private Git repository, and extensive customization options.
- All development tools run locally, ensuring no code or project data leaves the machine, making it enterprise-compliant for data privacy.
- Pilot is source-available under a commercial license, with different tiers for solo developers and teams.
GitHub - brave/brave-browser: Brave browser for Android, iOS, Linux, macOS, Windows.
2 months ago
- Repository contains build tools for Brave desktop browser on macOS, Windows, and Linux.
- Fetches and syncs code from Chromium and brave-core projects.
- Supports building for Android and iOS with specific commands.
- Offers different build types: Component, Release, Static, and Debug.
- Provides instructions for syncing code, applying patches, and updating dependencies.
- Includes guidelines for Google Safe Browsing API key setup.
- Lists security rules and IPC review guidelines from Chromium.
- Encourages community involvement through Q&A and translations.
- Provides troubleshooting solutions for common problems.
GitHub - moonshine-ai/moonshine: Fast and accurate automatic speech recognition (ASR) for edge devices
2 months ago
- Moonshine Voice is an open-source AI toolkit for real-time voice applications, running on-device for privacy and speed.
- Optimized for low latency, Moonshine processes audio while the user is speaking, improving responsiveness.
- Offers higher accuracy than Whisper Large V3 with models as small as 26MB, suitable for constrained devices.
- Cross-platform support includes Python, iOS, Android, MacOS, Linux, Windows, Raspberry Pi, IoT devices, and wearables.
- Includes high-level APIs for transcription, speaker identification (diarization), and command recognition, simplifying voice application development.
- Supports multiple languages: English, Spanish, Mandarin, Japanese, Korean, Vietnamese, Ukrainian, and Arabic.
- Community support is available via Discord, with examples and documentation to help developers get started.
- Moonshine outperforms Whisper in live speech scenarios with flexible input windows and caching for streaming, reducing latency.
- Language-specific models provide higher accuracy by focusing on one language, unlike Whisper's multilingual approach.
- The API is designed for ease of use, abstracting complex details to allow developers to focus on application logic.
GitHub - steipete/gogcli: Google Suite CLI: Gmail, GCal, GDrive, GContacts.
2 months ago
- gogcli is a fast, script-friendly CLI for various Google services including Gmail, Calendar, Drive, Docs, Sheets, and more.
- Supports JSON-first output, multiple accounts, and least-privilege authentication.
- Features include email tracking, calendar event management, document export, and task management.
- Allows secure credential storage using OS keyring or encrypted on-disk keyring.
- Supports auto-refreshing tokens and least-privilege auth with --readonly and --drive-scope options.
- Can be installed via Homebrew, AUR, or from source.
- Includes commands for managing OAuth credentials, service accounts, and domain-wide delegation.
- Offers detailed documentation for setup, usage, and integration testing.
GitHub - nautechsystems/nautilus_trader: A high-performance algorithmic trading platform and event-driven backtester
2 months ago
- NautilusTrader is an open-source, high-performance algorithmic trading platform.
- It supports backtesting and live trading with no code changes required.
- The platform is AI-first, designed for Python-native environments.
- Core components are written in Rust for performance and safety.
- Supports multiple asset classes including FX, Crypto, Futures, and more.
- Features include high-frequency trading, advanced order types, and multi-venue capabilities.
- Available on Linux, macOS, and Windows with Docker deployment options.
- Includes modular adapters for integrating various trading venues and data providers.
- Offers high-precision and standard-precision modes for core value types.
- Community contributions are welcome under the GNU Lesser General Public License v3.0.
GitHub - ruby/ruby: The Ruby Programming Language
2 months ago
- Ruby is an interpreted object-oriented programming language often used for web development.
- It offers scripting features for text processing, file management, and system tasks.
- Key features include simple syntax, object-oriented capabilities, operator overloading, and garbage collection.
- Ruby is highly portable, working on Unix-like systems, Windows, and macOS.
- Installation options and source code are available on the Ruby official website.
- Development branches and source code can be accessed via Git repositories.
- Community support includes mailing lists, Stack Overflow, and a bug tracking system.
- Ruby was created by Yukihiro Matsumoto (Matz) in 1995.
GitHub - letta-ai/letta-code: The memory-first coding agent
2 months ago
- Letta Code is a memory-first coding harness built on the Letta API.
- It uses persistent agents that learn over time and are portable across models like Claude, GPT, Gemini, and more.
- Install via npm: `npm install -g @letta-ai/letta-code`.
- Use `/connect` to configure your own LLM API keys and `/model` to swap models.
- Agents persist across sessions, improving with use, unlike session-based CLI tools.
- Use `/init` to initialize an agent's memory system and `/remember` to guide memory updates.
- Supports skill learning via `.skills` directory and `/skill` command.
- Community packages available for Arch Linux via AUR.
GitHub - alibaba/zvec: A lightweight, lightning-fast, in-process vector database
2 months ago
- Zvec is an open-source, in-process vector database built on Proxima, designed for embedding directly into applications.
- Features include blazing-fast searches, simple setup, support for dense and sparse vectors, hybrid search, and runs anywhere.
- Requirements include Python 3.10 - 3.12, with installation via pip or npm, and supports Linux and macOS.
- Example usage includes defining a collection schema, inserting documents, and querying by vector similarity.
- Zvec is ideal for production workloads, with detailed benchmarks available in the documentation.
- Community contributions are welcome, with a Contributing Guide available for those interested.
GitHub - minio/minio: MinIO is a high-performance, S3 compatible object store, open sourced under GNU AGPLv3 license.
2 months ago
- MinIO is a high-performance, S3-compatible object storage solution under GNU AGPL v3.0.
- Designed for AI/ML, analytics, and data-intensive workloads with industry-leading performance.
- MinIO AIStor offers enterprise-grade support for commercial or proprietary usage.
- MinIO community edition is now distributed as source code only; pre-compiled binaries are no longer maintained.
- Two options to use MinIO community edition: install from source or build a Docker image.
- Compile and run MinIO server from source using Golang (minimum version go1.24).
- MinIO deployment starts with default root credentials (minioadmin:minioadmin).
- Test MinIO deployment using the MinIO Console or S3-compatible tools like MinIO Client (mc).
- Build a Docker image for MinIO using the provided Dockerfile and run it with standard Docker commands.
- Deploy MinIO on Kubernetes using the MinIO Operator or community-maintained Helm charts.
- MinIO Server includes an embedded web-based object browser accessible via http://127.0.0.1:9000.
- Use MinIO Client (mc) for managing buckets and objects with commands similar to UNIX utilities.
- MinIO contributions are guided by the MinIO Contributor's Guide.
- MinIO source is licensed under GNU AGPLv3, and documentation under CC BY 4.0.
GitHub - DebugSwift/DebugSwift: A toolkit to make debugging iOS applications easier 🚀
2 months ago
- DebugSwift is a comprehensive debugging tool for iOS development, supporting iOS 14.0+, Swift 6.0+, and Xcode 16.0+.
- Features include HTTP monitoring, WebSocket inspection, request limiting, smart content formatting, encryption support, real-time metrics, memory leak detection, and thread checking.
- Additional tools include performance widgets, crash reports, console logs, device info, APNS tokens, custom actions, and visual debugging aids like grid overlays and view hierarchy inspection.
- Supports SwiftUI render tracking (beta), file browsing, UserDefaults and Keychain inspection, database browsing, and push notification simulation.
- Easy integration via Swift Package Manager, CocoaPods, or direct XCFramework download.
- Fully supports Apple Silicon Macs with native arm64 simulator builds, eliminating the need for architecture exclusions.
- Customizable setup options allow enabling/disabling specific features, beta features, and configuring app groups for file access.
- Advanced network debugging includes URL filtering, history clearing, and automatic decryption of API responses.
- SwiftUI-specific debugging features include render tracking, overlay styles, and logging.
- Encourages community contributions and is licensed under MIT.
GitHub - THUDM/slime: slime is an LLM post-training framework for RL Scaling.
2 months ago
- slime is an LLM post-training framework for RL scaling, offering high-performance training and flexible data generation.
- It supports models like GLM-4 series, Qwen3 series, DeepSeek V3 series, and Llama 3.
- Core modules include training (Megatron), rollout (SGLang + router), and data buffer.
- slime powers projects like P1 (physics reasoning), RLVE (verifiable environments), TritonForge (GPU kernels), APRIL (rollout optimization), and qqr (agent evolution).
- Arguments are categorized into Megatron, SGLang, and slime-specific, with detailed usage documentation available.
- Contributions are welcome, with guidelines for code style consistency and debugging provided.
- Special thanks to projects like SGLang, Megatron-LM, and others, with citation instructions included.
GitHub - google-deepmind/superhuman
2 months ago
- Hosted by Google DeepMind's Superhuman Reasoning team, led by Thang Luong.
- Includes Nature paper and 2024 IMO-silver achievement details.
- Features advanced benchmarks for AI mathematical reasoning evaluation: IMO-AnswerBench, IMO-ProofBench, IMO-GradingBench.
- Introduces a math research agent powered by Gemini Deep Think for iterative solution generation and verification.
- Software licensed under Apache License 2.0; other materials under CC-BY 4.0.
- Disclaimer: Not an official Google product.
GitHub - TelegramMessenger/MTProxy
2 months ago
- Install dependencies like git, curl, build-essential, libssl-dev, and zlib1g-dev on Debian/Ubuntu or openssl-devel, zlib-devel, and Development Tools on CentOS/RHEL.
- Clone the MTProxy repository and build it using the 'make' command.
- Obtain proxy-secret and proxy-multi.conf files from Telegram's core website.
- Generate a secret for users to connect to your proxy using '/dev/urandom' and 'xxd'.
- Run mtproto-proxy with specified parameters including secrets, ports, and worker count.
- Generate a proxy link or use @MTProxybot to register your proxy.
- Enable random padding for packets by adding 'dd' prefix to the secret to avoid ISP detection.
- Create and configure a systemd service file for MTProxy to ensure it runs at startup.
- Use Telegram's official Docker image, noting it may be outdated.
GitHub - SynkraAI/aios-core: Synkra AIOS: AI-Orchestrated System for Full Stack Development - Core Framework v4.0
2 months ago
- Synkra AIOS is an AI-powered self-modifying development framework based on Agentic Agile Development.
- It offers revolutionary capabilities for AI-driven development across various domains like software development, entertainment, creative writing, business strategy, and personal well-being.
- The framework follows a clear hierarchy: CLI First → Observability Second → UI Third.
- Key principles include CLI as the source of truth, new features must work via CLI before UI, and observability is for understanding, not controlling the CLI.
- Two key innovations: Agentic Planning (collaborative agents create detailed PRD and Architecture documents) and Contextualized Development Engineering (agents transform plans into hyper-detailed development stories).
- Installation is simple with a single command: `npx aios-core init meu-projeto`.
- The framework includes an interactive installer with modern features like real-time validation, progress indicators, and package manager selection.
- Synkra AIOS comes with 11 specialized agents including analyst, pm, architect, sm, dev, and qa.
- The Autonomous Development Engine (ADE) transforms requirements into functional code through a 7-epic system.
- Expansion packs (Squads) allow extending AIOS to any domain with modular teams of AI agents.
- AIOS Pro offers advanced features for larger teams, including advanced squads, memory layer, and enterprise integrations.
- The framework implements a multi-layered validation system for code quality and consistency.
- Community contributions are welcomed through forks, branches, and pull requests.
GitHub - cinnyapp/cinny: Yet another matrix client
2 months ago
- Cinny is a Matrix client with a focus on simple, elegant, and secure interfaces.
- Web app available at app.cinny.in, with dev branch at dev.cinny.in (may be unstable).
- Desktop app can be downloaded from the cinny-desktop repository.
- Self-hosting options include downloading a tarball from GitHub or using Docker images from DockerHub/GitHub Container Registry.
- Default homeservers and explore pages are defined in config.json.
- Redirects must be set up for assets; example configurations provided for Netlify, Nginx, and Caddy.
- Hash routing can be enabled to avoid configuring redirects (URLs will include /#/).
- For subdirectory deployment, rebuild the app after updating the base path in build.config.ts.
- PGP public key provided for tarball verification.
- Recommended to use a version manager like NVM due to frequent version changes.
- Development server can be started with 'npm ci' followed by 'npm start'.
- Build the app with 'npm run build' to compile into the dist/ directory.
- Dockerfile included for building and serving the app with Nginx on port 80.
- Run the Docker container with 'docker run -p 8080:80 cinny:latest' and access at http://localhost:8080.
GitHub - unslothai/unsloth: Fine-tuning & Reinforcement Learning for LLMs. 🦥 Train OpenAI gpt-oss, DeepSeek, Qwen, Llama, Gemma, TTS 2x faster with 70% less VRAM.
2 months ago
- Notebooks are beginner-friendly, allowing users to add datasets, run, and deploy trained models.
- Performance comparison of various models (e.g., gpt-oss, Qwen3, Gemma 3) showing speed and memory improvements.
- Unsloth supports faster embedding fine-tuning (~1.8-3.3x) and new batching algorithms for longer context RL.
- New RoPE & MLP Triton Kernels & Padding Free + Packing offer 3x faster training and 30% less VRAM.
- Training a 20B model with >500K context is now possible on an 80GB GPU.
- FP8 Reinforcement Learning is now supported on consumer GPUs.
- DeepSeek-OCR improves language understanding by 89%.
- Unsloth Docker image simplifies setup and environment issues.
- Vision RL now supports training VLMs with GRPO or GSPO.
- Quantization-Aware Training recovers ~70% accuracy.
- Memory-efficient RL introduces faster RL with 50% less VRAM and 10× more context.
- Support for Mistral 3, Gemma 3n, Qwen3, and other models.
- Dynamic 2.0 quants set new benchmarks on 5-shot MMLU & Aider Polyglot.
- Unsloth supports all models (TTS, BERT, Mamba), FFT, and MultiGPU.
- Long-context Reasoning (GRPO) allows training reasoning models with just 5GB VRAM.
- Unsloth Dynamic 4-bit Quantization increases accuracy with <10% more VRAM than BnB 4-bit.
- Support for Llama 4, Phi-4, Vision models, and Llama 3.3 (70B).
- Cut Cross Entropy supports 89K context for Llama 3.3 (70B) on an 80GB GPU.
- Memory usage cut by 30%, supporting 4x longer context windows.
- Installation guides for pip, Conda, and Docker.
- Example code for fine-tuning gpt-oss-20b provided.
- RL support includes GRPO, GSPO, FP8 training, DrGRPO, DAPO, PPO, and more.
- Benchmarks show Unsloth's speed, VRAM reduction, and longer context capabilities.
- Citations and acknowledgments for contributors and libraries used.
GitHub - danielmiessler/Personal_AI_Infrastructure: Agentic AI Infrastructure for magnifying HUMAN capabilities.
2 months ago
- PAI is a Personalized AI Platform designed to magnify human capabilities by focusing on user goals, preferences, and continuous learning.
- PAI's mission is to activate people by helping them identify and pursue their purpose through AI-augmented self-discovery and making high-quality AI accessible to everyone.
- PAI differentiates itself from other AI systems by being goal-oriented, pursuing optimal output, and continuously learning from user interactions.
- The platform is built on 16 principles, including user centricity, clear thinking, deterministic infrastructure, and continuous learning.
- PAI's architecture consists of primitives like persistent memory, custom skills, and intelligent routing, making it a complete system rather than just a tool.
- PAI is modular, with capabilities distributed as Packs—self-contained, AI-installable modules that can be added incrementally.
- Installation options include a full release for a complete setup, a bundle for customization, or individual packs for specific needs.
- PAI is open-source and platform-agnostic, working with various AI systems like Claude Code, OpenCode, and custom setups.
- Future roadmap includes features like local model support, granular model routing, remote access, and outbound phone calling.
- PAI is actively developed, with frequent updates and a community-driven approach to contributions and improvements.
GitHub - patchy631/ai-engineering-hub: In-depth tutorials on LLMs, RAGs and real-world AI agent applications.
2 months ago
- Comprehensive resource for learning and building with AI
- 93+ Production-Ready Projects across all skill levels
- In-depth tutorials on LLMs, RAG, Agents, and more
- Real-world AI agent applications
- Examples to implement, adapt, and scale in projects
- Resources for beginners, practitioners, and researchers
- FREE Data Science eBook with 150+ essential lessons
- Projects categorized into Beginner, Intermediate, and Advanced levels
- Multi-component systems and agentic workflows for experienced practitioners
- Complex systems, fine-tuning, and production deployments for advanced users
- Open for contributions with guidelines on how to get involved
- Licensed under the MIT License
GitHub - Jeffallan/claude-skills: 65 Specialized Skills for Full-Stack Developers. Transform Claude Code into your expert pair programmer.
2 months ago
- 66 specialized skills across 12 categories including languages, frameworks, infrastructure, APIs, testing, DevOps, security, data/ML, and platform specialists.
- Skills activate automatically based on requests, e.g., NestJS Expert for backend development or React Expert for frontend tasks.
- Complex tasks combine multiple skills like Feature Forge, Architecture Designer, Fullstack Guardian, Test Master, and DevOps Engineer.
- Includes context engineering with /common-ground to validate Claude's hidden assumptions.
- Requires Atlassian MCP server for workflow commands.
- Comprehensive documentation available: Quick Start Guide, Skills Guide, Common Ground Guide, Workflow Commands Guide, and more.
- Open-source project under MIT License, with contribution guidelines and community support via GitHub Issues and Discussions.

Hasty Briefsbeta

GitHub - seerr-team/seerr: Open-source media request and discovery manager for Jellyfin, Plex, and Emby.

GitHub - maxritter/claude-pilot: Claude Code is powerful. Pilot makes it reliable. Start a task, grab a coffee, come back to production-grade code. Tests enforced. Context preserved. Quality automated

GitHub - brave/brave-browser: Brave browser for Android, iOS, Linux, macOS, Windows.

GitHub - moonshine-ai/moonshine: Fast and accurate automatic speech recognition (ASR) for edge devices

GitHub - steipete/gogcli: Google Suite CLI: Gmail, GCal, GDrive, GContacts.

GitHub - nautechsystems/nautilus_trader: A high-performance algorithmic trading platform and event-driven backtester

GitHub - ruby/ruby: The Ruby Programming Language

GitHub - letta-ai/letta-code: The memory-first coding agent

GitHub - alibaba/zvec: A lightweight, lightning-fast, in-process vector database

GitHub - minio/minio: MinIO is a high-performance, S3 compatible object store, open sourced under GNU AGPLv3 license.

GitHub - DebugSwift/DebugSwift: A toolkit to make debugging iOS applications easier 🚀

GitHub - THUDM/slime: slime is an LLM post-training framework for RL Scaling.

GitHub - google-deepmind/superhuman

GitHub - TelegramMessenger/MTProxy

GitHub - SynkraAI/aios-core: Synkra AIOS: AI-Orchestrated System for Full Stack Development - Core Framework v4.0

GitHub - cinnyapp/cinny: Yet another matrix client

GitHub - unslothai/unsloth: Fine-tuning & Reinforcement Learning for LLMs. 🦥 Train OpenAI gpt-oss, DeepSeek, Qwen, Llama, Gemma, TTS 2x faster with 70% less VRAM.

GitHub - danielmiessler/Personal_AI_Infrastructure: Agentic AI Infrastructure for magnifying HUMAN capabilities.

GitHub - patchy631/ai-engineering-hub: In-depth tutorials on LLMs, RAGs and real-world AI agent applications.

GitHub - Jeffallan/claude-skills: 65 Specialized Skills for Full-Stack Developers. Transform Claude Code into your expert pair programmer.

GitHub - seerr-team/seerr: Open-source media request and discovery manager for Jellyfin, Plex, and Emby.

GitHub - maxritter/claude-pilot: Claude Code is powerful. Pilot makes it reliable. Start a task, grab a coffee, come back to production-grade code. Tests enforced. Context preserved. Quality automated

GitHub - brave/brave-browser: Brave browser for Android, iOS, Linux, macOS, Windows.

GitHub - moonshine-ai/moonshine: Fast and accurate automatic speech recognition (ASR) for edge devices

GitHub - steipete/gogcli: Google Suite CLI: Gmail, GCal, GDrive, GContacts.

GitHub - nautechsystems/nautilus_trader: A high-performance algorithmic trading platform and event-driven backtester

GitHub - ruby/ruby: The Ruby Programming Language

GitHub - letta-ai/letta-code: The memory-first coding agent

GitHub - alibaba/zvec: A lightweight, lightning-fast, in-process vector database

GitHub - minio/minio: MinIO is a high-performance, S3 compatible object store, open sourced under GNU AGPLv3 license.

GitHub - DebugSwift/DebugSwift: A toolkit to make debugging iOS applications easier 🚀

GitHub - THUDM/slime: slime is an LLM post-training framework for RL Scaling.

GitHub - google-deepmind/superhuman

GitHub - TelegramMessenger/MTProxy

GitHub - SynkraAI/aios-core: Synkra AIOS: AI-Orchestrated System for Full Stack Development - Core Framework v4.0

GitHub - cinnyapp/cinny: Yet another matrix client

GitHub - unslothai/unsloth: Fine-tuning & Reinforcement Learning for LLMs. 🦥 Train OpenAI gpt-oss, DeepSeek, Qwen, Llama, Gemma, TTS 2x faster with 70% less VRAM.

GitHub - danielmiessler/Personal_AI_Infrastructure: Agentic AI Infrastructure for magnifying HUMAN capabilities.

GitHub - patchy631/ai-engineering-hub: In-depth tutorials on LLMs, RAGs and real-world AI agent applications.

GitHub - Jeffallan/claude-skills: 65 Specialized Skills for Full-Stack Developers. Transform Claude Code into your expert pair programmer.