This media is not supported in your browser
VIEW IN TELEGRAM
π¦ copilotkit/opentag
Build Your Own AI Agent for Slack and Beyond with OpenTag
OpenTag is the open-source bot framework that lets you deploy your own AI agents directly into Slack, Discord, Telegram, and WhatsApp. By keeping the runtime on your own infrastructure, you maintain total control over your data, tools, and chosen AI models without worrying about per-seat pricing or vendor lock-in. It supports advanced features like generative UI for inline charts and tables, and includes built-in human-in-the-loop gates for secure actions like filing Linear tickets or managing Notion documents. Clone the repository, connect your preferred chat platform, and start building custom, self-hosted AI assistants today to automate your team's workflow.
π° https://news.ycombinator.com/item?id=48697420
π @hackernewsgithubprojects
Build Your Own AI Agent for Slack and Beyond with OpenTag
OpenTag is the open-source bot framework that lets you deploy your own AI agents directly into Slack, Discord, Telegram, and WhatsApp. By keeping the runtime on your own infrastructure, you maintain total control over your data, tools, and chosen AI models without worrying about per-seat pricing or vendor lock-in. It supports advanced features like generative UI for inline charts and tables, and includes built-in human-in-the-loop gates for secure actions like filing Linear tickets or managing Notion documents. Clone the repository, connect your preferred chat platform, and start building custom, self-hosted AI assistants today to automate your team's workflow.
π° https://news.ycombinator.com/item?id=48697420
π @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
π¦ eliasoulkadi/shokunin
Shokunin: 62 AI Agent Skills with Persistent Memory
Shokunin provides sixty-two specialized engineering skills and persistent AI memory for coding environments like OpenCode, Cursor, and Windsurf. This project bridges the gap between generic prompts and professional workflows by offering structured domain expertiseβranging from Docker and Kubernetes to frontend architecture and database managementβdirectly within your editor. By integrating ChromaDB for local vector memory and declarative self-updates, it allows AI agents to recall previous sessions and maintain consistent project context without relying on cloud services. It is a completely open-source toolkit designed to help developers standardize complex engineering tasks and improve overall coding precision.
π° https://news.ycombinator.com/item?id=48159744
π @hackernewsgithubprojects
Shokunin: 62 AI Agent Skills with Persistent Memory
Shokunin provides sixty-two specialized engineering skills and persistent AI memory for coding environments like OpenCode, Cursor, and Windsurf. This project bridges the gap between generic prompts and professional workflows by offering structured domain expertiseβranging from Docker and Kubernetes to frontend architecture and database managementβdirectly within your editor. By integrating ChromaDB for local vector memory and declarative self-updates, it allows AI agents to recall previous sessions and maintain consistent project context without relying on cloud services. It is a completely open-source toolkit designed to help developers standardize complex engineering tasks and improve overall coding precision.
π° https://news.ycombinator.com/item?id=48159744
π @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
π¦ magenta/the-infinite-crate
Build Live Music with The Infinite Crate
The Infinite Crate is the experimental digital audio workstation plugin that lets you steer live music generation through real-time text prompts. By combining a C++ JUCE engine with a responsive React interface, this project streams audio directly into your music software to create a dynamic practice partner or performance tool. You can manipulate musical output on the fly using intuitive text controls, making it easy to sample or jam in the moment. It is an impressive demonstration of integrating advanced AI models into standard audio production workflows for creative experimentation. Dive into this open-source project to start shaping your own sound.
π° https://news.ycombinator.com/item?id=47316378
π @hackernewsgithubprojects
Build Live Music with The Infinite Crate
The Infinite Crate is the experimental digital audio workstation plugin that lets you steer live music generation through real-time text prompts. By combining a C++ JUCE engine with a responsive React interface, this project streams audio directly into your music software to create a dynamic practice partner or performance tool. You can manipulate musical output on the fly using intuitive text controls, making it easy to sample or jam in the moment. It is an impressive demonstration of integrating advanced AI models into standard audio production workflows for creative experimentation. Dive into this open-source project to start shaping your own sound.
π° https://news.ycombinator.com/item?id=47316378
π @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
π¦ yuxumin/viq
ViQ: Unified Visual Representation Learning
Transform images into efficient, discrete codes with ViQ, a framework designed to bridge the gap between continuous visual encoders and text-based multimodal modeling. By utilizing a SigLIP2 vision tower paired with position-aware, head-wise Finite Scalar Quantization, this project enables images at any resolution to be represented as compact token sequences. This approach balances semantic richness with fine-grained detail, allowing you to feed visual data directly into language models like Qwen2.5 or reconstruct high-fidelity images. Because these codes can be extracted offline, you can significantly accelerate training pipelines, making ViQ a powerful tool for building more efficient and unified multimodal AI systems.
π @hackernewsgithubprojects
ViQ: Unified Visual Representation Learning
Transform images into efficient, discrete codes with ViQ, a framework designed to bridge the gap between continuous visual encoders and text-based multimodal modeling. By utilizing a SigLIP2 vision tower paired with position-aware, head-wise Finite Scalar Quantization, this project enables images at any resolution to be represented as compact token sequences. This approach balances semantic richness with fine-grained detail, allowing you to feed visual data directly into language models like Qwen2.5 or reconstruct high-fidelity images. Because these codes can be extracted offline, you can significantly accelerate training pipelines, making ViQ a powerful tool for building more efficient and unified multimodal AI systems.
π @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
π¦ drjacky/claude-android-ninja
Master Android Development with AI Agent Skills
This repository provides a comprehensive collection of structured instructions and templates designed to help AI agents build professional-grade Android applications using Kotlin and Jetpack Compose. By adopting the open Agent Skills format, it supplies developers with standardized guidance on modular architecture, Navigation3 routing, and Material 3 theming. It includes practical support for offline-first data synchronization, performance benchmarking, and rigorous testing practices using fakes and Hilt. Integrating this skill set allows development agents to generate clean, maintainable code across complex domains like media playback and secure networking, ensuring high-quality, scalable results that align with modern Android development standards.
π @hackernewsgithubprojects
Master Android Development with AI Agent Skills
This repository provides a comprehensive collection of structured instructions and templates designed to help AI agents build professional-grade Android applications using Kotlin and Jetpack Compose. By adopting the open Agent Skills format, it supplies developers with standardized guidance on modular architecture, Navigation3 routing, and Material 3 theming. It includes practical support for offline-first data synchronization, performance benchmarking, and rigorous testing practices using fakes and Hilt. Integrating this skill set allows development agents to generate clean, maintainable code across complex domains like media playback and secure networking, ensuring high-quality, scalable results that align with modern Android development standards.
π @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
π¦ xlang-ai/osworld-v2
OSWorld 2.0: Testing AI Agents on Real Desktop Tasks
OSWorld 2.0 provides a scalable framework to evaluate how well multimodal AI agents perform complex, long-horizon tasks across real-world computer environments. The repository serves as a unified benchmark where agents are tested on their ability to navigate desktop interfaces, manage files, and execute multi-application workflows as a human would. By supporting various infrastructure providers like Docker and AWS, it enables researchers to deploy consistent, reproducible testing environments. Developers can use the provided setup tools and task classes to measure an agent's success rate, helping to bridge the gap between simple model reasoning and practical, reliable desktop automation.
π @hackernewsgithubprojects
OSWorld 2.0: Testing AI Agents on Real Desktop Tasks
OSWorld 2.0 provides a scalable framework to evaluate how well multimodal AI agents perform complex, long-horizon tasks across real-world computer environments. The repository serves as a unified benchmark where agents are tested on their ability to navigate desktop interfaces, manage files, and execute multi-application workflows as a human would. By supporting various infrastructure providers like Docker and AWS, it enables researchers to deploy consistent, reproducible testing environments. Developers can use the provided setup tools and task classes to measure an agent's success rate, helping to bridge the gap between simple model reasoning and practical, reliable desktop automation.
π @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
π¦ fengdu2077/x-search-filters
Master X Search with this Simple Filter Tool
X Search Filters is the handy browser extension that helps you effortlessly build complex search queries on X without needing to memorize technical syntax. By adding a clean visual panel directly to the search bar, it allows you to toggle specific operators like date ranges, minimum engagement counts, and media types with ease. You can save your favorite searches, manage your history locally, and apply complex filters with a single click while keeping the native search bar fully functional. It is an essential tool for power users who want to find exactly what they need on the platform faster.
π @hackernewsgithubprojects
Master X Search with this Simple Filter Tool
X Search Filters is the handy browser extension that helps you effortlessly build complex search queries on X without needing to memorize technical syntax. By adding a clean visual panel directly to the search bar, it allows you to toggle specific operators like date ranges, minimum engagement counts, and media types with ease. You can save your favorite searches, manage your history locally, and apply complex filters with a single click while keeping the native search bar fully functional. It is an essential tool for power users who want to find exactly what they need on the platform faster.
π @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
π¦ yaojingang/yao-meta-skill
Yao Meta Skill: Building Governed AI Agent Skills
Yao Meta Skill is the engineering framework that finally brings rigorous governance and portability to your AI agent skills. This system transforms messy operational workflows, prompts, and transcripts into highly structured, reusable packages. By utilizing an intermediate representation and target-specific compilers, it ensures your skills perform reliably across platforms like OpenAI and Claude. With built-in tools for output evaluation, release gates, and permission probes, it prevents common failures and tracks adoption drift through a managed lifecycle. Use this system to turn your internal AI experiments into professional-grade, team-ready assets that are built to last.
π @hackernewsgithubprojects
Yao Meta Skill: Building Governed AI Agent Skills
Yao Meta Skill is the engineering framework that finally brings rigorous governance and portability to your AI agent skills. This system transforms messy operational workflows, prompts, and transcripts into highly structured, reusable packages. By utilizing an intermediate representation and target-specific compilers, it ensures your skills perform reliably across platforms like OpenAI and Claude. With built-in tools for output evaluation, release gates, and permission probes, it prevents common failures and tracks adoption drift through a managed lifecycle. Use this system to turn your internal AI experiments into professional-grade, team-ready assets that are built to last.
π @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
π¦ yaojingang/geoflow
GEOFlow: AI-Powered Multi-Site Content Engineering
GEOFlow is an open-source system designed to automate content engineering and distribution across multiple websites. It streamlines the entire workflow from AI-driven content generation and RAG-based semantic chunking to multi-platform publishing. By integrating knowledge bases, prompt management, and analytics into a centralized dashboard, it enables teams to manage and synchronize high-quality, SEO-optimized content across WordPress sites or custom GEOFlow agents. This platform simplifies complex tasks like automated scheduling, task monitoring, and cross-site theme replication. It provides a robust architecture for teams looking to manage scalable digital assets with reliable AI integration and real-time performance tracking.
π @hackernewsgithubprojects
GEOFlow: AI-Powered Multi-Site Content Engineering
GEOFlow is an open-source system designed to automate content engineering and distribution across multiple websites. It streamlines the entire workflow from AI-driven content generation and RAG-based semantic chunking to multi-platform publishing. By integrating knowledge bases, prompt management, and analytics into a centralized dashboard, it enables teams to manage and synchronize high-quality, SEO-optimized content across WordPress sites or custom GEOFlow agents. This platform simplifies complex tasks like automated scheduling, task monitoring, and cross-site theme replication. It provides a robust architecture for teams looking to manage scalable digital assets with reliable AI integration and real-time performance tracking.
π @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
π¦ yaojingang/yao-open-prompts
Master AI with Yao Open Prompts
Yao Open Prompts is the curated library that finally organizes high-quality AI instructions into actionable, professional workflows. This repository provides over one hundred tested Chinese and English prompts specifically designed for content creation, marketing strategy, learning methodologies, and everyday productivity. Instead of scattered advice, you get structured tools for tasks like generating PPTs, performing reverse engineering, or optimizing content for search visibility. By offering a clean, copy-paste format focused on real-world utility, it streamlines how you interact with AI models to achieve better results. Explore these refined templates to elevate your daily AI usage and save valuable time.
π @hackernewsgithubprojects
Master AI with Yao Open Prompts
Yao Open Prompts is the curated library that finally organizes high-quality AI instructions into actionable, professional workflows. This repository provides over one hundred tested Chinese and English prompts specifically designed for content creation, marketing strategy, learning methodologies, and everyday productivity. Instead of scattered advice, you get structured tools for tasks like generating PPTs, performing reverse engineering, or optimizing content for search visibility. By offering a clean, copy-paste format focused on real-world utility, it streamlines how you interact with AI models to achieve better results. Explore these refined templates to elevate your daily AI usage and save valuable time.
π @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
π¦ yaojingang/yao-geo-skills
Master Generative Engine Optimization with These Reusable Skill Packages
This repository provides twenty open-source skill packages designed to standardize workflows for Generative Engine Optimization. Instead of relying on scattered prompts, the project offers structured, verifiable execution assets that cover everything from strategic diagnosis and page technical audits to content production and performance tracking. Each skill includes clear input templates, evaluation rubrics, and reproducible examples to ensure quality and consistency in professional tasks. It serves as a practical toolkit for teams aiming to automate complex operations and maintain reliable results in generative search environments, ultimately turning repetitive optimization tasks into reusable and scalable professional workflows.
π @hackernewsgithubprojects
Master Generative Engine Optimization with These Reusable Skill Packages
This repository provides twenty open-source skill packages designed to standardize workflows for Generative Engine Optimization. Instead of relying on scattered prompts, the project offers structured, verifiable execution assets that cover everything from strategic diagnosis and page technical audits to content production and performance tracking. Each skill includes clear input templates, evaluation rubrics, and reproducible examples to ensure quality and consistency in professional tasks. It serves as a practical toolkit for teams aiming to automate complex operations and maintain reliable results in generative search environments, ultimately turning repetitive optimization tasks into reusable and scalable professional workflows.
π @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
π¦ ara-labs/agent-native-research-artifact
Scale Scientific Trust with Agent-Native Research Artifacts
Agent-Native Research Artifact is the ecosystem layer that finally makes AI research verifiable and observable. By providing a structured protocol and specialized agent skills, it solves the bottleneck of untraceable AI-generated scientific results. It works by capturing research journeys, crystallizing insights, and maintaining strict epistemic anchors that link claims to ground-truth evidence. You can use its built-in skills to manage sessions, compile documents, rigorously review logic, and visualize exploration graphs. This framework turns chaotic agent logs into reliable, machine-executable knowledge, ensuring that speed never compromises the integrity of your scientific work. Try it today to scale your research with absolute confidence.
π° https://news.ycombinator.com/item?id=48703174
π @hackernewsgithubprojects
Scale Scientific Trust with Agent-Native Research Artifacts
Agent-Native Research Artifact is the ecosystem layer that finally makes AI research verifiable and observable. By providing a structured protocol and specialized agent skills, it solves the bottleneck of untraceable AI-generated scientific results. It works by capturing research journeys, crystallizing insights, and maintaining strict epistemic anchors that link claims to ground-truth evidence. You can use its built-in skills to manage sessions, compile documents, rigorously review logic, and visualize exploration graphs. This framework turns chaotic agent logs into reliable, machine-executable knowledge, ensuring that speed never compromises the integrity of your scientific work. Try it today to scale your research with absolute confidence.
π° https://news.ycombinator.com/item?id=48703174
π @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
π¦ hasaneyldrm/exercises-dataset
Build Fitness Apps Fast with This Exercise Dataset
Access a massive, ready-to-use fitness dataset containing 1,324 unique exercises to power your next health or wellness project. This collection simplifies development by providing structured JSON data, including names, categories, target muscles, and equipment requirements, paired with thumbnail images and animated GIFs for every movement. It even includes bilingual instructions to help reach a broader audience. With its built-in interactive browser and developer setup guides for various programming languages, this repository is a perfect resource for building workout planners, machine learning models, or fitness prototypes. Explore the full library today to level up your application's exercise data.
π @hackernewsgithubprojects
Build Fitness Apps Fast with This Exercise Dataset
Access a massive, ready-to-use fitness dataset containing 1,324 unique exercises to power your next health or wellness project. This collection simplifies development by providing structured JSON data, including names, categories, target muscles, and equipment requirements, paired with thumbnail images and animated GIFs for every movement. It even includes bilingual instructions to help reach a broader audience. With its built-in interactive browser and developer setup guides for various programming languages, this repository is a perfect resource for building workout planners, machine learning models, or fitness prototypes. Explore the full library today to level up your application's exercise data.
π @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
π¦ heroicons-animated/heroicons-animated
Add Beautiful Animations to Heroicons in React
Enhance your user interface by integrating fluid, pre-built animations into standard iconography. This project provides a collection of beautifully animated versions of popular Heroicons, specifically designed for React applications using the Motion library. Developers can easily install these icons through a command-line interface or by adding the dedicated package to their existing projects. By simplifying the implementation of motion, the repository allows you to quickly add high-quality visual feedback to buttons and interactive elements without manual configuration. Explore the gallery to see these assets in action and bring your next design project to life.
π @hackernewsgithubprojects
Add Beautiful Animations to Heroicons in React
Enhance your user interface by integrating fluid, pre-built animations into standard iconography. This project provides a collection of beautifully animated versions of popular Heroicons, specifically designed for React applications using the Motion library. Developers can easily install these icons through a command-line interface or by adding the dedicated package to their existing projects. By simplifying the implementation of motion, the repository allows you to quickly add high-quality visual feedback to buttons and interactive elements without manual configuration. Explore the gallery to see these assets in action and bring your next design project to life.
π @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
π¦ vercel-labs/deepsec
Automate Code Security Audits with AI
Identify hidden vulnerabilities in large codebases by using this security harness powered by advanced coding agents. This tool scans your repository to surface complex issues that traditional methods often miss, allowing you to run deep, on-demand audits directly within your own infrastructure. For massive projects, it intelligently distributes tasks across worker machines, ensuring that scans remain efficient and resumable if interrupted. By leveraging specialized AI models to investigate and validate findings, it helps teams patch critical security flaws quickly. Use this system to maintain a robust security posture and catch bugs that have been lurking in your code.
π @hackernewsgithubprojects
Automate Code Security Audits with AI
Identify hidden vulnerabilities in large codebases by using this security harness powered by advanced coding agents. This tool scans your repository to surface complex issues that traditional methods often miss, allowing you to run deep, on-demand audits directly within your own infrastructure. For massive projects, it intelligently distributes tasks across worker machines, ensuring that scans remain efficient and resumable if interrupted. By leveraging specialized AI models to investigate and validate findings, it helps teams patch critical security flaws quickly. Use this system to maintain a robust security posture and catch bugs that have been lurking in your code.
π @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
π¦ st1ne/polymarket-assistant
Build a Live Crypto Trading Terminal Dashboard
Visualize real-time market trends by running this custom terminal dashboard that tracks your favorite cryptocurrencies. The project combines live Binance order flow with prediction market pricing from Polymarket to surface actionable signals directly in your command line. It automatically calculates eleven different indicators, including order book imbalance, volume profile, and technical signals like MACD or RSI, to output a clear bullish or bearish trend score. By aggregating these metrics into a polished interface, you gain a concise view of current market sentiment, helping you monitor potential price movements more effectively through a single consolidated live display.
π @hackernewsgithubprojects
Build a Live Crypto Trading Terminal Dashboard
Visualize real-time market trends by running this custom terminal dashboard that tracks your favorite cryptocurrencies. The project combines live Binance order flow with prediction market pricing from Polymarket to surface actionable signals directly in your command line. It automatically calculates eleven different indicators, including order book imbalance, volume profile, and technical signals like MACD or RSI, to output a clear bullish or bearish trend score. By aggregating these metrics into a polished interface, you gain a concise view of current market sentiment, helping you monitor potential price movements more effectively through a single consolidated live display.
π @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
π¦ gmapsscraper/google-maps-agent-skills
Automate Local Business Lead Generation with AI
Developers can now equip Claude Code and OpenClaw agents with specialized capabilities for scraping and processing Google Maps data. This collection provides modular skills that handle tasks ranging from extracting business contact information and verified emails to exporting leads into CRM formats like HubSpot. The system uses a two-layer design, combining paid API-backed skills for gathering fresh geographic data with free local-processing skills that analyze competitor trends or generate personalized cold outreach emails from existing CSV files. These tools offer a direct way to automate the entire lead generation pipeline, allowing users to move from search queries to actionable outreach.
π @hackernewsgithubprojects
Automate Local Business Lead Generation with AI
Developers can now equip Claude Code and OpenClaw agents with specialized capabilities for scraping and processing Google Maps data. This collection provides modular skills that handle tasks ranging from extracting business contact information and verified emails to exporting leads into CRM formats like HubSpot. The system uses a two-layer design, combining paid API-backed skills for gathering fresh geographic data with free local-processing skills that analyze competitor trends or generate personalized cold outreach emails from existing CSV files. These tools offer a direct way to automate the entire lead generation pipeline, allowing users to move from search queries to actionable outreach.
π @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
π¦ kstevica/captain-claw
Captain Claw: Autonomous Multi-Agent AI Orchestration
Captain Claw functions as an open-source AI agent system featuring sophisticated multi-agent orchestration and autonomous cognitive capabilities. It allows users to manage complex workflows through a centralized web dashboard, providing deep control over AI behavior. The system supports major language models and includes forty-eight built-in tools for tasks like live terminal access, file operations, and autonomous planning. Its standout feature is the ability to automatically assemble specialist councils for deliberation and execute parallel collaborative tasks. By closing the loop on notice, decision, action, and verification, the platform offers a robust environment for building reliable, self-directed AI teams.
π @hackernewsgithubprojects
Captain Claw: Autonomous Multi-Agent AI Orchestration
Captain Claw functions as an open-source AI agent system featuring sophisticated multi-agent orchestration and autonomous cognitive capabilities. It allows users to manage complex workflows through a centralized web dashboard, providing deep control over AI behavior. The system supports major language models and includes forty-eight built-in tools for tasks like live terminal access, file operations, and autonomous planning. Its standout feature is the ability to automatically assemble specialist councils for deliberation and execute parallel collaborative tasks. By closing the loop on notice, decision, action, and verification, the platform offers a robust environment for building reliable, self-directed AI teams.
π @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
π¦ open-free-llm-api/awesome-freellm-apis
Unlock 134+ Free LLM APIs with This Directory
More than one hundred and thirty free large language model APIs are now centralized in a single, machine-readable directory that eliminates the guesswork of finding active services. This project tracks over forty providers, including Google Gemini and Groq, by monitoring their free tiers daily to ensure users have accurate information on credit card requirements and rate limits. The repository offers one-click setup snippets for popular development tools like Cursor, Claude Code, and Codex, allowing developers to integrate these models seamlessly. It solves the frustration of managing fragmented provider access, ultimately providing a streamlined, reliable resource for testing AI models.
π @hackernewsgithubprojects
Unlock 134+ Free LLM APIs with This Directory
More than one hundred and thirty free large language model APIs are now centralized in a single, machine-readable directory that eliminates the guesswork of finding active services. This project tracks over forty providers, including Google Gemini and Groq, by monitoring their free tiers daily to ensure users have accurate information on credit card requirements and rate limits. The repository offers one-click setup snippets for popular development tools like Cursor, Claude Code, and Codex, allowing developers to integrate these models seamlessly. It solves the frustration of managing fragmented provider access, ultimately providing a streamlined, reliable resource for testing AI models.
π @hackernewsgithubprojects
This media is not supported in your browser
VIEW IN TELEGRAM
π¦ swarmone/agentic-swarm-bench
Stop Benchmarking LLMs Wrong: Testing Real Agentic Swarms
Agentic workflows produce LLM context windows that are eighty times larger than standard chatbot traffic, yet most benchmarks fail to simulate this reality. This repository provides an open-source tool specifically designed to benchmark LLM inference stacks under these complex, real-world coding scenarios. You can record your own live agentic sessions as replayable data, or use its synthetic modes to stress-test endpoints with growing context windows, tool definitions, and multi-turn error traces. It even tracks reasoning token overhead, ensuring you get accurate performance data. Use this tool to verify how your infrastructure actually performs during demanding, multi-turn software development.
π @hackernewsgithubprojects
Stop Benchmarking LLMs Wrong: Testing Real Agentic Swarms
Agentic workflows produce LLM context windows that are eighty times larger than standard chatbot traffic, yet most benchmarks fail to simulate this reality. This repository provides an open-source tool specifically designed to benchmark LLM inference stacks under these complex, real-world coding scenarios. You can record your own live agentic sessions as replayable data, or use its synthetic modes to stress-test endpoints with growing context windows, tool definitions, and multi-turn error traces. It even tracks reasoning token overhead, ensuring you get accurate performance data. Use this tool to verify how your infrastructure actually performs during demanding, multi-turn software development.
π @hackernewsgithubprojects