Skill UI

This SDK enables the development of real-time, conversational AI applications using Azure Cognitive Services. It supports bidirectional communication via WebSockets, allowing for continuous audio streaming (mic input) and receiving synthesized speech output. Ideal for building voice assistants, IVR systems, and interactive voice experiences.

This skill guides the development of advanced voice agents, representing the frontier of natural human-AI interaction. It covers both low-latency Speech-to-Speech (S2S) architectures (like OpenAI Realtime API) and fully controllable STT→LLM→TTS pipelines. Key focus areas include minimizing end-to-end latency (targeting <800ms), handling conversational nuances, background noise, and managing interruptions for applications like customer support, IVR, and live translation.

An expert guide for deploying and optimizing Large Language Models (LLMs) entirely on local hardware. Mastering the entire local AI ecosystem, including inference engines like Ollama, vLLM, and llama.cpp. Provides deep knowledge on quantization formats (GGUF, EXL2, AWQ), VRAM calculation, and ensuring privacy-first, offline deployment for state-of-the-art models.

This guide provides comprehensive techniques for optimizing ElevenLabs Text-to-Speech (TTS) performance. Learn how to reduce latency and maximize throughput by strategically selecting models (e.g., Flash vs. v3), implementing streaming endpoints (HTTP or WebSocket), and choosing optimal audio formats. It is essential for real-time conversational AI, IVR systems, and high-volume content generation.

Use Voiceover Studio to convert written scripts into high-quality, professional audio narration. This tool supports diverse applications including product demos, IVR prompts, podcast episodes, and short-video voiceovers. It manages complex requirements such as specific accents, emotions, speaking pace, and precise language/locale preservation, ensuring the resulting audio is ready for immediate use.

The `hf-mem` skill estimates the necessary VRAM or system memory required to load Safetensors or GGUF model weights for inference directly from the Hugging Face Hub. It uses efficient HTTP Range requests, avoiding local downloads. Users can determine if a model fits on their GPU or instance, including optional estimations for the KV cache and setting limits like maximum sequence length.

This professional tool helps developers and researchers identify the optimal AI model for a specific task, such as coding, reasoning, or image recognition. It queries official HuggingFace leaderboards, filters models based on hardware constraints (e.g., available VRAM or RAM), and provides a detailed comparison table including benchmark scores, parameter size, and licensing information.

This skill helps users find the best AI models for specific tasks by querying Hugging Face benchmark leaderboards. It analyzes hardware constraints like GPU VRAM or CPU memory to filter compatible models. The output includes a comparison table with benchmark scores, parameter sizes, licenses, and device compatibility status. It also suggests whether models can run locally or require API access.

An expert guide for deploying and optimizing Large Language Models (LLMs) entirely on local hardware. Mastering the entire local AI ecosystem, including inference engines like Ollama, vLLM, and llama.cpp. Provides deep knowledge on quantization formats (GGUF, EXL2, AWQ), VRAM calculation, and ensuring privacy-first, offline deployment for state-of-the-art models.

Fine-tune LLMs on a single consumer GPU with Unsloth Core. Covers VRAM sizing, LoRA/QLoRA, GRPO/DPO post-training, chat-template correctness, and GGUF export.