Skill UI

Comprehensive guide for configuring Bring Your Own Key (BYOK) for major AI providers, including OpenAI, Anthropic, Google Gemini, Azure, and custom endpoints. This allows users to bypass default quotas, manage costs directly with the source provider, and utilize a wide range of specialized models within the Cursor IDE. It covers setup steps, security best practices, and cost monitoring strategies.

A comprehensive guide for selecting and optimizing embedding models for advanced vector search and retrieval-augmented generation (RAG) applications. This guide covers model comparison, chunking strategies, dimension reduction techniques, and provides runnable Python templates for both OpenAI and local Sentence Transformer implementations.

This comprehensive guide details the end-to-end process of building Retrieval-Augmented Generation (RAG) pipelines using LangChain. It covers critical steps: loading diverse document types (PDF, CSV, TXT), splitting text into manageable chunks, generating high-quality embeddings (OpenAI), storing them in various vector databases (FAISS, Pinecone), and finally constructing the full runnable chain for accurate and context-aware querying.

Track and improve production LLMs with Langfuse observability: trace API calls, manage prompts, score evaluations, integrate with LangChain/LlamaIndex/OpenAI, and keep dashboards on cost, latency, and quality.

Implement comprehensive end-to-end tracing for Large Language Model (LLM) calls, chains, and complex agents using Langfuse. This workflow demonstrates techniques like OpenAI drop-in wrappers, manual span creation for RAG pipelines, tracking streamed responses, and integrating with various LLM providers (OpenAI, Anthropic). Use this to monitor latency, track token usage, and debug complex AI features in production.

LangSmith centralizes tracing, evaluation, and monitoring for LLM applications, letting teams debug prompts, run regression tests, collect latency/cost metrics, and integrate with OpenAI, Anthropic, LangChain, or LlamaIndex to keep production systems observable.

Loki Mode is a multi-agent autonomous startup system that progresses PRD to production with zero human intervention, orchestrating research-enhanced components such as OpenAI SDK, DeepMind, Anthropic, AWS Bedrock, and agent tooling while enforcing RARV reasoning and efficiency tracking.

High-performance RLHF framework that combines Ray, vLLM, and ZeRO-3 to accelerate PPO/GRPO/RLOO/DPO training for 7B-70B+ models on distributed GPU clusters, providing a single-stack workflow for reward models and policy optimization.

This skill provides advanced strategies for managing large language model (LLM) context windows and token usage via OpenRouter. Functions include context window lookup, context-aware model selection, conversation trimming, document chunking for RAG, and prompt caching. Use this when dealing with long conversations or large documents to maximize quality while minimizing API costs and preventing context limit overflow.

Enables Claude to orchestrate multi-agent systems using AI SDK v5, setting up agents, handoffs, intelligent routing, and workflows across OpenAI, Anthropic, and Google when users request complex agent coordination.

Covers prompt and response caching plus Cache Augmented Generation patterns for LLMs such as Claude and OpenAI to reuse stable context, reduce repeat work, and cut latency/cost on recurring queries.

Generate or edit high-quality photos, illustrations, and visual assets using OpenRouter models like Gemini 3.1 Flash and Flux 2 Pro; configure the API key once, then run the scripts for prompts, edits, or batch outputs for presentations, marketing, or general creative needs.