Skill UI

A comprehensive guide and code implementation for scaling data enrichment pipelines to handle high volumes (10K-100K+ leads monthly). It covers capacity planning, optimizing batch processing using queues (e.g., BullMQ/Redis), managing rate limits, and designing multi-table architectures for robust, large-scale data ingestion.

Chroma is an open-source, AI-native embedding database designed for building sophisticated LLM applications. It allows users to store high-dimensional embeddings and associated metadata, facilitating powerful vector search and full-text filtering. Ideal for implementing Retrieval-Augmented Generation (RAG) systems, semantic search, and document retrieval in local or self-hosted environments, making it perfect for open-source AI projects and prototyping.

Behavioral rules for LLMs writing or reviewing code, emphasizing caution, simplicity, surgical edits, and goal-driven execution to avoid common mistakes during implementation tasks, bug fixes, or refactors.

A comprehensive guide to understanding the complete state (context) available to Large Language Models during inference. Covers core components like System Prompts, Tool Definitions, Message History, and Retrieval Augmented Generation (RAG). Learning to engineer context is crucial for building robust, efficient, and reliable AI agent systems by managing the attention budget and ensuring high-signal information delivery.

This skill details advanced techniques for enhancing the effective context capacity of LLMs. It outlines four core strategies—Compaction, Observation Masking, KV-Cache Optimization, and Context Partitioning—to prevent context limits from constraining task complexity. By strategically managing token usage, users can reduce operational costs, minimize latency, and implement robust, long-running agent systems for handling large documents or conversations.

This skill set provides advanced, structured strategies for optimizing Large Language Model (LLM) interactions. It covers techniques such as intelligent context summarization, dynamic trimming, context routing, and precise token budget allocation to prevent 'context rot'. It is essential for building robust, long-running, multi-turn conversational AI systems.

Persistent LLM conversation memory that manages short-term, long-term, and entity-focused layers to store, retrieve, and consolidate memorable context for smarter conversational AI agents.

A comprehensive pattern for building robust and cost-efficient LLM application pipelines. This guide details combining model routing based on task complexity, immutable cost tracking, limited retry logic, and prompt caching to manage API expenditures while maintaining high quality. Essential for production systems handling variable-complexity, high-volume tasks.

A comprehensive toolkit for building production-grade AI agents focused on crypto business development. This skill guides users through the entire token listing lifecycle: autonomously discovering promising assets across multiple chains, applying a weighted 100-point scoring model, verifying safety via advanced wallet forensics, and managing outreach pipelines with necessary human oversight and LLM cascade patterns.

A protocol that provides LLM agents with persistent structural memory of a codebase. It maps the project as a directed graph, tracking not only dependencies (what imports what) but also the 'reasons' (why it connects). This allows agents to understand context, perform impact analysis, and navigate large repositories without loading the entire source code into the context window.

Provides reference architectures for building high-scale, reliable transcription systems using Deepgram. It covers four main patterns: synchronous REST for short files, async queues (BullMQ) for batch processing, WebSocket for real-time streaming, and a hybrid router for mixed workloads. Use this when designing production-grade audio processing pipelines.

Orchestrates PyTorch-native TorchTitan 4D parallelism (FSDP2, TP, PP, CP) for large-scale LLM pretraining, scaling from 8 to 512+ GPUs with Float8, torch.compile, and distributed checkpointing to boot Llama 3.1, DeepSeek V3, or custom models.