> browse/registry

> streamlit

Expert guidance for Streamlit, the Python framework for building interactive data applications and dashboards. Helps developers create web apps for data exploration, ML model demos, and internal tools using pure Python — no frontend skills required.

> e2b

You are an expert in E2B, the cloud platform for running AI-generated code in secure sandboxes. You help developers give AI agents the ability to execute code, install packages, read/write files, and run long processes in isolated cloud environments — each sandbox is a lightweight VM that boots in ~150ms with full Linux, filesystem, and networking.

> openai-agents

You are an expert in the OpenAI Agents SDK (formerly Swarm), the official framework for building multi-agent systems. You help developers create agents with tool calling, guardrails, agent handoffs, streaming, tracing, and MCP integration — building production-grade AI agents that coordinate, delegate tasks, and execute tools with built-in safety controls.

> openai-codex-cli

You are an expert in OpenAI's Codex CLI, the open-source terminal-based coding agent that reads your codebase, generates and edits code, runs shell commands, and applies changes — all within your terminal. You help developers use Codex CLI for code generation, refactoring, debugging, and automation with configurable approval modes (suggest, auto-edit, full-auto) and sandboxed execution for safety.

> modal

Serverless GPU compute platform for running Python functions in the cloud. Deploy ML models, run training jobs, and serve inference endpoints without managing infrastructure. Supports A100/H100 GPUs, custom container images, and scales to zero automatically.

> kaggle-finetune

End-to-end workflow for fine-tuning LLMs using Kaggle datasets. Use when downloading datasets from Kaggle for model training, preparing conversation/customer service data for chatbot fine-tuning, or building domain-specific AI assistants. Covers dataset discovery, download, preprocessing into chat format, and integration with PEFT/LoRA training.

> peft-fine-tuning

Parameter-efficient fine-tuning for LLMs using LoRA, QLoRA, and 25+ methods. Use when a user asks to fine-tune a language model, train a custom LLM, adapt a model to their data, use LoRA or QLoRA, fine-tune Llama or Mistral, or train a model on consumer GPUs. Covers PEFT methods for 7B-70B parameter models.

> fireworks-ai

Expert guidance for Fireworks AI, the platform for running open-source LLMs (Llama, Mixtral, Qwen, etc.) with enterprise-grade speed and reliability. Helps developers integrate Fireworks' inference API, fine-tune models, and deploy custom model endpoints with function calling and structured output support.

> comet-ml

Expert guidance for Comet ML, the platform for tracking machine learning experiments, managing models, and monitoring production ML systems. Helps developers log experiments, compare model versions, and build reproducible ML pipelines with automatic code/data versioning.

> cursor-ai

Build software with Cursor, the AI-powered code editor. Use when a user asks to configure Cursor rules, set up .cursorrules files, use Composer for multi-file edits, integrate MCP servers, or optimize AI-assisted coding workflows.

> haystack

You are an expert in Haystack, the open-source framework by deepset for building production RAG pipelines and LLM applications. You help developers create composable pipelines with document stores, retrievers, readers, generators, and custom components — connecting to 20+ LLM providers and vector databases with a pipeline-as-code approach.

> bull-mq

You are an expert in BullMQ, the high-performance job queue for Node.js built on Redis. You help developers build reliable background processing systems with delayed jobs, rate limiting, prioritization, repeatable cron jobs, job dependencies, concurrency control, and dead-letter handling — powering email sending, image processing, webhook delivery, report generation, and any async workload.

> braintrust

You are an expert in Braintrust, the evaluation and observability platform for AI applications. You help developers run systematic evaluations, compare model versions, track experiments, log production traces, and measure quality metrics — with a focus on making AI development as rigorous as traditional software testing.

> helicone

LLM observability proxy that sits between your app and LLM providers. Logs every request, enables caching, rate limiting, and provides cost analytics. Works with OpenAI, Anthropic, and other providers with a one-line integration change.

> openai-sdk

Integrate OpenAI APIs into applications. Use when a user asks to add GPT or ChatGPT to an app, generate text with OpenAI, build a chatbot, use GPT-4 or o1 models, generate embeddings, use function calling, stream chat completions, build AI features, moderate content, generate images with DALL-E, transcribe audio with Whisper API, or integrate any OpenAI model. Covers Chat Completions, Assistants API, function calling, embeddings, streaming, vision, DALL-E, Whisper, and moderation.

> instructor

You are an expert in Instructor, the library for getting structured, validated output from LLMs. You help developers extract typed data from unstructured text using Pydantic models (Python) or Zod schemas (TypeScript), with automatic retries on validation failures, streaming partial objects, and support for OpenAI, Anthropic, Google, and local models — turning LLMs into reliable data extraction engines.

> jotai

You are an expert in Jotai, the primitive and flexible state management library for React based on atomic state. You help developers build React applications with fine-grained reactivity using atoms (state primitives), derived atoms (computed values), async atoms (data fetching), and atom families — providing bottom-up state management where only components subscribing to changed atoms re-render.

> job-queue

Build job queues and background worker systems using BullMQ, Celery, or Sidekiq. Use when you need to offload slow tasks from request handlers — email sending, PDF generation, image processing, data exports, or any work that takes more than a few hundred milliseconds. Covers job priorities, concurrency control, scheduled jobs, progress tracking, and graceful shutdown. Trigger words: background job, worker, queue, async task, BullMQ, Celery, cron job, scheduled task, job retry.

> e2b-sandbox

Run AI-generated code safely in cloud sandboxes with E2B — secure execution environments for LLM agents. Use when someone asks to "run code in a sandbox", "E2B", "execute AI-generated code safely", "code interpreter for AI", "sandboxed code execution", "run untrusted code", or "give my AI agent a computer". Covers sandbox creation, code execution, file system, process management, and custom environments.

> gradio

Python library for building ML demo UIs with minimal code. Create interactive web interfaces for models with text, image, audio, and video inputs/outputs. Share demos via public links or deploy to Hugging Face Spaces.