数据与 AI技能

Rewrite lengthy prompts, tool descriptions, and agent instructions into compact text while preserving requirements, exceptions, and exact values. The process records removed content and compares token counts, the units models use to measure text length.

中文简介正在准备中。当前显示英语简介。

• Shortening system prompts
• Compressing tool documentation and skill bodies
• Reducing text length for model input

Write and edit system prompts, tool instructions, and agent definitions with clear structure and explicit rule strength. It guides tag use, placement of critical constraints, concise wording, and practical examples that show how tools should be used.

中文简介正在准备中。当前显示英语简介。

• Drafting system prompts for agents
• Clarifying required and optional instructions
• Writing tool explanations with runnable examples

Examine AI product reliability, learning methods, and model limitations through frameworks drawn from Andrej Karpathy’s public statements. Assess the gap between demonstrations and real use, uneven capabilities, and where people should retain control.

中文简介正在准备中。当前显示英语简介。

• Assessing AI product reliability beyond a demo
• Planning hands-on study of neural networks
• Examining LLM strengths and failure patterns

Create a reusable SKILL.md from research into a person’s thinking or a topic’s methods. Start with a name or a broad need, then extract mental models, decision rules, expression patterns, and limitations into a working skill.

中文简介正在准备中。当前显示英语简介。

• Creating a thinking adviser based on a named person
• Finding a suitable perspective for an unclear need
• Building a skill from personal writing and decision notes

Examine AI research directions and safety strategies through thinking models drawn from Ilya Sutskever’s public statements. The skill checks current facts when needed and presents a concise judgment with an everyday analogy.

中文简介正在准备中。当前显示英语简介。

• Assessing AI research directions
• Examining model scaling and data limits
• Discussing AI safety strategies

Use the weknora CLI to manage knowledge bases and documents, search stored material, and ask questions grounded in it. This shared guide covers authentication, structured JSON results, error handling, and previews of changes.

中文简介正在准备中。当前显示英语简介。

• Setting up authentication for a WeKnora server
• Managing knowledge bases and uploaded documents
• Searching stored material and asking questions about it

Choose the right WeKnora command to answer questions from a knowledge base or retrieve source passages. Covers cited answers, custom agents, document searches, and continuing a conversation through the command-line tool.

中文简介正在准备中。当前显示英语简介。

• Ask questions with citations from a knowledge base
• Retrieve source passages for further analysis
• Find documents by title or filename

Run diffusers generation pipelines from a terminal and save their outputs. Inspect accepted inputs without downloading model weights, apply LoRA adapters during generation, run remotely through HF Jobs, and inspect software and GPU environment details.

中文简介正在准备中。当前显示英语简介。

• Running one-off generation or batch jobs from a terminal
• Checking a new model's basic operation
• Inspecting accepted inputs before running a pipeline

Forecast sales, demand, or sensor readings with Google's pretrained TimesFM model, without custom training. Check hardware and dataset memory needs, include external factors such as prices or holidays, and obtain forecasts with uncertainty ranges.

中文简介正在准备中。当前显示英语简介。

• Forecasting future sales or demand from CSV data
• Including prices and holidays in sales forecasts
• Forecasting multiple stores' time series in batches

Compose compact, structured prompts for coding, debugging, review, and research with Codex or GPT-5.4. This internal skill defines task scope, completion criteria, output format, verification steps, and evidence requirements using consistent XML tags.

中文简介正在准备中。当前显示英语简介。

• Define implementation scope and completion criteria
• Include verification steps in debugging requests
• Specify evidence requirements for code reviews

Work with BigQuery to create datasets and tables, run SQL queries, ingest data, and manage processing jobs. The skill also covers connections to BigQuery ML and Gemini for machine learning and AI analysis.

中文简介正在准备中。当前显示英语简介。

• Creating datasets and tables
• Analyzing data with SQL queries
• Ingesting data and managing processing jobs

Helps design ClickHouse tables, write analytical queries, and build data ingestion workflows. Covers batch inserts, precomputed summaries, and performance monitoring for teams working with large datasets and time-based activity analysis.

中文简介正在准备中。当前显示英语简介。

• Designing tables for large analytical datasets
• Improving slow aggregation queries
• Analyzing user retention and conversion steps

Build custom interactive charts, maps, and network diagrams with d3.js. The skill guides data preparation, styling, resizing, and interactions such as zooming and tooltips, giving detailed control over how data appears and responds to users.

中文简介正在准备中。当前显示英语简介。

• Creating charts beyond standard library options
• Exploring relationships in a network diagram
• Adding zoom and tooltips to visualizations

Set up model training with torchforge for reinforcement learning, where rewards guide improvement. Covers training configurations, custom reward and loss functions, and distributed execution with Monarch and TorchTitan, keeping algorithm experiments separate from infrastructure concerns.

中文简介正在准备中。当前显示英语简介。

• Train mathematical reasoning models with GRPO
• Experiment with custom reward and loss functions
• Scale training across multiple GPUs or nodes

Guides Python agent development with PydanticAI, covering tool calls and structured outputs checked against defined data types. Explains service injection, provider switching, streaming responses, and tests that run without calling a real language model.

中文简介正在准备中。当前显示英语简介。

• Building agents that return validated structured data
• Connecting customer support agents to order data
• Testing agent logic without real model calls

Build similarity search with FAISS to find nearby matches in large collections of numeric vectors. The skill guides index selection, memory reduction, GPU use, and saving indexes while balancing search speed and accuracy.

中文简介正在准备中。当前显示英语简介。

• Find similar items in large vector datasets
• Choose an index for search speed and accuracy
• Reduce memory use for vector search

Compress LLM weights using HQQ quantization, which stores weights with fewer bits without calibration datasets. Follow workflows for precision settings, model saving, inference integration, and LoRA fine-tuning of quantized models.

中文简介正在准备中。当前显示英语简介。

• Compressing LLMs without calibration datasets
• Serving quantized models with vLLM or HuggingFace Transformers
• Comparing bit widths and inference backends

Plan voice AI systems by comparing direct speech processing with separate transcription, language generation, and speech synthesis stages. Covers response delays, detecting when people speak, interruptions, and noise handling to guide conversational interface design.

中文简介正在准备中。当前显示英语简介。

• Compare architectures for a voice AI system
• Review delays across conversation processing stages
• Design handling for interruptions during speech

Train language models with SimPO using pairs of preferred and rejected answers. Configure preference training without a separate reference model, adapt settings for base or instruction models, and address unstable training, memory limits, or lost capabilities.

中文简介正在准备中。当前显示英语简介。

• Training a base model with preferred and rejected answers
• Further tuning an instruction model with preference data
• Adjusting training settings for math and code tasks

Plan caching for repeated LLM requests by choosing what to reuse: shared prompt beginnings, complete responses, or documents loaded in advance. Covers Anthropic prompt caching, cache invalidation, and problems caused by changing inputs or outdated answers.

中文简介正在准备中。当前显示英语简介。

• Reusing shared instructions across Claude requests
• Caching answers to repeated or similar questions
• Planning document preloading with CAG

Plan and configure large language model training with OpenRLHF to better match human preferences. Covers training method selection, distributed GPU setup, inference acceleration with vLLM, and troubleshooting memory or training stability issues.

中文简介正在准备中。当前显示英语简介。

• Choosing a method for training models on human preferences
• Setting up PPO training with a reward model
• Configuring GRPO training without a critic model

Build AI workflows with DSPy by defining inputs and outputs and combining reusable components. Create question answering systems that use retrieved documents, tool-using agents, and automatically optimized prompts, then compare results with evaluation data.

中文简介正在准备中。当前显示英语简介。

• Build question answering systems that use retrieved documents
• Optimize prompts using examples and evaluation metrics
• Combine AI tasks into reusable processing stages

Design and improve RAG systems, which retrieve documents to support language model answers. Explore meaningful document splitting, vector storage, combined keyword and semantic search, and relevance ranking while checking common causes of poor retrieval.

中文简介正在准备中。当前显示英语简介。

• Design a system that answers questions using documents
• Review document splitting when search misses relevant content
• Combine keyword search with search based on meaning

Prepare text for language models using HuggingFace Tokenizers, which splits text into model-readable units. Train custom vocabularies, process batches, track original text positions, and configure input lengths for use with transformers.

中文简介正在准备中。当前显示英语简介。

• Preparing large text collections for language models
• Training a tokenizer on domain-specific text
• Mapping model outputs back to original text