brandonhimpfen/notebooks-gallery：ML、LLM、RAG 与评估实战 Notebook 合集

A curated cookbook of practical ML, LLM, RAG, evaluation, and AI systems notebooks.

notebooks-gallery is designed as a documentation-first repository for learning, demos, reproducible workflows, and practical examples that connect machine learning concepts to production-style AI systems thinking.

What is included

Beginner-friendly ML and LLM notebooks
Local examples that do not require API keys
RAG, evaluation, prompt template, and structured output examples
Sample datasets for reproducible demos
Notebook standards and contribution guidance
A validation script for notebook hygiene

Repository structure

notebooks-gallery/
├── notebooks/
│   ├── ml-basics/
│   ├── llm-basics/
│   ├── rag/
│   ├── evaluation/
│   ├── prompt-engineering/
│   └── ai-systems/
├── datasets/sample/
├── docs/
├── scripts/
├── README.md
├── requirements.txt
└── CHANGELOG.md

Notebooks in v0.1

Notebook	Topic	Purpose
`01-ml-classification-basics.ipynb`	ML basics	Train and evaluate a small text classifier
`02-embeddings-similarity-search.ipynb`	Retrieval	Build local similarity search using TF-IDF
`03-prompt-template-basics.ipynb`	Prompting	Create reusable prompt templates
`04-simple-rag-pipeline.ipynb`	RAG	Retrieve context and format a grounded answer stub
`05-llm-output-evaluation.ipynb`	Evaluation	Validate structured outputs with deterministic checks
`06-structured-output-contracts.ipynb`	AI systems	Define and validate output contracts
`07-dataset-quality-checks.ipynb`	Data quality	Run basic dataset checks
`08-model-comparison-workflow.ipynb`	Model selection	Compare candidates with a scorecard

Quick start

git clone https://github.com/brandonhimpfen/notebooks-gallery.git
cd notebooks-gallery
python -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt
jupyter lab

On Windows PowerShell:

python -m venv .venv
.\.venv\Scripts\Activate.ps1
pip install -r requirements.txt
jupyter lab

Validate notebooks

python scripts/validate_notebooks.py

The validator checks that notebooks are valid JSON, contain at least one markdown cell and one code cell, avoid committed outputs by default, and include a top-level title.

Design principles

Keep notebooks practical and reproducible.
Prefer local examples over API-dependent demos.
Explain the system pattern, not only the code.
State limitations clearly.
Make notebooks safe to run and easy to review.

License

This project is licensed under the MIT License.