pocket-guide¶

A complete pocket guide for data, AI and deployment work: from the terminal and Git to LLM apps, agents and cloud deployment. Numbered from basic to advanced; every guide explains what a tool is, why it exists, how it works (with mental-model diagrams) and when to use each command.

New here? Start with 00 - Big Picture (how everything connects, the journey of a request through an AI app, learning paths), then 01 - Core Concepts (the basic vocabulary: packages, APIs, SDKs, configuration, reading errors).

Start¶

#	Guide	Covers
00	Big Picture	Map of the stack, request journey, dev lifecycle, learning paths, "I want to..." finder
97	Capstone Project	Build, test, evaluate, containerise, automate and deploy a document chatbot, step by step
98	Glossary	Every key term from all guides, A to Z, linked to its guide
99	Quick Reference	The most-used commands of every guide on one page

Practice: every guide ends with a Try It section (exercises with hidden solutions), and examples/ has runnable, tested mini-projects: LLM basics, a tool-using agent, a RAG chatbot API and an MCP server.

Website: the same content as a searchable site with dark mode: https://harmandeep2993.github.io/pocket-guide/

Foundations¶

#	Guide	Covers
01	Core Concepts	Code and runtimes, terminal, paths, packages and dependencies, libraries vs frameworks, APIs, SDKs in depth, config and secrets, reading errors
02	Markdown	Headings, formatting, lists, links, images, code, tables, anchors
03	Terminal and PowerShell	Navigation, files, search, pipes, env vars, processes, network, winget, CMD/Bash equivalents
04	Linux	Navigation, files, grep/find, permissions, apt, processes, systemd, network, SSH, tar, cron
05	Git and GitHub	Commit, branch, merge/rebase, conflicts, undo, stash, .gitignore, gh CLI, PR workflow
06	VS Code	Shortcuts, multi-cursor, search, debugging, Python setup, Git, extensions, settings, remote
07	Regex	Classes, anchors, quantifiers, groups, lookarounds, common patterns, Python re, pandas, grep, SQL
08	YAML, JSON, TOML and .env	Syntax, Python parsing, jq, JSONL, JSON Schema, config formats, secrets files
09	HTTP and APIs	Methods, status codes, headers, REST, auth, curl, requests/httpx, retries, rate limits, SSE, webhooks

Python¶

#	Guide	Covers
10	Python Basics	Types, strings, f-strings, lists, dicts, loops, comprehensions, functions, errors, files, classes, logging
11	Python Virtual Environment	venv create/activate, pip, requirements.txt, VS Code, troubleshooting
12	uv	Projects, add/remove, run, lock/sync, Python versions, pip interface, uvx tools, Docker
13	Pydantic	Models, validation, constraints, validators, JSON Schema, settings, LLM structured output
14	Async Python	async/await, gather, TaskGroup, semaphores, timeouts, async HTTP and LLM calls, queues
15	pytest	Fixtures, parametrize, markers, mocking APIs and LLMs, FastAPI tests, coverage
16	Jupyter	Kernels from venv, shortcuts, magics, display options, autoreload, export, notebooks in Git

Data and Machine Learning¶

#	Guide	Covers
17	NumPy	Create arrays, indexing, filtering, reshape, math, broadcasting, axis, random, linear algebra
18	Pandas	Read/write, inspect, select, filter, clean, groupby, pivot, merge
19	Polars and DuckDB	Parquet, expressions, lazy mode, pandas translation, SQL on files and DataFrames
20	SQL	SELECT, WHERE, GROUP BY, JOINs, CTEs, window functions, DDL/DML, psql/sqlite3, SQL from pandas
21	Matplotlib	Line, scatter, bar, hist, box, pie, labels, legend, subplots, styles, save
22	Seaborn	Distribution, categorical, relationship, regression, heatmap, pairplot, facets, palettes
23	Scikit-learn	Split, preprocessing, pipelines, models, metrics, cross-validation, tuning, saving models
24	PyTorch	Tensors, GPU, autograd, nn.Module, training loop, evaluation, saving, overfitting
25	Hugging Face	Hub, pipelines, tokenizers, open LLMs, chat templates, quantization, embeddings, datasets

AI Engineering¶

#	Guide	Covers
26	LLM Fundamentals	Tokens, next-token prediction, training, context, sampling, reasoning, hallucinations, cost, model choice
27	LLM APIs	Claude SDK in depth, streaming, structured outputs, vision/PDF, thinking, caching, batches, OpenAI equivalents
28	Prompt Engineering	Clarity, context, XML tags, examples, output formats, long docs, templates, chaining, checklist
29	Tool Use	Tool definitions, the tool loop, tool runner, parallel calls, errors, server tools, tool design, safety
30	Embeddings and Vector DBs	Embedding models, similarity, ANN/HNSW, FAISS, Chroma, pgvector, Qdrant, hybrid search
31	RAG	Loading, chunking, retrieval, reranking, citations, contextual/agentic RAG, evaluation, security
32	AI Agents	Agent loop, workflows vs agents, patterns, memory, planning, multi-agent, human-in-the-loop, limits
33	Agent Frameworks	Claude Agent SDK, OpenAI Agents SDK, LangChain/LangGraph, LlamaIndex, PydanticAI, CrewAI, choosing
34	MCP	Model Context Protocol: architecture, Python servers, Claude Code/Desktop/VS Code, clients, security
35	Evals and Observability	Eval sets, graders, LLM-as-judge, CI evals, tracing, logging, cost monitoring, feedback
36	Local LLMs	Ollama in depth, hardware sizing, quantization, llama.cpp, LM Studio, vLLM, Docker
37	Fine-tuning	When to fine-tune, LoRA/QLoRA, datasets, TRL training, evaluation, GGUF export, DPO, hosted options
38	AI Security	OWASP LLM Top 10, prompt injection, excessive agency, data leakage, guardrails, regulation, red teaming
39	AI UIs	Streamlit, Gradio, Chainlit chat apps, streaming, state, secrets, FastAPI frontend, deployment

APIs and Deployment¶

#	Guide	Covers
40	FastAPI	Routes, Pydantic validation, dependencies, settings, routers, testing, ML model API, Docker
41	Uvicorn	ASGI server: running apps, reload, workers, Gunicorn, proxy headers, HTTPS, timeouts, logging, Docker, systemd
42	Redis and Task Queues	Caching, LLM response cache, rate limiting, sessions, RQ, Celery, arq, job status pattern
43	Docker	Images, containers, run options, Dockerfile, volumes, networks, Compose, cleanup, registry
44	GitHub Actions	Workflows, triggers, Python CI with uv, secrets, caching, evals in CI, Docker builds, Azure OIDC deploy
45	Nginx and HTTPS	Reverse proxy, Let's Encrypt, streaming/WebSockets, basic auth, rate limits, systemd, Caddy
46	Kubernetes	Pods, Deployments, Services, Ingress, config, probes, scaling, rollouts, GPUs, Helm, AKS
47	Terraform	HCL, providers, resources, variables, state, modules, environments, Azure example, CI/CD
48	Azure	Concepts, CLI, resource groups, VMs, storage, ACR, Container Apps, App Service, Key Vault, databases, RBAC, Azure OpenAI, cost, Bicep
49	Azure VM + Linux + Ollama	Azure CLI, VM, NSG, SSH, Linux basics, Ollama, SSH tunnel
50	Project Structure	Monorepo for Python microservices in Docker + React frontend: layers, uv workspace, proxy, config, Compose, tests, CI, deploy (with a runnable starter)
51	Project Templates	When to make a template, GitHub template repos, Copier (questions, placeholders, update), Cookiecutter, testing templates in CI

What's in This Repository¶

pocket-guide/
  README.md                  this page: the index of all guides
  guides/                    every guide, 00_big-picture.md ... 99_quick-reference.md
  examples/                  runnable mini-projects with tests (no API key needed)
    llm_basics/              one call, streaming, structured output
    tool_agent/              tool definitions and a manual agent loop
    docs_chatbot/            RAG chatbot: FastAPI, vector index, evals, Docker (the capstone code)
    mcp_server/              MCP server exposing document search
  templates/                 starting points to copy into your own projects
    fullstack-microservices/ Python services in Docker + React frontend + Nginx proxy
    service-template/        Copier template that adds a new service to the starter
  tools/                     scripts: doc checks, navigation, glossary, website build
  site_assets/               website stylesheet (extra.css)
  mkdocs.yml                 website configuration and page order
  requirements-docs.txt      packages needed to build the website
  lychee.toml                link checker settings
  .markdownlint-cli2.jsonc   Markdown style rules
  .github/workflows/         CI: doc checks, link check, examples, templates, website deploy

Folder	Details
guides/	Numbering groups, the parts every guide has, what is generated, how to add a guide
examples/	What each example shows, setup, running tests and demos
templates/	The microservices starter and the Copier service template, quick start
tools/	What each script checks or generates, building the website locally, CI workflows

Conventions¶

File names: guides/NN_topic.md (two-digit number, lowercase, hyphens), ordered from basic to advanced.
Each Introduction starts with Before you start: what to read first, the problem the tool solves, how it was done before, and an everyday analogy.
Each guide opens with an Introduction: what the tool is, why we use it, a mental model, key terms and where it fits with the other guides.
Each Introduction ends with an Official docs table: the tool's home page and key reference pages for the latest information.
Then a numbered Contents list; sections are numbered to match.
Section 0. Flags and Parameters breaks a sample command into its parts and explains every flag / parameter used in that guide.
Each section opens with a short summary: what it is, how it works, and when you would use it.
Previous / Next links at the top and bottom of every guide follow the reading order (generated by tools/build_nav.py).
Commands have a short comment on the right explaining what they do.
Most guides end with a Troubleshooting table of common errors and fixes.
Each guide shows a Last verified date and ends with a Try It exercise section.
Quality checks run in CI: tools/check_docs.py (structure, links between guides, anchors), markdownlint, a weekly external link check, the site build and the example tests.
Fast-moving tools (LLM models, agent frameworks, cloud services) change often: the concepts are stable, but check the Official docs links in each guide for exact current versions and names.