Tech Stack and Tools

The full, opinionated 2026 stack for this course. Use it as your "what to install and why" reference.

TL;DR

Python 3.11+ | Streamlit | OpenAI / Gemini / Anthropic APIs
ChromaDB or pgvector | tiktoken | python-dotenv
LangChain (light) | Hugging Face (optional) | FastAPI (light intro)
GitHub | VS Code or Cursor | Colab or Replit for fast experiments
Streamlit Cloud | Hugging Face Spaces | Vercel | Railway

Languages and runtimes

  • Python 3.11+ is the default. Easy install via python.org or pyenv.
  • Node 20+ if you take the Next.js stretch. Install via nvm.

Environments

  • Local: VS Code (free) or Cursor (AI-first IDE).
  • Cloud notebooks: Google Colab (free with limits), Kaggle Notebooks, Replit.
  • Containers: Docker Desktop or Podman, optional.

Core Python libraries

LibraryPurposeInstall
openaiOpenAI API clientpip install openai
anthropicClaude API clientpip install anthropic
google-genaiGemini API clientpip install google-genai
python-dotenvload .env keyspip install python-dotenv
tiktokencount tokenspip install tiktoken
streamlitUI frameworkpip install streamlit
chromadbvector DBpip install chromadb
pypdfPDF parsingpip install pypdf
pymupdfbetter PDF parsingpip install pymupdf
numpyvector mathpip install numpy
tenacityretries with backoffpip install tenacity
pydanticdata validationpip install pydantic
fastapibackend API (light intro)pip install fastapi uvicorn
langchain / langchain_openaioptional orchestrationpip install langchain langchain-openai
sentence-transformersself-hosted embeddingspip install sentence-transformers
rank-bm25keyword searchpip install rank-bm25

LLM providers and free tiers

ProviderNotable free tier (check current limits)When to pick
OpenAITrial credit on new accountsDefault for this course
Google AI Studio (Gemini)Generous free tier on Flash modelsCheap experimentation, long context
Anthropic ClaudeTrial creditLong-context document analysis
Together AI / GroqFree tier with open modelsCheap inference, open-weights testing
Hugging Face InferenceFree tier on many modelsOpen-model experiments
CohereGenerous free tier on multilingual embeddingsMultilingual RAG

Tip: at the start of every project, set a hard spending limit in each provider dashboard. $5 is plenty for the whole course.

Embeddings

ModelProviderStrengths
text-embedding-3-smallOpenAICheap, balanced, the default
text-embedding-3-largeOpenAIHigher quality, larger dim
voyage-3Voyage AIOften top of MTEB
embed-multilingual-v3CohereMultilingual strength
bge-base-en-v1.5BAAI (open)Self-hostable, free
nomic-embed-text-v1.5Nomic (open)Self-hostable, long context

Vector databases

DBBest forSetup difficulty
ChromaDBPrototypes, local devEasy
FAISSResearch, in-memory speedEasy (library only)
pgvectorTeams already on Postgres or SupabaseEasy
QdrantProduction, hybrid searchMedium
WeaviateProduction, native hybridMedium
PineconeHosted, easy production scaleEasy (paid)
MilvusMassive scaleHard

Deployment platforms

PlatformApp typesCost
Streamlit CloudStreamlitFree tier
Hugging Face SpacesStreamlit, Gradio, DockerFree tier
VercelNext.js, Edge FunctionsGenerous free tier
RailwayFastAPI, workers, DBsPay-as-you-grow
Fly.ioContainers, long-runningFree tier with caveats
RenderWeb servicesFree tier with sleeps

Observability and monitoring

ToolPurpose
logging (stdlib)Local logs
Logtail / Better StackHosted log aggregation (free tier)
HeliconeLLM-specific observability proxy
LangSmithLangChain-native tracing
Phoenix (Arize)Open-source LLM tracing

Eval frameworks

ToolStrength
PromptfooSide-by-side prompt eval
RagasRAG-specific metrics
TruLensProduction LLM evals
OpenAI EvalsOpen-source eval harness
LangSmith EvalsLangChain-native

Setup commands

Create a project from zero:

mkdir my-ai-app && cd my-ai-app
python -m venv .venv
source .venv/bin/activate          # Mac/Linux
.venv\Scripts\activate            # Windows

pip install streamlit openai python-dotenv tiktoken chromadb pypdf tenacity pydantic

echo ".env" > .gitignore
echo ".venv/" >> .gitignore
echo "chroma_db/" >> .gitignore
echo "data/" >> .gitignore

git init

.env:

OPENAI_API_KEY=sk-...
GEMINI_API_KEY=...
ANTHROPIC_API_KEY=sk-ant-...

Verify:

python -c "from openai import OpenAI; print(OpenAI().models.list().data[0].id)"

Cost-aware defaults

  • Default model for prototypes: gpt-4o-mini (or gemini-2.5-flash)
  • Default embedding model: text-embedding-3-small
  • Default chunk size: 500 tokens with 80 overlap
  • Default top-K: 5
  • Default max_tokens for chat: 600 to 800
  • Default temperature: 0.4 for factual, 0.8 for creative

What we deliberately avoid in this course

  • Heavy ML training (no PyTorch deep dives at beginner level)
  • Custom transformer implementations
  • Manual fine-tuning runs (better in the intermediate course)
  • AWS / GCP / Azure deep provisioning (overkill for beginners)
  • Kubernetes (not for first apps)

Optional power-ups

  • Cursor: AI-pair-programming IDE; great for refactors.
  • Ollama: run small open-weight models on your laptop offline.
  • LM Studio: GUI for local models, good for evaluating Llama, Qwen, etc.
  • Supabase: Postgres + auth + storage + pgvector in one free tier.

SEO Notes

  • Primary keyword: "AI tech stack for beginners 2026"
  • Featured snippet target: the "TL;DR" block at the top and the "Setup commands" code block
  • Internal links: Modules 5, 6, 9, 10