•
Install Haystack, wire a tiny Agent or pipeline with one tool, and learn production-minded NLP agents after work.
Haystack (by deepset) is a Python framework for LLM pipelines and agents—components you compose into RAG, search, and tool-using loops. Docs: docs.haystack.deepset.ai · product: haystack.deepset.ai.
After work it rewards people who like explicit pipelines: generators, retrievers, and agents as clear nodes—not a mysterious mega-prompt.
Working engineers and serious after-work Python learners who want mature NLP/pipeline building blocks. If you prefer “one class does everything,” try smolagents first; come back when you want composability.
Overview: Top 10 AI agent tools for after-work Python.
python -m venv .venv
source .venv/bin/activate
pip install -U haystack-ai
# add provider extras named in the Agent quickstart you follow
export OPENAI_API_KEY=... # or the provider shown in docs
From docs.haystack.deepset.ai, take the minimal Agent example. Prefer one that registers a single tool. Run: “What is 9 * 8? Use the calculator tool.”
If the sample exposes intermediate results, print them. Haystack’s lesson is seeing the graph, not only the chat string.
On paper: ToolInvoker → Agent (or whatever names the current docs use). Label inputs/outputs. Five minutes; huge clarity.
Add a unit_convert_km_to_miles(km: float) -> float tool. Ask for “Convert 5 km.” Verify the tool ran. No retriever yet.
Pipelines map well to weekday brains: small, named steps you can re-open tomorrow without rediscovering magic.
When components feel ordinary, strengthen Python with Intro to Python Programming so custom tools stay short and testable.