How to Start Using LlamaIndex Workflows

LlamaIndex helps you connect LLMs to your data. Workflows are the event-driven way to build multi-step agents/RAG pipelines in Python. Start from docs.llamaindex.ai · product home llamaindex.ai.

After work, the winning move is one document + one workflow—not a company wiki mirror on night one.

Who this is for

Learners who want document Q&A or a small data agent after the day job. Comfortable opening files and writing functions; vector DB jargon can wait.

Overview: Top 10 AI agent tools for after-work Python.

Steps to start tonight

1. Install into a fresh venv

python -m venv .venv
source .venv/bin/activate
pip install -U llama-index
# add extras named in the Workflows / agents quickstart you open tonight

2. Set one LLM key

export OPENAI_API_KEY=...  # or the provider the quickstart uses

3. Drop exactly one markdown/text file in ./data

Example: data/notes.md with ten lines about your study plan. Resist the urge to dump a whole Drive export.

4. Run the shortest Workflow or agent quickstart

Copy the minimal Workflow sample from docs.llamaindex.ai (or the current “agents / workflows” guide). Index that one file, ask: “Summarize my notes in two bullets.” Print intermediate events if the sample supports it—that is the lesson.

5. Name the steps in NOTES.md

Write: which step loaded data, which called the LLM, where the answer returned. Workflows click when you can narrate the event chain.

Tiny exercise (20–30 minutes)

Add a second question path: “List any action items.” If none exist, the answer should say so. Do not add a second data source yet.

Pitfalls

• Ingesting gigabytes on night one. One file.
• Jumping straight to production vector DBs. Local/default storage from the quickstart is enough.
• Ignoring Workflow events and only reading the final string. You came for the orchestration model.
• Mixing three LlamaIndex major-version tutorials. Stay on the docs version you installed.
• Secrets in repo.

Why this fits after work

Document Q&A is a realistic after-work use case, and Workflows teach structure without a visual builder tax.

Next step

When file + workflow feels boring (in a good way), deepen Python with Intro to Python Programming so loaders and transforms read as ordinary code.