Skip to content
Build, Evaluate & Secure an AI Agent

Before the event

Two to three weeks before

Section titled “Two to three weeks before”
  • Pick a format from Formats and timing
  • Do the whole workshop yourself, start to finish, on the model attendees will use
  • Find helpers. One per 10 to 15 attendees is ideal
  • Publish your event. Copy is ready on Promote your event
  • Check the venue: power for every seat, a projector, and wifi that can handle the room

One week before

Section titled “One week before”
  • Email attendees the Setup link and ask them to finish it at home, including the model download
  • Ask them to run python -m agentlab.check and reply or post in your chat if it fails
  • Brief your helpers. Ask them to read Troubleshooting and the facilitator notes
  • Decide on a backup for laptops that can’t run a model. Options: pair them with a neighbor, use Codespaces with a cloud provider, or run a shared Ollama server (see below)

The day before

Section titled “The day before”
  • Pull the latest version of the repository. Run pytest in the code folder to make sure the code works
  • Run python -m agentlab.check on your presenting laptop
  • Put the Ollama installers and the model on a few USB sticks for people who arrive unprepared (see below)
  • Prepare a short URL or QR code for the workshop site

On the day

Section titled “On the day”
  • Put the workshop URL and your event’s help channel on the screen while people arrive
  • Start with a setup check: everyone runs python -m agentlab.check. Helpers walk the room
  • Remind people that the solution folders exist. Falling behind is fine
  • Take photos for social media (ask permission first)

After the event

Section titled “After the event”
  • Share highlights with your event’s community
  • Open issues on GitHub for anything that confused people

Offline and low-bandwidth tips

Section titled “Offline and low-bandwidth tips”

Copy the model to a USB stick. Ollama stores models in ~/.ollama/models (on Windows, C:\Users\<you>\.ollama\models). Copy that folder from a laptop that has the model onto a USB stick, and copy it into the same place on an attendee’s laptop. Also bring the Ollama installers for macOS, Windows and Linux.

Run a shared model server. One strong laptop or desktop can serve the whole room. On that machine, start Ollama so it accepts connections from the network:

Terminal window
OLLAMA_HOST=0.0.0.0 ollama serve

Attendees then set LLM_BASE_URL=http://<that machine's IP address>:11434/v1 in their .env. A single machine handles a limited number of requests at once, so this works best for a small group or as a fallback for a few people.

Python packages offline. The workshop needs only httpx and python-dotenv. If you expect poor wifi, download them ahead of time with pip download -d wheels httpx python-dotenv setuptools and share the folder. Attendees install with pip install --no-index --find-links wheels -e .