Avis Agent — Take-Home

A starting point for the Avis rental-servicing agent take-home. Start with BRIEF.md — it describes the task. This repo is a thin scaffold to save you setup time; it is not a solution.

What's here

.
├── BRIEF.md                          # the task — read this first
├── docs/
│   └── api-reference.md              # the Avis API your agent calls
├── data/
│   └── knowledge-base/articles.json  # Avis help-center articles (for RAG)
├── src/
│   ├── agent.py                      # a runnable hello-world agent (NOT a solution)
│   └── avis_client.py                # one worked example call to the Avis API
├── env.example
└── requirements.txt

Setup

1. Python 3.10+ required (the OpenAI Agents SDK needs it). Check first — on macOS the system python3 is often 3.9, which won't work:

   python3 --version            # need 3.10 or higher
   # if it's < 3.10, install a newer one (e.g. `brew install python@3.12`) and use that:
   python3 -m venv .venv && source .venv/bin/activate

2. Install dependencies:

   pip install -r requirements.txt

3. Configure your environment:

   cp env.example .env

   AVIS_API_URL is already filled in. Add your AVIS_API_KEY — grab it from this 1Password link: https://share.1password.com/s#yaM3mKXX9P8_BHxWxpaSzhqqe4YIq4p6AaUafkLJ9vY — plus your own LLM key: we highly recommend the OpenAI Agents SDK (set OPENAI_API_KEY) — the scaffold uses it — but you may use Google's ADK instead (set GOOGLE_API_KEY and adapt the scaffold).
4. Verify the API connection (looks up a sample reservation):

   python src/avis_client.py

5. Run the starter agent:

   python src/agent.py

Then build

Head to BRIEF.md. Build the RAG foundation and escalation logic first, then choose which workflow(s) to support and why. The design is yours — agent.py and avis_client.py just prove the wiring works; replace and extend them however you like.

When you submit, please update this README to cover: your design decisions, which workflow(s) you chose and why, what you cut, how to run your code, and where your logs are. Submit as a zip archive of this repo (uploaded, not a GitHub link) — exclude .venv/, __pycache__/, and any local secrets.