Call a hosted model and read the bill

1. 01Put your key and a model name in the environment

   Keep the key in your shell or a .env file that git ignores. Never paste it into a script.

   test -n "$OPENROUTER_API_KEY" && test -n "$MODEL" && echo ready

   You pass when: It prints ready. MODEL holds a model id you picked from the OpenRouter model list (filter by price: free).

   Check step 1

2. 02Write chat.py: one system message, one user message, print the answer and the usage

   Use the openai package with base_url set to https://openrouter.ai/api/v1. The same script talks to Ollama if you point base_url at http://localhost:11434/v1.

   uv run --with openai chat.py "Explain a context window to a ten year old"

   You pass when: An answer prints, then one line with prompt tokens and completion tokens.

   Check step 2 chat.py

3. 03Add a --chat mode that resends the whole conversation each turn

   The model remembers nothing. The conversation is a list you keep and send again, which is why long chats cost more per turn.

   uv run --with openai chat.py --chat

   You pass when: Tell it your name in turn one and ask for it in turn two: it answers correctly, and the prompt token count grows every turn.

   Check step 3 chat.py

4. 04Print the cost of each call from the model's listed prices

   uv run --with openai chat.py "Give me three names for a parcel company"

   You pass when: A cost line prints in dollars. On a free model it is 0, so also print what the same call would cost on one paid model you name.

   Check step 4 chat.py

The check runs on the computer, not on your word: a step passes when its command exits with 0 there.