Skip to content

Getting started

From zero to your first briefing in about five minutes.

flowchart LR
    A[1. Install tools] --> B[2. Pick a model]
    B --> C[3. Start API]
    C --> D[4. Start web app]
    D --> E([5. Summarise AAPL])

1. Prerequisites

Tool Why Check
uv Python 3.14 and the backend environment uv --version
Node 22+ and corepack The web app (yarn 4) node -v then corepack enable
Ollama or an OpenAI key The model that writes the briefing ollama --version
just (optional) Short commands used below just --version

2. Pick a model

flowchart TD
    Q{What do you have?} -->|"Ollama installed,<br/>signed in"| C["Ollama cloud model<br/>gpt-oss:120b-cloud<br/>(default, fastest)"]
    Q -->|"Ollama, want everything<br/>on this machine"| L["Local model<br/>qwen3 via ollama pull"]
    Q -->|"Remote Ollama or<br/>ollama.com with a token"| T["Ollama + API token"]
    Q -->|"OpenAI key"| O["OpenAI<br/>gpt-4o-mini"]
ollama signin          # once; cloud models need no download

Nothing else to do: the default model is gpt-oss:120b-cloud.

Privacy

Cloud models send the prompt (public news headlines) to Ollama's cloud. Choose a local model if that is not acceptable.

ollama pull qwen3

Then, after the app is running, open Settings and set the model to qwen3:latest. Expect about 30 s per briefing; see Thinking mode.

Nothing to install. After the app is running, open Settings, choose OpenAI, and paste your API key.

See Models for every option, including an Ollama with an API token.

3. Install and run

just setup      # uv sync + yarn install
just api        # API on http://localhost:8000   (interactive docs at /docs)
just web        # UI  on http://localhost:5173
Without just 
cd backend  && uv sync && uv run uvicorn zeroai.asgi:app --reload
cd frontend && corepack enable && yarn install && yarn dev

4. Your first briefing

  1. Open http://localhost:5173.
  2. Type AAPL (or click a suggestion) and press Summarise.
  3. Watch the status line (⠹ Summarising… 4.2 s), the reasoning panel and the briefing fill in.
  4. When it ends you see the time and token counts. Settings shows your usage per model.

Try the API directly

curl localhost:8000/api/v1/stocks/AAPL/summary          # one JSON document
curl -N localhost:8000/api/v1/stocks/AAPL/summary/stream  # live SSE events

Something not working? Go to Troubleshooting.