Skip to content

ZeroAI

A one-page, AI-written briefing of the latest news for a stock. Type a ticker, and ZeroAI reads the newest headlines from your news sources and streams a structured summary into the page while the model writes it: sentiment, key points, risks and an outlook, with the sources beside it.

flowchart LR
    U([You: AAPL]) --> W[Web app<br/>React + Mantine]
    W <-->|"SSE stream"| A[API<br/>FastAPI]
    A -->|"RSS/Atom, concurrent"| N[(News sources<br/>Yahoo, Google, Nasdaq...)]
    A -->|"PydanticAI"| M{{LLM<br/>Ollama or OpenAI}}
    A --- D[(SQLite<br/>sources, model config, usage)]

What you get
Live streaming Status, reasoning, partial summary and final answer arrive over Server-Sent Events.
Your sources RSS/Atom feeds on the server's host allow-list, with optional {ticker} templates. Manage them in the UI.
Your model Local Ollama, an Ollama cloud model, an Ollama with an API token, or OpenAI. Switch in the UI.
Thinking mode Watch the model's reasoning live, or switch it off for speed and reliability.
Cost visibility Time and token usage after every run, aggregated per model.
Safe by design One model call at a time (semaphore); closing the tab aborts the run.

Where to go next

  1. Getting started: install, pick a model, first briefing.
  2. Models: Ollama (local, cloud, token) or OpenAI.
  1. Architecture: the pieces and how they connect.
  2. Request flow: one briefing, step by step.
  3. Streaming: the SSE events and the UI states.
  1. Troubleshooting: symptom to fix, with a decision chart.
  1. Configuration overview: what lives where.
  2. API reference and Frontend guide.
  3. Testing: strategy, commands and CI.