Auger/mvp_watchlist.md

Augor — Watchlist MVP

Goal

Replace the current bulk-feed home screen with a stock-watchlist-centric UX. Users add stocks they care about; the app fetches relevant news, runs LLM analysis, and surfaces per-stock growth signals with a confidence score.

Scope (in)

• Watchlist: add/remove stock tickers (e.g. AAPL, TSLA)
• Home screen: grid of stock cards showing ticker, last signal (growth/decline/neutral), probability, and a sparkline placeholder
• Stock dashboard: full signal list for that stock, rationale text, source articles
• Per-stock pipeline: filter aggregated feed by ticker relevance, run clustering + signal generation scoped to that ticker
• Settings: API key, feed list (unchanged), watchlist persisted via SharedPreferences

Scope (out — do later)

• Real-time or live price data
• Actual candlestick/OHLC charts (use placeholder or flat line for now)
• Backtesting
• Push notifications
• Backend / cloud processing

Data Model

class WatchedStock {
  final String ticker;       // e.g. "AAPL"
  final String companyName;  // e.g. "Apple Inc."
  EventSignal? latestSignal;
  List<EventSignal> signalHistory;
}

WatchedStock is persisted as JSON in SharedPreferences under key watched_stocks.

State

Add a WatchlistProvider (ChangeNotifier):

• List<WatchedStock> stocks
• addStock(ticker, companyName)
• removeStock(ticker)
• updateSignals(ticker, List<EventSignal>)
• save() / load()

Routing (go_router)

Path	Page
/	Home — stock card grid
/stock/:ticker	Stock dashboard
/settings	Settings (existing)
/watchlist/add	Add stock dialog or page

Home Screen

• GridView of StockCard widgets, one per watched stock
• Each card: ticker symbol (large), company name, signal badge (colored chip), probability percentage, placeholder mini-chart area
• Floating action button or top-right icon to add a stock
• Empty state: prompt to add a stock

Stock Dashboard

• Header: ticker + company name
• "Run Analysis" button — triggers the per-stock pipeline
• Status/progress text
• Placeholder chart (Container with grey background for now)
• Scrollable list of EventSignalCard widgets (reuse existing widget)
• Each card shows summary, signal, probability, confidence, rationale, source count

Per-Stock Pipeline

Scoped version of the existing pipeline:

1. Fetch all enabled feeds (same as now)
2. Filter FeedItem list to items relevant to the ticker — use company name + ticker as keywords alongside existing keyword list
3. Generate embeddings on filtered items only
4. Group by event (existing clustering)
5. Generate signals (existing SignalGenerator)
6. Store results in WatchlistProvider.updateSignals(ticker, signals)

The simplest first pass: add ticker + company name into the relevance keyword check. No separate embedding per ticker yet — just string matching on title/description.

Build Order

1. WatchedStock model + WatchlistProvider
2. Persist watchlist in SharedPreferences
3. New home screen with StockCard grid and add-stock flow
4. go_router routes updated
5. Stock dashboard page (static layout first)
6. Wire per-stock pipeline to dashboard "Run Analysis" button
7. Verbose progress logging (status string passed down to UI)
8. Placeholder chart widget

Definition of Done

• User can add AAPL and TSLA to watchlist
• Home screen shows both cards
• Tapping a card opens its dashboard
• Tapping "Run Analysis" runs the pipeline and populates signal cards
• At least one signal card renders with summary + probability
• No crashes on empty watchlist or missing API key