Steadier market data, and fewer stuck scheduled runs

A handful of fixes, mostly for assistants that watch markets or run on a schedule.

Market data. Stock prices and market overviews could occasionally show up blank instead of a number — usually during a trading session that hadn’t fully settled yet. That’s fixed, with a real fallback instead of a blank. Sector performance for the 1-day view had a similar problem and was returning nothing at all for any sector; it now shows real numbers. When a price or sector genuinely isn’t available, you’ll see that explicitly instead of a silent gap.

Prediction tracking. Assistants that track a lot of predictions can now filter and page through them instead of always pulling back the whole collection at once. This is headroom for the future rather than a fix for anything going wrong today — collections have stayed well within the old approach’s limits — but it removes a truncation risk as they grow.

Scheduled runs. A rare naming collision — your workspace happening to have an agent with the same name as one in a different workspace — could cause that agent’s saved state or canvas updates to fail on a scheduled run. Nothing about your data was ever at risk; the collision was in an internal identifier, not your records. It’s fixed either way.

Asking for an integration Ward doesn’t have. If you ask Ward to connect to something it doesn’t support, it now consistently tells you so directly, rather than sometimes filing it away as an open request. Ward doesn’t build one-off integrations for individual asks — if what you need exists as an MCP server, ask and it’ll look for it and offer to connect it.