Skip to content
AI OSAI OS Docs

Local operations

Trend Finder collection and scoring run locally through the scoped scheduler job. Source and AI runtime readiness are independent: healthy public evidence can still use deterministic analysis, and a ready model runtime does not make an unreviewed source valid.

Run and inspect

Section titled “Run and inspect”
Terminal window
bun run scheduler:trend-finder:status
bun run scheduler:trend-finder:run
bun run scheduler:trend-finder:status

The first status can correctly report that no run exists yet. After a run, read the data, analysis, and source labels in the UI instead of assuming that every configured path succeeded.

Browser run control

Section titled “Browser run control”

The Trends run button is a local development control. It saves a dirty Creator Lens first, starts the scoped job, shows bounded progress, and refreshes the current browser data after completion.

The button remains usable when a provider is disabled or a source is missing. Those conditions belong in explicit readiness, degraded, blocked, or fallback states. Starting a second run while one is active returns an overlap state instead of launching another writer.

Account-backed analysis

Section titled “Account-backed analysis”

For the implemented local text provider, authenticate and test the intended model first:

Terminal window
bun run auth:openai -- status
bun run codex:smoke -- --model gpt-5.6-sol --json

Then run Trend Finder and inspect its own status. Runtime readiness alone is not end-to-end collector proof. See OpenAI subscription runtime.

Brief export

Section titled “Brief export”
Terminal window
bun run trend-finder:export-brief

The export uses the validated current payload and writes a local static Brief. It preserves the current provenance and excludes local diagnostics and source internals.

Historical backtests

Section titled “Historical backtests”
Terminal window
bun run trend-finder:backtest -- --sources producthunt-ai-launches --window-start 2026-01-01 --window-end 2026-01-07 --windows 4 --dry-run

Backtests are script-only. Full archives remain private local artifacts. Only an explicitly published bounded aggregate can reach browser data. Engine Replay remains a sanitized latest-run surface; it is not a historical backtest UI.

Before sharing a run or branch, use bun run runtime:check-private and make sure fixture, fallback, degraded, and live labels match what you claim.