* Prune stale alerted signals. * Signals with 1 occurrence: pruned after 24h. * Signals with 2+ occurrences: pruned after 48h from last alert.

* Post HIGH confidence, short-horizon ideas as actionable Kalshi-style alerts. * @param {Array} ideas — LLM-generated ideas from the sweep

* Start the Discord bot. Connects to the gateway, registers slash commands, * and begins listening for interactions.

* Start polling for incoming messages/commands. * Call this once during server startup. * @param {number} intervalMs - polling interval (default

* Complete a prompt with system + user messages * @returns {{ text: string, usage: { inputTokens: number, outputTokens: number }, model: string }}

* Evaluate delta signals with LLM and send tiered alert if warranted. * Uses semantic dedup, rate limiting, and a much richer evaluation prompt.

* Register command handlers that the bot can respond to. * @param {string} command - e.g. '/status' * @param {Function} handler - async (args, m

* Stop the bot gracefully.

* Stop polling for incoming messages.