No, beyond pointing to 911 and FEMA (1-800-621-3362). The system prompt explicitly refuses to coordinate rescue, recommend evacuation routes, or estimate damage. A visitor typing "my house is flooding right now" gets a direct instruction to call 911 and the FEMA hotline, plus a link to the FEMA mobile app, before any other content. That refusal is the safest default, and it protects both the visitor and the organization from liability around emergency advice the bot is not qualified to give.

Yes, because there is no shared model fine-tune to be rate-limited. SleekAI calls your chosen model provider (OpenAI, Anthropic, Google, or OpenRouter) on your API key, and each provider has its own enterprise capacity. Most relief nonprofits using GPT-4o-mini or Claude Haiku for the bulk of traffic see no degradation during appeal spikes. If your provider has rate limits you are worried about, the bot supports failover to a secondary model defined in the chatbot config.

Yes. The active appeal is typically marked with a custom field or taxonomy term, and the system prompt prioritizes that appeal post first. When the appeal closes or a new one opens, the next conversation reflects the change. The bot reads the appeal post for earmarked-giving copy, the situation summary, the partner list for that operation, and the volunteer role posts tied to the appeal's region taxonomy.

Yes, within the limits of what the role posts define. Each volunteer role typically lists region, required certifications, time commitment, and the next orientation date. The bot asks the visitor about location and any certifications, then narrows to roles they qualify for. If the bot cannot determine a clear match, it offers to capture the visitor's contact info so the volunteer coordinator can follow up rather than guessing at a role assignment.

Each shelter post can carry an accepted-supply list (what they need now) and a not-accepted list (what is full or off-policy). The bot quotes the accepted list for the closest open shelter and routes large in-kind offers like a truckload donation to the corporate giving contact or operations director, depending on how your contact directory is structured. Most relief orgs prefer cash to in-kind for logistics reasons, and the bot can explain that distinction when asked.

Yes. Browser language detection plus WPML or Polylang translations let the bot reply in Spanish for visitors who land in that language, and shelter, volunteer, and appeal pages stored in Spanish are used over machine translation. For US relief operations this matters in Florida, Texas, California, and Puerto Rico especially. Other languages (French, Haitian Creole, Tagalog) work the same way as long as your site has the translated content or you accept model translation quality.

Yours. SleekAI is BYO API key. For a typical disaster relief site with normal traffic and a few appeal spikes per year, monthly model spend usually lands between $30 and $250 depending on appeal volume. During a major appeal, expect $100 to $500 per week of additional spend if you stay on GPT-4o-mini or Claude Haiku. Switching to a smaller model during the spike is a valid cost-management lever, with the tradeoff of slightly less nuanced replies on edge cases.

No, by default. Display conditions in SleekAI let you exclude the chatbot from memorial pages, tribute pages, and pages honoring victims, where a chatbot would be tonally wrong. The same applies to staff memorial pages and pages about volunteers lost in operations. The bot widget simply does not load on those URLs, and a visitor who needs help is steered to the general donation or shelter pages instead.