New Feature: Web Search Mode

What's New

Backboard now supports live web search as a first-class mode — allowing developers to combine real-time data with persistent memory and model routing in a single API call for more contextual, up-to-date AI responses.

How It Works

• Set web_search: true in your chat completion request.

• Backboard queries the web, ranks results by relevance, and injects them as context alongside your memory and RAG data.

• Results are automatically attributed with source URLs and freshness timestamps.

• Combine with memory for responses that are both grounded in real-time data and personalized.

Configuration Options

• Search depth: Choose between quick (3 results) or deep (10+ results) modes.

• Recency filter: Restrict results to last hour, day, week, month, or all time.

• Domain filtering: Allowlist or blocklist specific domains.

• Language preference: Target results in specific languages.

• Region targeting: Localize results by geographic region.

Use Cases

• Customer support: Answer questions about recent product updates or outages.

• Research agents: Gather real-time data alongside stored knowledge.

• News-aware assistants: Keep responses current without manual updates.

• Competitive intelligence: Monitor market changes in real-time.

Pricing

Web search is billed at $0.005 per search query. Results are cached for 1 hour to reduce costs on repeated similar queries.